{
  "schema_version": 1,
  "title": "Tool Selection Lab: 100 tools, 329 tasks, 9 strategies",
  "methodology": {
    "source": "article-analysis.zip / tokens_by_model_strategy.csv, results_by_task_type.csv, multistep_breakdown.csv; steplimit-16-analysis (16-step reruns)",
    "evaluations": 8883,
    "models": 3,
    "strategies": 9,
    "tasks_per_model_strategy": 329,
    "repetitions": 1,
    "notes": [
      "Model input/output tokens use canonical final executions, not superseded reruns.",
      "Aggregated success values combine equal task counts across three models.",
      "Tokens per completed task includes tokens used on failed evaluations in the numerator.",
      "Tokens per completed task are token counts, not dollar cost.",
      "Single-tool and multi-step groups exclude clarification and unsupported tasks, which are separate groups.",
      "Some API requests had missing usage; observed token totals are lower bounds.",
      "Chart data contains no secret keys or raw user prompts.",
      "Execution budget: up to 16 model steps per task, including tool discovery calls. Evaluations that finished within 8 steps were retained; evaluations that exhausted that budget (509 of 8,883) were rerun from scratch with 16 steps, and 508 reruns replace their originals. One evaluation (nemotron-3-super, all, hE-23) keeps its 8-step result because its rerun ended in an API error.",
      "357 of 8,883 evaluations still ran out of steps in the final results (gpt-oss-20b 0, nemotron-3.5-lightning 5, nemotron-3-super 352) and are counted as not completed.",
      "Token totals cover the final retained evaluations, failed ones included, and exclude superseded attempts. They are lower bounds and not the total resources spent on the experiment."
    ],
    "max_steps": 16
  },
  "models": [
    {
      "key": "all-models",
      "label": "All 3 models",
      "aggregate": true
    },
    {
      "key": "gpt-oss-20b",
      "label": "GPT-OSS 20B",
      "aggregate": false
    },
    {
      "key": "nemotron-3-super-120b-a12b",
      "label": "Nemotron Super",
      "aggregate": false
    },
    {
      "key": "nemotron-3.5-lightning-30b-a3b",
      "label": "Nemotron Lightning",
      "aggregate": false
    }
  ],
  "strategies": [
    {
      "key": "all",
      "label": "All 100 tools",
      "type": "expose"
    },
    {
      "key": "category",
      "label": "LLM category",
      "type": "route"
    },
    {
      "key": "category-laya",
      "label": "Laya category",
      "type": "route"
    },
    {
      "key": "semantic@10",
      "label": "spaCy top-10",
      "type": "fixed"
    },
    {
      "key": "keyword@10",
      "label": "BM25 top-10",
      "type": "fixed"
    },
    {
      "key": "semantic-nim@10",
      "label": "NIM semantic top-10",
      "type": "fixed"
    },
    {
      "key": "agent-search",
      "label": "Agent search (semantic)",
      "type": "search"
    },
    {
      "key": "agent-search-kw",
      "label": "Agent search (BM25)",
      "type": "search"
    },
    {
      "key": "hybrid",
      "label": "Hybrid",
      "type": "hybrid"
    }
  ],
  "metrics": {
    "success_pct": "percentage of cases completed correctly",
    "tokens_per_completed_task": "total model input + output tokens divided by completed tasks",
    "input_tokens_total": "model input tokens across canonical tasks",
    "output_tokens_total": "model output tokens across canonical tasks"
  },
  "overall": [
    {
      "model": "all-models",
      "strategy": "all",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 805,
      "success_pct": 81.6,
      "input_tokens_total": 27478225,
      "output_tokens_total": 382613,
      "tokens_per_completed_task": 34610.0,
      "input_tokens_median": 20863,
      "output_tokens_median": 254,
      "model_api_requests_total": 4133,
      "successful_model_calls_total": 3141,
      "model_turns_mean": 3.2,
      "business_tool_calls_total": 2213,
      "search_tools_calls_total": 0,
      "step_limit_failures": 39,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 100,
      "total_model_tokens": 27860838
    },
    {
      "model": "all-models",
      "strategy": "category",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 631,
      "success_pct": 63.9,
      "input_tokens_total": 4287817,
      "output_tokens_total": 703015,
      "tokens_per_completed_task": 7909.0,
      "input_tokens_median": 3185,
      "output_tokens_median": 528,
      "model_api_requests_total": 4184,
      "successful_model_calls_total": 3354,
      "model_turns_mean": 2.4,
      "business_tool_calls_total": 1531,
      "search_tools_calls_total": 0,
      "step_limit_failures": 25,
      "routing_errors": 119,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 4990832
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 604,
      "success_pct": 61.2,
      "input_tokens_total": 5339966,
      "output_tokens_total": 371257,
      "tokens_per_completed_task": 9456.0,
      "input_tokens_median": 4106,
      "output_tokens_median": 252,
      "model_api_requests_total": 3141,
      "successful_model_calls_total": 2442,
      "model_turns_mean": 2.48,
      "business_tool_calls_total": 1484,
      "search_tools_calls_total": 0,
      "step_limit_failures": 24,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 20,
      "total_model_tokens": 5711223
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 461,
      "success_pct": 46.7,
      "input_tokens_total": 2960056,
      "output_tokens_total": 371784,
      "tokens_per_completed_task": 7227.0,
      "input_tokens_median": 1545,
      "output_tokens_median": 236,
      "model_api_requests_total": 2809,
      "successful_model_calls_total": 2116,
      "model_turns_mean": 2.14,
      "business_tool_calls_total": 1164,
      "search_tools_calls_total": 0,
      "step_limit_failures": 35,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 3331840
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 653,
      "success_pct": 66.2,
      "input_tokens_total": 3665987,
      "output_tokens_total": 382465,
      "tokens_per_completed_task": 6200.0,
      "input_tokens_median": 2766,
      "output_tokens_median": 229,
      "model_api_requests_total": 3523,
      "successful_model_calls_total": 2694,
      "model_turns_mean": 2.74,
      "business_tool_calls_total": 1760,
      "search_tools_calls_total": 0,
      "step_limit_failures": 42,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 4048452
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 721,
      "success_pct": 73.0,
      "input_tokens_total": 3872069,
      "output_tokens_total": 362988,
      "tokens_per_completed_task": 5874.0,
      "input_tokens_median": 2827,
      "output_tokens_median": 230,
      "model_api_requests_total": 3706,
      "successful_model_calls_total": 2835,
      "model_turns_mean": 2.87,
      "business_tool_calls_total": 1888,
      "search_tools_calls_total": 0,
      "step_limit_failures": 43,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 4235057
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 714,
      "success_pct": 72.3,
      "input_tokens_total": 6553055,
      "output_tokens_total": 573830,
      "tokens_per_completed_task": 9982.0,
      "input_tokens_median": 3398,
      "output_tokens_median": 362,
      "model_api_requests_total": 5859,
      "successful_model_calls_total": 4569,
      "model_turns_mean": 4.62,
      "business_tool_calls_total": 1902,
      "search_tools_calls_total": 1725,
      "step_limit_failures": 48,
      "routing_errors": 2,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 7126885
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 705,
      "success_pct": 71.4,
      "input_tokens_total": 6956766,
      "output_tokens_total": 610680,
      "tokens_per_completed_task": 10734.0,
      "input_tokens_median": 3500,
      "output_tokens_median": 388,
      "model_api_requests_total": 6016,
      "successful_model_calls_total": 4741,
      "model_turns_mean": 4.8,
      "business_tool_calls_total": 1940,
      "search_tools_calls_total": 1862,
      "step_limit_failures": 51,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 7567446
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "jobs": 987,
      "unique_tasks": 329,
      "completed": 748,
      "success_pct": 75.8,
      "input_tokens_total": 6166017,
      "output_tokens_total": 458191,
      "tokens_per_completed_task": 8856.0,
      "input_tokens_median": 3343,
      "output_tokens_median": 266,
      "model_api_requests_total": 4530,
      "successful_model_calls_total": 3492,
      "model_turns_mean": 3.53,
      "business_tool_calls_total": 2031,
      "search_tools_calls_total": 521,
      "step_limit_failures": 50,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 6624208
    }
  ],
  "by_model": [
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 271,
      "success_pct": 82.4,
      "input_tokens_total": 2651870,
      "output_tokens_total": 78799,
      "tokens_per_completed_task": 10076.0,
      "input_tokens_median": 6575,
      "output_tokens_median": 179,
      "model_api_requests_total": 797,
      "successful_model_calls_total": 797,
      "model_turns_mean": 2.42,
      "business_tool_calls_total": 470,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 100,
      "total_model_tokens": 2730669
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 230,
      "success_pct": 69.9,
      "input_tokens_total": 705013,
      "output_tokens_total": 123318,
      "tokens_per_completed_task": 3601.0,
      "input_tokens_median": 1646,
      "output_tokens_median": 298,
      "model_api_requests_total": 1051,
      "successful_model_calls_total": 1050,
      "model_turns_mean": 2.19,
      "business_tool_calls_total": 395,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 3,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 828331
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 201,
      "success_pct": 61.1,
      "input_tokens_total": 645213,
      "output_tokens_total": 68227,
      "tokens_per_completed_task": 3549.0,
      "input_tokens_median": 1899,
      "output_tokens_median": 152,
      "model_api_requests_total": 653,
      "successful_model_calls_total": 653,
      "model_turns_mean": 1.98,
      "business_tool_calls_total": 324,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 20,
      "total_model_tokens": 713440
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 151,
      "success_pct": 45.9,
      "input_tokens_total": 328280,
      "output_tokens_total": 55796,
      "tokens_per_completed_task": 2544.0,
      "input_tokens_median": 667,
      "output_tokens_median": 140,
      "model_api_requests_total": 495,
      "successful_model_calls_total": 494,
      "model_turns_mean": 1.5,
      "business_tool_calls_total": 165,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 384076
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 217,
      "success_pct": 66.0,
      "input_tokens_total": 432071,
      "output_tokens_total": 67789,
      "tokens_per_completed_task": 2304.0,
      "input_tokens_median": 1280,
      "output_tokens_median": 160,
      "model_api_requests_total": 646,
      "successful_model_calls_total": 646,
      "model_turns_mean": 1.96,
      "business_tool_calls_total": 317,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 499860
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 242,
      "success_pct": 73.6,
      "input_tokens_total": 458711,
      "output_tokens_total": 67471,
      "tokens_per_completed_task": 2174.0,
      "input_tokens_median": 1295,
      "output_tokens_median": 155,
      "model_api_requests_total": 683,
      "successful_model_calls_total": 682,
      "model_turns_mean": 2.07,
      "business_tool_calls_total": 353,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 526182
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 217,
      "success_pct": 66.0,
      "input_tokens_total": 1008393,
      "output_tokens_total": 131603,
      "tokens_per_completed_task": 5253.0,
      "input_tokens_median": 2089,
      "output_tokens_median": 274,
      "model_api_requests_total": 1206,
      "successful_model_calls_total": 1206,
      "model_turns_mean": 3.66,
      "business_tool_calls_total": 374,
      "search_tools_calls_total": 503,
      "step_limit_failures": 0,
      "routing_errors": 2,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 1139996
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 222,
      "success_pct": 67.5,
      "input_tokens_total": 1140272,
      "output_tokens_total": 145191,
      "tokens_per_completed_task": 5790.0,
      "input_tokens_median": 2121,
      "output_tokens_median": 289,
      "model_api_requests_total": 1291,
      "successful_model_calls_total": 1290,
      "model_turns_mean": 3.92,
      "business_tool_calls_total": 390,
      "search_tools_calls_total": 571,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 1285463
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 254,
      "success_pct": 77.2,
      "input_tokens_total": 734794,
      "output_tokens_total": 91047,
      "tokens_per_completed_task": 3251.0,
      "input_tokens_median": 1582,
      "output_tokens_median": 180,
      "model_api_requests_total": 837,
      "successful_model_calls_total": 836,
      "model_turns_mean": 2.54,
      "business_tool_calls_total": 406,
      "search_tools_calls_total": 101,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 825841
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 276,
      "success_pct": 83.9,
      "input_tokens_total": 9367650,
      "output_tokens_total": 110678,
      "tokens_per_completed_task": 34342.0,
      "input_tokens_median": 20915,
      "output_tokens_median": 254,
      "model_api_requests_total": 1112,
      "successful_model_calls_total": 892,
      "model_turns_mean": 2.71,
      "business_tool_calls_total": 573,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 100,
      "total_model_tokens": 9478328
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 225,
      "success_pct": 68.4,
      "input_tokens_total": 1538826,
      "output_tokens_total": 381910,
      "tokens_per_completed_task": 8537.0,
      "input_tokens_median": 3352,
      "output_tokens_median": 1132,
      "model_api_requests_total": 1271,
      "successful_model_calls_total": 1026,
      "model_turns_mean": 2.12,
      "business_tool_calls_total": 410,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 40,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 1920736
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 204,
      "success_pct": 62.0,
      "input_tokens_total": 1825484,
      "output_tokens_total": 118387,
      "tokens_per_completed_task": 9529.0,
      "input_tokens_median": 5068,
      "output_tokens_median": 260,
      "model_api_requests_total": 912,
      "successful_model_calls_total": 715,
      "model_turns_mean": 2.17,
      "business_tool_calls_total": 388,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 20,
      "total_model_tokens": 1943871
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 162,
      "success_pct": 49.2,
      "input_tokens_total": 851961,
      "output_tokens_total": 109345,
      "tokens_per_completed_task": 5934.0,
      "input_tokens_median": 2671,
      "output_tokens_median": 253,
      "model_api_requests_total": 711,
      "successful_model_calls_total": 558,
      "model_turns_mean": 1.7,
      "business_tool_calls_total": 229,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 961306
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 229,
      "success_pct": 69.6,
      "input_tokens_total": 1099131,
      "output_tokens_total": 105365,
      "tokens_per_completed_task": 5260.0,
      "input_tokens_median": 2988,
      "output_tokens_median": 223,
      "model_api_requests_total": 952,
      "successful_model_calls_total": 738,
      "model_turns_mean": 2.24,
      "business_tool_calls_total": 409,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 1204496
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 252,
      "success_pct": 76.6,
      "input_tokens_total": 1152491,
      "output_tokens_total": 102456,
      "tokens_per_completed_task": 4980.0,
      "input_tokens_median": 2978,
      "output_tokens_median": 223,
      "model_api_requests_total": 951,
      "successful_model_calls_total": 775,
      "model_turns_mean": 2.36,
      "business_tool_calls_total": 447,
      "search_tools_calls_total": 0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 1254947
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 257,
      "success_pct": 78.1,
      "input_tokens_total": 2063679,
      "output_tokens_total": 171352,
      "tokens_per_completed_task": 8697.0,
      "input_tokens_median": 3585,
      "output_tokens_median": 362,
      "model_api_requests_total": 1761,
      "successful_model_calls_total": 1379,
      "model_turns_mean": 4.19,
      "business_tool_calls_total": 497,
      "search_tools_calls_total": 555,
      "step_limit_failures": 2,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 2235031
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 244,
      "success_pct": 74.2,
      "input_tokens_total": 2200954,
      "output_tokens_total": 189455,
      "tokens_per_completed_task": 9797.0,
      "input_tokens_median": 3578,
      "output_tokens_median": 367,
      "model_api_requests_total": 1751,
      "successful_model_calls_total": 1407,
      "model_turns_mean": 4.29,
      "business_tool_calls_total": 498,
      "search_tools_calls_total": 587,
      "step_limit_failures": 3,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 2390409
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 254,
      "success_pct": 77.2,
      "input_tokens_total": 1875459,
      "output_tokens_total": 133389,
      "tokens_per_completed_task": 7909.0,
      "input_tokens_median": 3477,
      "output_tokens_median": 256,
      "model_api_requests_total": 1203,
      "successful_model_calls_total": 975,
      "model_turns_mean": 2.96,
      "business_tool_calls_total": 497,
      "search_tools_calls_total": 156,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 2008848
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 258,
      "success_pct": 78.4,
      "input_tokens_total": 15458705,
      "output_tokens_total": 193136,
      "tokens_per_completed_task": 60666.0,
      "input_tokens_median": 31449,
      "output_tokens_median": 319,
      "model_api_requests_total": 2224,
      "successful_model_calls_total": 1452,
      "model_turns_mean": 4.44,
      "business_tool_calls_total": 1170,
      "search_tools_calls_total": 0,
      "step_limit_failures": 39,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 100,
      "total_model_tokens": 15651841
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 176,
      "success_pct": 53.5,
      "input_tokens_total": 2043978,
      "output_tokens_total": 197787,
      "tokens_per_completed_task": 12737.0,
      "input_tokens_median": 3325,
      "output_tokens_median": 398,
      "model_api_requests_total": 1862,
      "successful_model_calls_total": 1278,
      "model_turns_mean": 2.9,
      "business_tool_calls_total": 726,
      "search_tools_calls_total": 0,
      "step_limit_failures": 25,
      "routing_errors": 76,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 2241765
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 199,
      "success_pct": 60.5,
      "input_tokens_total": 2869269,
      "output_tokens_total": 184643,
      "tokens_per_completed_task": 15346.0,
      "input_tokens_median": 5116,
      "output_tokens_median": 389,
      "model_api_requests_total": 1576,
      "successful_model_calls_total": 1074,
      "model_turns_mean": 3.27,
      "business_tool_calls_total": 772,
      "search_tools_calls_total": 0,
      "step_limit_failures": 24,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 20,
      "total_model_tokens": 3053912
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 148,
      "success_pct": 45.0,
      "input_tokens_total": 1779815,
      "output_tokens_total": 206643,
      "tokens_per_completed_task": 13422.0,
      "input_tokens_median": 2995,
      "output_tokens_median": 406,
      "model_api_requests_total": 1603,
      "successful_model_calls_total": 1064,
      "model_turns_mean": 3.23,
      "business_tool_calls_total": 770,
      "search_tools_calls_total": 0,
      "step_limit_failures": 35,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 1986458
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 207,
      "success_pct": 62.9,
      "input_tokens_total": 2134785,
      "output_tokens_total": 209311,
      "tokens_per_completed_task": 11324.0,
      "input_tokens_median": 3107,
      "output_tokens_median": 385,
      "model_api_requests_total": 1925,
      "successful_model_calls_total": 1310,
      "model_turns_mean": 4.02,
      "business_tool_calls_total": 1034,
      "search_tools_calls_total": 0,
      "step_limit_failures": 42,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 2344096
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 227,
      "success_pct": 69.0,
      "input_tokens_total": 2260867,
      "output_tokens_total": 193061,
      "tokens_per_completed_task": 10810.0,
      "input_tokens_median": 3082,
      "output_tokens_median": 335,
      "model_api_requests_total": 2072,
      "successful_model_calls_total": 1378,
      "model_turns_mean": 4.18,
      "business_tool_calls_total": 1088,
      "search_tools_calls_total": 0,
      "step_limit_failures": 43,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 2453928
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 240,
      "success_pct": 72.9,
      "input_tokens_total": 3480983,
      "output_tokens_total": 270875,
      "tokens_per_completed_task": 15633.0,
      "input_tokens_median": 4809,
      "output_tokens_median": 504,
      "model_api_requests_total": 2892,
      "successful_model_calls_total": 1984,
      "model_turns_mean": 6.02,
      "business_tool_calls_total": 1031,
      "search_tools_calls_total": 667,
      "step_limit_failures": 46,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 3751858
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 239,
      "success_pct": 72.6,
      "input_tokens_total": 3615540,
      "output_tokens_total": 276034,
      "tokens_per_completed_task": 16283.0,
      "input_tokens_median": 5027,
      "output_tokens_median": 540,
      "model_api_requests_total": 2974,
      "successful_model_calls_total": 2044,
      "model_turns_mean": 6.19,
      "business_tool_calls_total": 1052,
      "search_tools_calls_total": 704,
      "step_limit_failures": 48,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 5,
      "total_model_tokens": 3891574
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "jobs": 329,
      "unique_tasks": 329,
      "completed": 240,
      "success_pct": 72.9,
      "input_tokens_total": 3555764,
      "output_tokens_total": 233755,
      "tokens_per_completed_task": 15790.0,
      "input_tokens_median": 5090,
      "output_tokens_median": 402,
      "model_api_requests_total": 2490,
      "successful_model_calls_total": 1681,
      "model_turns_mean": 5.08,
      "business_tool_calls_total": 1128,
      "search_tools_calls_total": 264,
      "step_limit_failures": 50,
      "routing_errors": 0,
      "api_errors": 0,
      "median_shortlist": 10,
      "total_model_tokens": 3789519
    }
  ],
  "by_task_group": [
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "group": "single_tool",
      "count": 185,
      "completed": 158,
      "success_pct": 85.4,
      "input_tokens_mean": 7046.4,
      "output_tokens_mean": 176.5,
      "tokens_per_completed_task": 8457.0,
      "model_turns_mean": 2.14,
      "business_tool_calls_mean": 1.14,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "group": "multi_step",
      "count": 100,
      "completed": 69,
      "success_pct": 69.0,
      "input_tokens_mean": 11655.9,
      "output_tokens_mean": 378.8,
      "tokens_per_completed_task": 17442.0,
      "model_turns_mean": 3.45,
      "business_tool_calls_mean": 2.47,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "group": "ambiguous",
      "count": 20,
      "completed": 20,
      "success_pct": 100.0,
      "input_tokens_mean": 3744.2,
      "output_tokens_mean": 136.7,
      "tokens_per_completed_task": 3881.0,
      "model_turns_mean": 1.15,
      "business_tool_calls_mean": 0.15,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 4003.2,
      "output_tokens_mean": 235.5,
      "tokens_per_completed_task": 4239.0,
      "model_turns_mean": 1.23,
      "business_tool_calls_mean": 0.23,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 9877.0,
      "output_tokens_mean": 174.0,
      "tokens_per_completed_task": 10051.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "group": "single_tool",
      "count": 185,
      "completed": 133,
      "success_pct": 71.9,
      "input_tokens_mean": 1643.1,
      "output_tokens_mean": 288.8,
      "tokens_per_completed_task": 2687.0,
      "model_turns_mean": 1.91,
      "business_tool_calls_mean": 0.93,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 3,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "group": "multi_step",
      "count": 100,
      "completed": 54,
      "success_pct": 54.0,
      "input_tokens_mean": 3486.3,
      "output_tokens_mean": 548.0,
      "tokens_per_completed_task": 7471.0,
      "model_turns_mean": 3.15,
      "business_tool_calls_mean": 2.15,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 1128.8,
      "output_tokens_mean": 333.3,
      "tokens_per_completed_task": 1539.0,
      "model_turns_mean": 1.15,
      "business_tool_calls_mean": 0.15,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 1183.9,
      "output_tokens_mean": 364.3,
      "tokens_per_completed_task": 1548.0,
      "model_turns_mean": 1.09,
      "business_tool_calls_mean": 0.09,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 1891.0,
      "output_tokens_mean": 209.5,
      "tokens_per_completed_task": 2100.0,
      "model_turns_mean": 2.5,
      "business_tool_calls_mean": 1.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "group": "single_tool",
      "count": 185,
      "completed": 122,
      "success_pct": 65.9,
      "input_tokens_mean": 1750.1,
      "output_tokens_mean": 156.0,
      "tokens_per_completed_task": 2890.0,
      "model_turns_mean": 1.83,
      "business_tool_calls_mean": 0.83,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "group": "multi_step",
      "count": 100,
      "completed": 41,
      "success_pct": 41.0,
      "input_tokens_mean": 2802.9,
      "output_tokens_mean": 338.0,
      "tokens_per_completed_task": 7661.0,
      "model_turns_mean": 2.69,
      "business_tool_calls_mean": 1.69,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "group": "ambiguous",
      "count": 20,
      "completed": 15,
      "success_pct": 75.0,
      "input_tokens_mean": 914.8,
      "output_tokens_mean": 116.8,
      "tokens_per_completed_task": 1375.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 913.7,
      "output_tokens_mean": 136.3,
      "tokens_per_completed_task": 1050.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 1379.5,
      "output_tokens_mean": 113.0,
      "tokens_per_completed_task": 2985.0,
      "model_turns_mean": 1.5,
      "business_tool_calls_mean": 0.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "group": "single_tool",
      "count": 185,
      "completed": 94,
      "success_pct": 50.8,
      "input_tokens_mean": 1048.4,
      "output_tokens_mean": 135.5,
      "tokens_per_completed_task": 2330.0,
      "model_turns_mean": 1.58,
      "business_tool_calls_mean": 0.58,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "group": "multi_step",
      "count": 100,
      "completed": 16,
      "success_pct": 16.0,
      "input_tokens_mean": 1053.7,
      "output_tokens_mean": 249.3,
      "tokens_per_completed_task": 8144.0,
      "model_turns_mean": 1.55,
      "business_tool_calls_mean": 0.55,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 633.5,
      "output_tokens_mean": 112.8,
      "tokens_per_completed_task": 829.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 651.6,
      "output_tokens_mean": 149.8,
      "tokens_per_completed_task": 801.0,
      "model_turns_mean": 1.05,
      "business_tool_calls_mean": 0.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 975.0,
      "output_tokens_mean": 123.5,
      "tokens_per_completed_task": 2197.0,
      "model_turns_mean": 1.5,
      "business_tool_calls_mean": 0.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "group": "single_tool",
      "count": 185,
      "completed": 131,
      "success_pct": 70.8,
      "input_tokens_mean": 1191.6,
      "output_tokens_mean": 146.5,
      "tokens_per_completed_task": 1890.0,
      "model_turns_mean": 1.84,
      "business_tool_calls_mean": 0.84,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "group": "multi_step",
      "count": 100,
      "completed": 43,
      "success_pct": 43.0,
      "input_tokens_mean": 1821.2,
      "output_tokens_mean": 335.0,
      "tokens_per_completed_task": 5014.0,
      "model_turns_mean": 2.57,
      "business_tool_calls_mean": 1.57,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 611.1,
      "output_tokens_mean": 121.1,
      "tokens_per_completed_task": 771.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 617.9,
      "output_tokens_mean": 203.1,
      "tokens_per_completed_task": 821.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 1843.5,
      "output_tokens_mean": 152.0,
      "tokens_per_completed_task": 1996.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "group": "single_tool",
      "count": 185,
      "completed": 144,
      "success_pct": 77.8,
      "input_tokens_mean": 1241.8,
      "output_tokens_mean": 145.8,
      "tokens_per_completed_task": 1783.0,
      "model_turns_mean": 1.91,
      "business_tool_calls_mean": 0.91,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "group": "multi_step",
      "count": 100,
      "completed": 55,
      "success_pct": 55.0,
      "input_tokens_mean": 1968.5,
      "output_tokens_mean": 331.9,
      "tokens_per_completed_task": 4183.0,
      "model_turns_mean": 2.76,
      "business_tool_calls_mean": 1.76,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 650.6,
      "output_tokens_mean": 113.8,
      "tokens_per_completed_task": 805.0,
      "model_turns_mean": 1.05,
      "business_tool_calls_mean": 0.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 698.1,
      "output_tokens_mean": 214.3,
      "tokens_per_completed_task": 912.0,
      "model_turns_mean": 1.14,
      "business_tool_calls_mean": 0.14,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 1875.0,
      "output_tokens_mean": 155.5,
      "tokens_per_completed_task": 2030.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "group": "single_tool",
      "count": 185,
      "completed": 134,
      "success_pct": 72.4,
      "input_tokens_mean": 2333.2,
      "output_tokens_mean": 272.7,
      "tokens_per_completed_task": 3598.0,
      "model_turns_mean": 3.17,
      "business_tool_calls_mean": 0.95,
      "search_tools_calls_mean": 1.22,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "group": "multi_step",
      "count": 100,
      "completed": 43,
      "success_pct": 43.0,
      "input_tokens_mean": 4567.4,
      "output_tokens_mean": 602.1,
      "tokens_per_completed_task": 12022.0,
      "model_turns_mean": 4.79,
      "business_tool_calls_mean": 1.89,
      "search_tools_calls_mean": 1.92,
      "step_limit_failures": 0,
      "routing_errors": 2,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 1132.8,
      "output_tokens_mean": 186.5,
      "tokens_per_completed_task": 1389.0,
      "model_turns_mean": 1.75,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.75,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "group": "unsupported",
      "count": 22,
      "completed": 19,
      "success_pct": 86.4,
      "input_tokens_mean": 4167.1,
      "output_tokens_mean": 760.5,
      "tokens_per_completed_task": 5706.0,
      "model_turns_mean": 4.36,
      "business_tool_calls_mean": 0.27,
      "search_tools_calls_mean": 3.09,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 2835.5,
      "output_tokens_mean": 241.0,
      "tokens_per_completed_task": 3076.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "group": "single_tool",
      "count": 185,
      "completed": 140,
      "success_pct": 75.7,
      "input_tokens_mean": 2619.0,
      "output_tokens_mean": 304.6,
      "tokens_per_completed_task": 3863.0,
      "model_turns_mean": 3.41,
      "business_tool_calls_mean": 1.04,
      "search_tools_calls_mean": 1.36,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "group": "multi_step",
      "count": 100,
      "completed": 39,
      "success_pct": 39.0,
      "input_tokens_mean": 5219.8,
      "output_tokens_mean": 675.8,
      "tokens_per_completed_task": 15117.0,
      "model_turns_mean": 5.15,
      "business_tool_calls_mean": 1.88,
      "search_tools_calls_mean": 2.27,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "group": "ambiguous",
      "count": 20,
      "completed": 20,
      "success_pct": 100.0,
      "input_tokens_mean": 1558.7,
      "output_tokens_mean": 274.6,
      "tokens_per_completed_task": 1833.0,
      "model_turns_mean": 2.05,
      "business_tool_calls_mean": 0.15,
      "search_tools_calls_mean": 0.9,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 4408.9,
      "output_tokens_mean": 692.4,
      "tokens_per_completed_task": 5344.0,
      "model_turns_mean": 4.36,
      "business_tool_calls_mean": 0.09,
      "search_tools_calls_mean": 3.27,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 2795.0,
      "output_tokens_mean": 263.5,
      "tokens_per_completed_task": 3058.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "group": "single_tool",
      "count": 185,
      "completed": 151,
      "success_pct": 81.6,
      "input_tokens_mean": 1760.5,
      "output_tokens_mean": 186.0,
      "tokens_per_completed_task": 2385.0,
      "model_turns_mean": 2.15,
      "business_tool_calls_mean": 1.01,
      "search_tools_calls_mean": 0.15,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "group": "multi_step",
      "count": 100,
      "completed": 61,
      "success_pct": 61.0,
      "input_tokens_mean": 2983.8,
      "output_tokens_mean": 384.1,
      "tokens_per_completed_task": 5521.0,
      "model_turns_mean": 3.31,
      "business_tool_calls_mean": 2.08,
      "search_tools_calls_mean": 0.23,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 1142.8,
      "output_tokens_mean": 198.6,
      "tokens_per_completed_task": 1490.0,
      "model_turns_mean": 1.4,
      "business_tool_calls_mean": 0.25,
      "search_tools_calls_mean": 0.15,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 3791.0,
      "output_tokens_mean": 632.9,
      "tokens_per_completed_task": 4424.0,
      "model_turns_mean": 3.32,
      "business_tool_calls_mean": 0.14,
      "search_tools_calls_mean": 2.18,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 2229.0,
      "output_tokens_mean": 164.0,
      "tokens_per_completed_task": 2393.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "group": "single_tool",
      "count": 185,
      "completed": 162,
      "success_pct": 87.6,
      "input_tokens_mean": 23440.9,
      "output_tokens_mean": 235.1,
      "tokens_per_completed_task": 27037.0,
      "model_turns_mean": 2.24,
      "business_tool_calls_mean": 1.28,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "group": "multi_step",
      "count": 100,
      "completed": 72,
      "success_pct": 72.0,
      "input_tokens_mean": 41505.9,
      "output_tokens_mean": 460.9,
      "tokens_per_completed_task": 58287.0,
      "model_turns_mean": 3.93,
      "business_tool_calls_mean": 2.96,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 17325.0,
      "output_tokens_mean": 245.9,
      "tokens_per_completed_task": 18496.0,
      "model_turns_mean": 1.65,
      "business_tool_calls_mean": 0.65,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 21423.9,
      "output_tokens_mean": 711.9,
      "tokens_per_completed_task": 23190.0,
      "model_turns_mean": 2.05,
      "business_tool_calls_mean": 1.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 31339.5,
      "output_tokens_mean": 254.5,
      "tokens_per_completed_task": 31594.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "group": "single_tool",
      "count": 185,
      "completed": 133,
      "success_pct": 71.9,
      "input_tokens_mean": 3323.4,
      "output_tokens_mean": 1052.4,
      "tokens_per_completed_task": 6087.0,
      "model_turns_mean": 1.77,
      "business_tool_calls_mean": 0.89,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 22,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "group": "multi_step",
      "count": 100,
      "completed": 54,
      "success_pct": 54.0,
      "input_tokens_mean": 7917.4,
      "output_tokens_mean": 1314.4,
      "tokens_per_completed_task": 17096.0,
      "model_turns_mean": 3.11,
      "business_tool_calls_mean": 2.24,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 12,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "group": "ambiguous",
      "count": 20,
      "completed": 17,
      "success_pct": 85.0,
      "input_tokens_mean": 2808.0,
      "output_tokens_mean": 1096.5,
      "tokens_per_completed_task": 4593.0,
      "model_turns_mean": 1.25,
      "business_tool_calls_mean": 0.4,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 3,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "group": "unsupported",
      "count": 22,
      "completed": 19,
      "success_pct": 86.4,
      "input_tokens_mean": 3106.2,
      "output_tokens_mean": 1425.5,
      "tokens_per_completed_task": 5247.0,
      "model_turns_mean": 1.32,
      "business_tool_calls_mean": 0.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 3,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 3880.0,
      "output_tokens_mean": 1245.0,
      "tokens_per_completed_task": 5125.0,
      "model_turns_mean": 2.5,
      "business_tool_calls_mean": 1.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "group": "single_tool",
      "count": 185,
      "completed": 122,
      "success_pct": 65.9,
      "input_tokens_mean": 4769.0,
      "output_tokens_mean": 269.3,
      "tokens_per_completed_task": 7640.0,
      "model_turns_mean": 1.9,
      "business_tool_calls_mean": 0.9,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "group": "multi_step",
      "count": 100,
      "completed": 41,
      "success_pct": 41.0,
      "input_tokens_mean": 7989.4,
      "output_tokens_mean": 536.5,
      "tokens_per_completed_task": 20795.0,
      "model_turns_mean": 3.04,
      "business_tool_calls_mean": 2.06,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 2706.9,
      "output_tokens_mean": 200.3,
      "tokens_per_completed_task": 3230.0,
      "model_turns_mean": 1.1,
      "business_tool_calls_mean": 0.1,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 3756.0,
      "output_tokens_mean": 474.3,
      "tokens_per_completed_task": 4230.0,
      "model_turns_mean": 1.55,
      "business_tool_calls_mean": 0.55,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 3754.0,
      "output_tokens_mean": 241.0,
      "tokens_per_completed_task": 7990.0,
      "model_turns_mean": 1.5,
      "business_tool_calls_mean": 0.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "group": "single_tool",
      "count": 185,
      "completed": 99,
      "success_pct": 53.5,
      "input_tokens_mean": 2511.4,
      "output_tokens_mean": 250.2,
      "tokens_per_completed_task": 5161.0,
      "model_turns_mean": 1.65,
      "business_tool_calls_mean": 0.65,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "group": "multi_step",
      "count": 100,
      "completed": 21,
      "success_pct": 21.0,
      "input_tokens_mean": 3127.4,
      "output_tokens_mean": 517.7,
      "tokens_per_completed_task": 17358.0,
      "model_turns_mean": 2.02,
      "business_tool_calls_mean": 1.02,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 1491.0,
      "output_tokens_mean": 168.8,
      "tokens_per_completed_task": 1747.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 1755.2,
      "output_tokens_mean": 337.3,
      "tokens_per_completed_task": 2093.0,
      "model_turns_mean": 1.23,
      "business_tool_calls_mean": 0.23,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 3087.0,
      "output_tokens_mean": 246.5,
      "tokens_per_completed_task": 6667.0,
      "model_turns_mean": 2.0,
      "business_tool_calls_mean": 1.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "group": "single_tool",
      "count": 185,
      "completed": 139,
      "success_pct": 75.1,
      "input_tokens_mean": 2890.2,
      "output_tokens_mean": 231.8,
      "tokens_per_completed_task": 4155.0,
      "model_turns_mean": 1.97,
      "business_tool_calls_mean": 0.97,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "group": "multi_step",
      "count": 100,
      "completed": 47,
      "success_pct": 47.0,
      "input_tokens_mean": 4781.1,
      "output_tokens_mean": 497.0,
      "tokens_per_completed_task": 11230.0,
      "model_turns_mean": 3.13,
      "business_tool_calls_mean": 2.13,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 1490.6,
      "output_tokens_mean": 162.5,
      "tokens_per_completed_task": 1740.0,
      "model_turns_mean": 1.05,
      "business_tool_calls_mean": 0.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 2207.0,
      "output_tokens_mean": 413.7,
      "tokens_per_completed_task": 2621.0,
      "model_turns_mean": 1.5,
      "business_tool_calls_mean": 0.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 3987.0,
      "output_tokens_mean": 211.0,
      "tokens_per_completed_task": 4198.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "group": "single_tool",
      "count": 185,
      "completed": 154,
      "success_pct": 83.2,
      "input_tokens_mean": 2939.6,
      "output_tokens_mean": 220.6,
      "tokens_per_completed_task": 3796.0,
      "model_turns_mean": 2.02,
      "business_tool_calls_mean": 1.02,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "group": "multi_step",
      "count": 100,
      "completed": 55,
      "success_pct": 55.0,
      "input_tokens_mean": 5085.2,
      "output_tokens_mean": 451.6,
      "tokens_per_completed_task": 10067.0,
      "model_turns_mean": 3.32,
      "business_tool_calls_mean": 2.33,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 20,
      "success_pct": 100.0,
      "input_tokens_mean": 1548.5,
      "output_tokens_mean": 154.5,
      "tokens_per_completed_task": 1703.0,
      "model_turns_mean": 1.1,
      "business_tool_calls_mean": 0.1,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 2780.3,
      "output_tokens_mean": 591.6,
      "tokens_per_completed_task": 3532.0,
      "model_turns_mean": 1.91,
      "business_tool_calls_mean": 0.91,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4002.0,
      "output_tokens_mean": 190.0,
      "tokens_per_completed_task": 4192.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "group": "single_tool",
      "count": 185,
      "completed": 148,
      "success_pct": 80.0,
      "input_tokens_mean": 4191.4,
      "output_tokens_mean": 363.4,
      "tokens_per_completed_task": 5694.0,
      "model_turns_mean": 3.37,
      "business_tool_calls_mean": 1.08,
      "search_tools_calls_mean": 1.29,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "group": "multi_step",
      "count": 100,
      "completed": 68,
      "success_pct": 68.0,
      "input_tokens_mean": 9272.7,
      "output_tokens_mean": 672.1,
      "tokens_per_completed_task": 14625.0,
      "model_turns_mean": 5.61,
      "business_tool_calls_mean": 2.58,
      "search_tools_calls_mean": 2.03,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 3005.1,
      "output_tokens_mean": 393.6,
      "tokens_per_completed_task": 3577.0,
      "model_turns_mean": 2.45,
      "business_tool_calls_mean": 0.35,
      "search_tools_calls_mean": 1.1,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "group": "unsupported",
      "count": 22,
      "completed": 20,
      "success_pct": 90.9,
      "input_tokens_mean": 13277.0,
      "output_tokens_mean": 1295.8,
      "tokens_per_completed_task": 16030.0,
      "model_turns_mean": 6.27,
      "business_tool_calls_mean": 1.32,
      "search_tools_calls_mean": 4.05,
      "step_limit_failures": 2,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4407.5,
      "output_tokens_mean": 265.0,
      "tokens_per_completed_task": 4672.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "group": "single_tool",
      "count": 185,
      "completed": 145,
      "success_pct": 78.4,
      "input_tokens_mean": 4498.7,
      "output_tokens_mean": 403.1,
      "tokens_per_completed_task": 6254.0,
      "model_turns_mean": 3.48,
      "business_tool_calls_mean": 1.11,
      "search_tools_calls_mean": 1.38,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "group": "multi_step",
      "count": 100,
      "completed": 56,
      "success_pct": 56.0,
      "input_tokens_mean": 10906.0,
      "output_tokens_mean": 826.5,
      "tokens_per_completed_task": 20951.0,
      "model_turns_mean": 5.98,
      "business_tool_calls_mean": 2.64,
      "search_tools_calls_mean": 2.37,
      "step_limit_failures": 2,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "group": "ambiguous",
      "count": 20,
      "completed": 19,
      "success_pct": 95.0,
      "input_tokens_mean": 2167.9,
      "output_tokens_mean": 293.8,
      "tokens_per_completed_task": 2591.0,
      "model_turns_mean": 2.0,
      "business_tool_calls_mean": 0.1,
      "search_tools_calls_mean": 0.9,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 10260.5,
      "output_tokens_mean": 1173.1,
      "tokens_per_completed_task": 11434.0,
      "model_turns_mean": 5.45,
      "business_tool_calls_mean": 1.05,
      "search_tools_calls_mean": 3.41,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4512.5,
      "output_tokens_mean": 279.0,
      "tokens_per_completed_task": 4792.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "group": "single_tool",
      "count": 185,
      "completed": 150,
      "success_pct": 81.1,
      "input_tokens_mean": 3932.4,
      "output_tokens_mean": 248.3,
      "tokens_per_completed_task": 5156.0,
      "model_turns_mean": 2.22,
      "business_tool_calls_mean": 1.03,
      "search_tools_calls_mean": 0.23,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "group": "multi_step",
      "count": 100,
      "completed": 62,
      "success_pct": 62.0,
      "input_tokens_mean": 8001.3,
      "output_tokens_mean": 568.3,
      "tokens_per_completed_task": 13822.0,
      "model_turns_mean": 4.09,
      "business_tool_calls_mean": 2.65,
      "search_tools_calls_mean": 0.45,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 3934.7,
      "output_tokens_mean": 259.4,
      "tokens_per_completed_task": 4660.0,
      "model_turns_mean": 2,
      "business_tool_calls_mean": 0.55,
      "search_tools_calls_mean": 0.45,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 11819.4,
      "output_tokens_mean": 1141.0,
      "tokens_per_completed_task": 12960.0,
      "model_turns_mean": 4.95,
      "business_tool_calls_mean": 1.23,
      "search_tools_calls_mean": 2.73,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4551.0,
      "output_tokens_mean": 174.5,
      "tokens_per_completed_task": 4726.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "group": "single_tool",
      "count": 185,
      "completed": 151,
      "success_pct": 81.6,
      "input_tokens_mean": 41560.4,
      "output_tokens_mean": 448.4,
      "tokens_per_completed_task": 51468.0,
      "model_turns_mean": 3.95,
      "business_tool_calls_mean": 3.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "group": "multi_step",
      "count": 100,
      "completed": 66,
      "success_pct": 66.0,
      "input_tokens_mean": 67838.0,
      "output_tokens_mean": 730.9,
      "tokens_per_completed_task": 103892.0,
      "model_turns_mean": 6.36,
      "business_tool_calls_mean": 5.55,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 19,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 19465.0,
      "output_tokens_mean": 330.7,
      "tokens_per_completed_task": 21995.0,
      "model_turns_mean": 1.85,
      "business_tool_calls_mean": 0.85,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 23803.9,
      "output_tokens_mean": 1356.3,
      "tokens_per_completed_task": 26358.0,
      "model_turns_mean": 2.27,
      "business_tool_calls_mean": 1.27,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 36624.0,
      "output_tokens_mean": 320.5,
      "tokens_per_completed_task": 36944.0,
      "model_turns_mean": 3.5,
      "business_tool_calls_mean": 2.5,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "group": "single_tool",
      "count": 185,
      "completed": 107,
      "success_pct": 57.8,
      "input_tokens_mean": 4966.1,
      "output_tokens_mean": 460.2,
      "tokens_per_completed_task": 9382.0,
      "model_turns_mean": 2.63,
      "business_tool_calls_mean": 1.89,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 12,
      "routing_errors": 36,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "group": "multi_step",
      "count": 100,
      "completed": 32,
      "success_pct": 32.0,
      "input_tokens_mean": 9475.3,
      "output_tokens_mean": 817.1,
      "tokens_per_completed_task": 32164.0,
      "model_turns_mean": 3.84,
      "business_tool_calls_mean": 3.33,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 12,
      "routing_errors": 37,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "group": "ambiguous",
      "count": 20,
      "completed": 14,
      "success_pct": 70.0,
      "input_tokens_mean": 2316.8,
      "output_tokens_mean": 321.8,
      "tokens_per_completed_task": 3769.0,
      "model_turns_mean": 1.1,
      "business_tool_calls_mean": 0.25,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 3,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 5549.9,
      "output_tokens_mean": 1085.5,
      "tokens_per_completed_task": 6951.0,
      "model_turns_mean": 2.55,
      "business_tool_calls_mean": 1.59,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4637.5,
      "output_tokens_mean": 321.0,
      "tokens_per_completed_task": 4958.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "group": "single_tool",
      "count": 185,
      "completed": 114,
      "success_pct": 61.6,
      "input_tokens_mean": 8143.1,
      "output_tokens_mean": 470.2,
      "tokens_per_completed_task": 13978.0,
      "model_turns_mean": 3.07,
      "business_tool_calls_mean": 2.15,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 14,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "group": "multi_step",
      "count": 100,
      "completed": 45,
      "success_pct": 45.0,
      "input_tokens_mean": 12144.5,
      "output_tokens_mean": 744.6,
      "tokens_per_completed_task": 28642.0,
      "model_turns_mean": 4.49,
      "business_tool_calls_mean": 3.59,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 10,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "group": "ambiguous",
      "count": 20,
      "completed": 17,
      "success_pct": 85.0,
      "input_tokens_mean": 2445.8,
      "output_tokens_mean": 316.8,
      "tokens_per_completed_task": 3250.0,
      "model_turns_mean": 1,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 4061.0,
      "output_tokens_mean": 742.1,
      "tokens_per_completed_task": 4803.0,
      "model_turns_mean": 1.64,
      "business_tool_calls_mean": 0.64,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 5046.0,
      "output_tokens_mean": 266.5,
      "tokens_per_completed_task": 10625.0,
      "model_turns_mean": 2.0,
      "business_tool_calls_mean": 1.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "group": "single_tool",
      "count": 185,
      "completed": 91,
      "success_pct": 49.2,
      "input_tokens_mean": 5583.0,
      "output_tokens_mean": 488.6,
      "tokens_per_completed_task": 12343.0,
      "model_turns_mean": 3.31,
      "business_tool_calls_mean": 2.43,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 21,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "group": "multi_step",
      "count": 100,
      "completed": 23,
      "success_pct": 23.0,
      "input_tokens_mean": 5952.8,
      "output_tokens_mean": 945.6,
      "tokens_per_completed_task": 29993.0,
      "model_turns_mean": 3.54,
      "business_tool_calls_mean": 2.65,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 11,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 14,
      "success_pct": 70.0,
      "input_tokens_mean": 1491.0,
      "output_tokens_mean": 258.9,
      "tokens_per_completed_task": 2500.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "group": "unsupported",
      "count": 22,
      "completed": 19,
      "success_pct": 86.4,
      "input_tokens_mean": 5258.8,
      "output_tokens_mean": 709.5,
      "tokens_per_completed_task": 6911.0,
      "model_turns_mean": 3.32,
      "business_tool_calls_mean": 2.45,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 3,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 1,
      "success_pct": 50.0,
      "input_tokens_mean": 3087.0,
      "output_tokens_mean": 453.5,
      "tokens_per_completed_task": 7081.0,
      "model_turns_mean": 2.0,
      "business_tool_calls_mean": 1.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "group": "single_tool",
      "count": 185,
      "completed": 126,
      "success_pct": 68.1,
      "input_tokens_mean": 6179.7,
      "output_tokens_mean": 495.2,
      "tokens_per_completed_task": 9800.0,
      "model_turns_mean": 3.85,
      "business_tool_calls_mean": 2.98,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 24,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "group": "multi_step",
      "count": 100,
      "completed": 40,
      "success_pct": 40.0,
      "input_tokens_mean": 8971.9,
      "output_tokens_mean": 894.9,
      "tokens_per_completed_task": 24667.0,
      "model_turns_mean": 5.43,
      "business_tool_calls_mean": 4.61,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 18,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 17,
      "success_pct": 85.0,
      "input_tokens_mean": 1490.6,
      "output_tokens_mean": 286.6,
      "tokens_per_completed_task": 2091.0,
      "model_turns_mean": 1.05,
      "business_tool_calls_mean": 0.05,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "group": "unsupported",
      "count": 22,
      "completed": 22,
      "success_pct": 100.0,
      "input_tokens_mean": 2571.4,
      "output_tokens_mean": 998.6,
      "tokens_per_completed_task": 3570.0,
      "model_turns_mean": 1.73,
      "business_tool_calls_mean": 0.73,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 3987.0,
      "output_tokens_mean": 252.0,
      "tokens_per_completed_task": 4239.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "group": "single_tool",
      "count": 185,
      "completed": 139,
      "success_pct": 75.1,
      "input_tokens_mean": 5932.3,
      "output_tokens_mean": 420.8,
      "tokens_per_completed_task": 8456.0,
      "model_turns_mean": 3.67,
      "business_tool_calls_mean": 2.78,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "group": "multi_step",
      "count": 100,
      "completed": 47,
      "success_pct": 47.0,
      "input_tokens_mean": 10380.4,
      "output_tokens_mean": 854.0,
      "tokens_per_completed_task": 23903.0,
      "model_turns_mean": 6.12,
      "business_tool_calls_mean": 5.34,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 22,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "group": "ambiguous",
      "count": 20,
      "completed": 18,
      "success_pct": 90.0,
      "input_tokens_mean": 1555.0,
      "output_tokens_mean": 253.0,
      "tokens_per_completed_task": 2009.0,
      "model_turns_mean": 1.1,
      "business_tool_calls_mean": 0.1,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "group": "unsupported",
      "count": 22,
      "completed": 21,
      "success_pct": 95.5,
      "input_tokens_mean": 3920.7,
      "output_tokens_mean": 1099.5,
      "tokens_per_completed_task": 5259.0,
      "model_turns_mean": 2.5,
      "business_tool_calls_mean": 1.55,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4002.0,
      "output_tokens_mean": 278.5,
      "tokens_per_completed_task": 4280.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "group": "single_tool",
      "count": 185,
      "completed": 145,
      "success_pct": 78.4,
      "input_tokens_mean": 6425.2,
      "output_tokens_mean": 508.4,
      "tokens_per_completed_task": 8846.0,
      "model_turns_mean": 4.77,
      "business_tool_calls_mean": 2.75,
      "search_tools_calls_mean": 1.14,
      "step_limit_failures": 22,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "group": "multi_step",
      "count": 100,
      "completed": 61,
      "success_pct": 61.0,
      "input_tokens_mean": 14389.8,
      "output_tokens_mean": 1013.5,
      "tokens_per_completed_task": 25251.0,
      "model_turns_mean": 7.78,
      "business_tool_calls_mean": 4.82,
      "search_tools_calls_mean": 2.14,
      "step_limit_failures": 18,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "group": "ambiguous",
      "count": 20,
      "completed": 16,
      "success_pct": 80.0,
      "input_tokens_mean": 6684.4,
      "output_tokens_mean": 719.6,
      "tokens_per_completed_task": 9255.0,
      "model_turns_mean": 3.85,
      "business_tool_calls_mean": 0.45,
      "search_tools_calls_mean": 2.4,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "group": "unsupported",
      "count": 22,
      "completed": 16,
      "success_pct": 72.7,
      "input_tokens_mean": 32291.3,
      "output_tokens_mean": 2745.5,
      "tokens_per_completed_task": 48176.0,
      "model_turns_mean": 10.73,
      "business_tool_calls_mean": 1.27,
      "search_tools_calls_mean": 8.73,
      "step_limit_failures": 6,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4626.5,
      "output_tokens_mean": 344.5,
      "tokens_per_completed_task": 4971.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "group": "single_tool",
      "count": 185,
      "completed": 143,
      "success_pct": 77.3,
      "input_tokens_mean": 7081.5,
      "output_tokens_mean": 532.3,
      "tokens_per_completed_task": 9850.0,
      "model_turns_mean": 4.99,
      "business_tool_calls_mean": 2.85,
      "search_tools_calls_mean": 1.26,
      "step_limit_failures": 23,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "group": "multi_step",
      "count": 100,
      "completed": 60,
      "success_pct": 60.0,
      "input_tokens_mean": 16016.0,
      "output_tokens_mean": 1079.2,
      "tokens_per_completed_task": 28492.0,
      "model_turns_mean": 8.16,
      "business_tool_calls_mean": 4.79,
      "search_tools_calls_mean": 2.56,
      "step_limit_failures": 19,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "group": "ambiguous",
      "count": 20,
      "completed": 17,
      "success_pct": 85.0,
      "input_tokens_mean": 6795.3,
      "output_tokens_mean": 703.8,
      "tokens_per_completed_task": 8822.0,
      "model_turns_mean": 4,
      "business_tool_calls_mean": 0.55,
      "search_tools_calls_mean": 2.5,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "group": "unsupported",
      "count": 22,
      "completed": 17,
      "success_pct": 77.3,
      "input_tokens_mean": 25400.4,
      "output_tokens_mean": 2493.0,
      "tokens_per_completed_task": 36097.0,
      "model_turns_mean": 9.55,
      "business_tool_calls_mean": 1.41,
      "search_tools_calls_mean": 7.36,
      "step_limit_failures": 5,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4566.5,
      "output_tokens_mean": 354.5,
      "tokens_per_completed_task": 4921.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "group": "single_tool",
      "count": 185,
      "completed": 147,
      "success_pct": 79.5,
      "input_tokens_mean": 7202.5,
      "output_tokens_mean": 438.9,
      "tokens_per_completed_task": 9617.0,
      "model_turns_mean": 3.9,
      "business_tool_calls_mean": 2.85,
      "search_tools_calls_mean": 0.16,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "group": "multi_step",
      "count": 100,
      "completed": 63,
      "success_pct": 63.0,
      "input_tokens_mean": 12667.4,
      "output_tokens_mean": 807.5,
      "tokens_per_completed_task": 21389.0,
      "model_turns_mean": 6.46,
      "business_tool_calls_mean": 5.36,
      "search_tools_calls_mean": 0.3,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "group": "ambiguous",
      "count": 20,
      "completed": 14,
      "success_pct": 70.0,
      "input_tokens_mean": 7983.7,
      "output_tokens_mean": 630.8,
      "tokens_per_completed_task": 12306.0,
      "model_turns_mean": 3.4,
      "business_tool_calls_mean": 1.35,
      "search_tools_calls_mean": 1.15,
      "step_limit_failures": 2,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "group": "unsupported",
      "count": 22,
      "completed": 14,
      "success_pct": 63.6,
      "input_tokens_mean": 35808.9,
      "output_tokens_mean": 2666.4,
      "tokens_per_completed_task": 60461.0,
      "model_turns_mean": 10.41,
      "business_tool_calls_mean": 1.55,
      "search_tools_calls_mean": 8.23,
      "step_limit_failures": 8,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "group": "confusable_2op",
      "count": 2,
      "completed": 2,
      "success_pct": 100.0,
      "input_tokens_mean": 4551.0,
      "output_tokens_mean": 264.5,
      "tokens_per_completed_task": 4816.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "group": "single_tool",
      "count": 555,
      "completed": 471,
      "success_pct": 84.9,
      "input_tokens_mean": 24015.9,
      "output_tokens_mean": 286.7,
      "tokens_per_completed_task": 28637.0,
      "model_turns_mean": 2.78,
      "business_tool_calls_mean": 1.82,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "group": "multi_step",
      "count": 300,
      "completed": 207,
      "success_pct": 69.0,
      "input_tokens_mean": 40333.3,
      "output_tokens_mean": 523.5,
      "tokens_per_completed_task": 59213.0,
      "model_turns_mean": 4.58,
      "business_tool_calls_mean": 3.66,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 19,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "group": "ambiguous",
      "count": 60,
      "completed": 57,
      "success_pct": 95.0,
      "input_tokens_mean": 13511.4,
      "output_tokens_mean": 237.8,
      "tokens_per_completed_task": 14473.0,
      "model_turns_mean": 1.55,
      "business_tool_calls_mean": 0.55,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "group": "unsupported",
      "count": 66,
      "completed": 64,
      "success_pct": 97.0,
      "input_tokens_mean": 16410.3,
      "output_tokens_mean": 767.9,
      "tokens_per_completed_task": 17715.0,
      "model_turns_mean": 1.85,
      "business_tool_calls_mean": 0.85,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 25946.8,
      "output_tokens_mean": 249.7,
      "tokens_per_completed_task": 26196.0,
      "model_turns_mean": 3.17,
      "business_tool_calls_mean": 2.17,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "group": "single_tool",
      "count": 555,
      "completed": 373,
      "success_pct": 67.2,
      "input_tokens_mean": 3310.9,
      "output_tokens_mean": 600.5,
      "tokens_per_completed_task": 5820.0,
      "model_turns_mean": 2.1,
      "business_tool_calls_mean": 1.24,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 12,
      "routing_errors": 61,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "group": "multi_step",
      "count": 300,
      "completed": 140,
      "success_pct": 46.7,
      "input_tokens_mean": 6959.7,
      "output_tokens_mean": 893.2,
      "tokens_per_completed_task": 16828.0,
      "model_turns_mean": 3.37,
      "business_tool_calls_mean": 2.57,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 12,
      "routing_errors": 49,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "group": "ambiguous",
      "count": 60,
      "completed": 50,
      "success_pct": 83.3,
      "input_tokens_mean": 2084.5,
      "output_tokens_mean": 583.9,
      "tokens_per_completed_task": 3202.0,
      "model_turns_mean": 1.17,
      "business_tool_calls_mean": 0.27,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 6,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "group": "unsupported",
      "count": 66,
      "completed": 62,
      "success_pct": 93.9,
      "input_tokens_mean": 3280.0,
      "output_tokens_mean": 958.4,
      "tokens_per_completed_task": 4512.0,
      "model_turns_mean": 1.65,
      "business_tool_calls_mean": 0.73,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 1,
      "routing_errors": 3,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3469.5,
      "output_tokens_mean": 591.8,
      "tokens_per_completed_task": 4061.0,
      "model_turns_mean": 2.67,
      "business_tool_calls_mean": 1.67,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "group": "single_tool",
      "count": 555,
      "completed": 358,
      "success_pct": 64.5,
      "input_tokens_mean": 4887.4,
      "output_tokens_mean": 298.5,
      "tokens_per_completed_task": 8040.0,
      "model_turns_mean": 2.27,
      "business_tool_calls_mean": 1.29,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 14,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "group": "multi_step",
      "count": 300,
      "completed": 127,
      "success_pct": 42.3,
      "input_tokens_mean": 7645.6,
      "output_tokens_mean": 539.7,
      "tokens_per_completed_task": 19335.0,
      "model_turns_mean": 3.41,
      "business_tool_calls_mean": 2.45,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 10,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "group": "ambiguous",
      "count": 60,
      "completed": 50,
      "success_pct": 83.3,
      "input_tokens_mean": 2022.5,
      "output_tokens_mean": 211.3,
      "tokens_per_completed_task": 2681.0,
      "model_turns_mean": 1.03,
      "business_tool_calls_mean": 0.03,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "group": "unsupported",
      "count": 66,
      "completed": 66,
      "success_pct": 100.0,
      "input_tokens_mean": 2910.2,
      "output_tokens_mean": 450.9,
      "tokens_per_completed_task": 3361.0,
      "model_turns_mean": 1.4,
      "business_tool_calls_mean": 0.4,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "group": "confusable_2op",
      "count": 6,
      "completed": 3,
      "success_pct": 50.0,
      "input_tokens_mean": 3393.2,
      "output_tokens_mean": 206.8,
      "tokens_per_completed_task": 7200.0,
      "model_turns_mean": 1.67,
      "business_tool_calls_mean": 0.67,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "group": "single_tool",
      "count": 555,
      "completed": 284,
      "success_pct": 51.2,
      "input_tokens_mean": 3047.6,
      "output_tokens_mean": 291.4,
      "tokens_per_completed_task": 6525.0,
      "model_turns_mean": 2.18,
      "business_tool_calls_mean": 1.22,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 21,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "group": "multi_step",
      "count": 300,
      "completed": 60,
      "success_pct": 20.0,
      "input_tokens_mean": 3378.0,
      "output_tokens_mean": 570.9,
      "tokens_per_completed_task": 19744.0,
      "model_turns_mean": 2.37,
      "business_tool_calls_mean": 1.41,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 11,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "group": "ambiguous",
      "count": 60,
      "completed": 51,
      "success_pct": 85.0,
      "input_tokens_mean": 1205.2,
      "output_tokens_mean": 180.2,
      "tokens_per_completed_task": 1630.0,
      "model_turns_mean": 1.0,
      "business_tool_calls_mean": 0.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "group": "unsupported",
      "count": 66,
      "completed": 63,
      "success_pct": 95.5,
      "input_tokens_mean": 2555.2,
      "output_tokens_mean": 398.9,
      "tokens_per_completed_task": 3095.0,
      "model_turns_mean": 1.87,
      "business_tool_calls_mean": 0.91,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 3,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "group": "confusable_2op",
      "count": 6,
      "completed": 3,
      "success_pct": 50.0,
      "input_tokens_mean": 2383.0,
      "output_tokens_mean": 274.5,
      "tokens_per_completed_task": 5315.0,
      "model_turns_mean": 1.83,
      "business_tool_calls_mean": 0.83,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "group": "single_tool",
      "count": 555,
      "completed": 396,
      "success_pct": 71.4,
      "input_tokens_mean": 3420.5,
      "output_tokens_mean": 291.2,
      "tokens_per_completed_task": 5202.0,
      "model_turns_mean": 2.55,
      "business_tool_calls_mean": 1.6,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 24,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "group": "multi_step",
      "count": 300,
      "completed": 130,
      "success_pct": 43.3,
      "input_tokens_mean": 5191.4,
      "output_tokens_mean": 575.6,
      "tokens_per_completed_task": 13309.0,
      "model_turns_mean": 3.71,
      "business_tool_calls_mean": 2.77,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 18,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "group": "ambiguous",
      "count": 60,
      "completed": 55,
      "success_pct": 91.7,
      "input_tokens_mean": 1197.4,
      "output_tokens_mean": 190.1,
      "tokens_per_completed_task": 1514.0,
      "model_turns_mean": 1.03,
      "business_tool_calls_mean": 0.03,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "group": "unsupported",
      "count": 66,
      "completed": 66,
      "success_pct": 100.0,
      "input_tokens_mean": 1798.8,
      "output_tokens_mean": 538.5,
      "tokens_per_completed_task": 2337.0,
      "model_turns_mean": 1.41,
      "business_tool_calls_mean": 0.41,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3272.5,
      "output_tokens_mean": 205.0,
      "tokens_per_completed_task": 3478.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "group": "single_tool",
      "count": 555,
      "completed": 437,
      "success_pct": 78.7,
      "input_tokens_mean": 3371.2,
      "output_tokens_mean": 262.4,
      "tokens_per_completed_task": 4615.0,
      "model_turns_mean": 2.53,
      "business_tool_calls_mean": 1.57,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "group": "multi_step",
      "count": 300,
      "completed": 157,
      "success_pct": 52.3,
      "input_tokens_mean": 5811.4,
      "output_tokens_mean": 545.8,
      "tokens_per_completed_task": 12148.0,
      "model_turns_mean": 4.07,
      "business_tool_calls_mean": 3.14,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 22,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "group": "ambiguous",
      "count": 60,
      "completed": 57,
      "success_pct": 95.0,
      "input_tokens_mean": 1251.4,
      "output_tokens_mean": 173.8,
      "tokens_per_completed_task": 1500.0,
      "model_turns_mean": 1.08,
      "business_tool_calls_mean": 0.08,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "group": "unsupported",
      "count": 66,
      "completed": 64,
      "success_pct": 97.0,
      "input_tokens_mean": 2466.4,
      "output_tokens_mean": 635.1,
      "tokens_per_completed_task": 3198.0,
      "model_turns_mean": 1.85,
      "business_tool_calls_mean": 0.87,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3293.0,
      "output_tokens_mean": 208.0,
      "tokens_per_completed_task": 3501.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "group": "single_tool",
      "count": 555,
      "completed": 427,
      "success_pct": 76.9,
      "input_tokens_mean": 4316.6,
      "output_tokens_mean": 381.5,
      "tokens_per_completed_task": 6106.0,
      "model_turns_mean": 3.77,
      "business_tool_calls_mean": 1.59,
      "search_tools_calls_mean": 1.22,
      "step_limit_failures": 22,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "group": "multi_step",
      "count": 300,
      "completed": 172,
      "success_pct": 57.3,
      "input_tokens_mean": 9410.0,
      "output_tokens_mean": 762.6,
      "tokens_per_completed_task": 17743.0,
      "model_turns_mean": 6.06,
      "business_tool_calls_mean": 3.1,
      "search_tools_calls_mean": 2.03,
      "step_limit_failures": 18,
      "routing_errors": 2,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "group": "ambiguous",
      "count": 60,
      "completed": 54,
      "success_pct": 90.0,
      "input_tokens_mean": 3607.4,
      "output_tokens_mean": 433.2,
      "tokens_per_completed_task": 4490.0,
      "model_turns_mean": 2.68,
      "business_tool_calls_mean": 0.27,
      "search_tools_calls_mean": 1.42,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "group": "unsupported",
      "count": 66,
      "completed": 55,
      "success_pct": 83.3,
      "input_tokens_mean": 16578.5,
      "output_tokens_mean": 1600.6,
      "tokens_per_completed_task": 21815.0,
      "model_turns_mean": 7.12,
      "business_tool_calls_mean": 0.95,
      "search_tools_calls_mean": 5.29,
      "step_limit_failures": 8,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3956.5,
      "output_tokens_mean": 283.5,
      "tokens_per_completed_task": 4240.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "group": "single_tool",
      "count": 555,
      "completed": 428,
      "success_pct": 77.1,
      "input_tokens_mean": 4733.1,
      "output_tokens_mean": 413.3,
      "tokens_per_completed_task": 6673.0,
      "model_turns_mean": 3.96,
      "business_tool_calls_mean": 1.67,
      "search_tools_calls_mean": 1.33,
      "step_limit_failures": 24,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "group": "multi_step",
      "count": 300,
      "completed": 155,
      "success_pct": 51.7,
      "input_tokens_mean": 10713.9,
      "output_tokens_mean": 860.5,
      "tokens_per_completed_task": 22402.0,
      "model_turns_mean": 6.43,
      "business_tool_calls_mean": 3.1,
      "search_tools_calls_mean": 2.4,
      "step_limit_failures": 21,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "group": "ambiguous",
      "count": 60,
      "completed": 56,
      "success_pct": 93.3,
      "input_tokens_mean": 3507.3,
      "output_tokens_mean": 424.1,
      "tokens_per_completed_task": 4212.0,
      "model_turns_mean": 2.68,
      "business_tool_calls_mean": 0.27,
      "search_tools_calls_mean": 1.43,
      "step_limit_failures": 1,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "group": "unsupported",
      "count": 66,
      "completed": 60,
      "success_pct": 90.9,
      "input_tokens_mean": 13356.6,
      "output_tokens_mean": 1452.8,
      "tokens_per_completed_task": 16290.0,
      "model_turns_mean": 6.45,
      "business_tool_calls_mean": 0.85,
      "search_tools_calls_mean": 4.68,
      "step_limit_failures": 5,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3958.0,
      "output_tokens_mean": 299.0,
      "tokens_per_completed_task": 4257.0,
      "model_turns_mean": 4.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 1.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "group": "single_tool",
      "count": 555,
      "completed": 448,
      "success_pct": 80.7,
      "input_tokens_mean": 4298.5,
      "output_tokens_mean": 291.1,
      "tokens_per_completed_task": 5686.0,
      "model_turns_mean": 2.76,
      "business_tool_calls_mean": 1.63,
      "search_tools_calls_mean": 0.18,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "group": "multi_step",
      "count": 300,
      "completed": 186,
      "success_pct": 62.0,
      "input_tokens_mean": 7884.2,
      "output_tokens_mean": 586.6,
      "tokens_per_completed_task": 13663.0,
      "model_turns_mean": 4.62,
      "business_tool_calls_mean": 3.36,
      "search_tools_calls_mean": 0.33,
      "step_limit_failures": 20,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "group": "ambiguous",
      "count": 60,
      "completed": 50,
      "success_pct": 83.3,
      "input_tokens_mean": 4353.7,
      "output_tokens_mean": 362.9,
      "tokens_per_completed_task": 5660.0,
      "model_turns_mean": 2.27,
      "business_tool_calls_mean": 0.72,
      "search_tools_calls_mean": 0.58,
      "step_limit_failures": 2,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "group": "unsupported",
      "count": 66,
      "completed": 58,
      "success_pct": 87.9,
      "input_tokens_mean": 17139.8,
      "output_tokens_mean": 1480.1,
      "tokens_per_completed_task": 21188.0,
      "model_turns_mean": 6.23,
      "business_tool_calls_mean": 0.97,
      "search_tools_calls_mean": 4.38,
      "step_limit_failures": 8,
      "routing_errors": 0,
      "api_errors": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "group": "confusable_2op",
      "count": 6,
      "completed": 6,
      "success_pct": 100.0,
      "input_tokens_mean": 3777.0,
      "output_tokens_mean": 201.0,
      "tokens_per_completed_task": 3978.0,
      "model_turns_mean": 3.0,
      "business_tool_calls_mean": 2.0,
      "search_tools_calls_mean": 0.0,
      "step_limit_failures": 0,
      "routing_errors": 0,
      "api_errors": 0
    }
  ],
  "multi_step_detail": [
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 69,
      "success_pct": 69.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 17,
      "success_pct": 73.9,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 52,
      "success_pct": 67.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 32,
      "success_pct": 76.2,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 37,
      "success_pct": 63.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 22,
      "success_pct": 81.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 30,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "all",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 54,
      "success_pct": 54.0,
      "retrieval_missed_needed_tool": 21
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 38,
      "success_pct": 49.4,
      "retrieval_missed_needed_tool": 21
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 27,
      "success_pct": 64.3,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 27,
      "success_pct": 46.6,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 17,
      "success_pct": 63.0,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 21,
      "success_pct": 42.0,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 41,
      "success_pct": 41.0,
      "retrieval_missed_needed_tool": 47
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 14,
      "success_pct": 60.9,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 27,
      "success_pct": 35.1,
      "retrieval_missed_needed_tool": 43
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 17,
      "success_pct": 40.5,
      "retrieval_missed_needed_tool": 20
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 24,
      "success_pct": 41.4,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 7,
      "success_pct": 46.7,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 10,
      "success_pct": 37.0,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 17,
      "success_pct": 34.0,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "category-laya",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 16,
      "success_pct": 16.0,
      "retrieval_missed_needed_tool": 84
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 4,
      "success_pct": 17.4,
      "retrieval_missed_needed_tool": 19
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 12,
      "success_pct": 15.6,
      "retrieval_missed_needed_tool": 65
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 13,
      "success_pct": 31.0,
      "retrieval_missed_needed_tool": 29
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 3,
      "success_pct": 5.2,
      "retrieval_missed_needed_tool": 55
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 4,
      "success_pct": 26.7,
      "retrieval_missed_needed_tool": 11
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 0,
      "success_pct": 0.0,
      "retrieval_missed_needed_tool": 8
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 9,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 3,
      "success_pct": 6.0,
      "retrieval_missed_needed_tool": 47
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 43,
      "success_pct": 43.0,
      "retrieval_missed_needed_tool": 44
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 13,
      "success_pct": 56.5,
      "retrieval_missed_needed_tool": 7
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 30,
      "success_pct": 39.0,
      "retrieval_missed_needed_tool": 37
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 28,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 15,
      "success_pct": 25.9,
      "retrieval_missed_needed_tool": 38
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 8,
      "success_pct": 53.3,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 20,
      "success_pct": 74.1,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 10,
      "success_pct": 20.0,
      "retrieval_missed_needed_tool": 35
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "keyword@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 55,
      "success_pct": 55.0,
      "retrieval_missed_needed_tool": 29
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 39,
      "success_pct": 50.6,
      "retrieval_missed_needed_tool": 26
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 32,
      "success_pct": 76.2,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 23,
      "success_pct": 39.7,
      "retrieval_missed_needed_tool": 25
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 22,
      "success_pct": 81.5,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 17,
      "success_pct": 34.0,
      "retrieval_missed_needed_tool": 24
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "semantic-nim@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 43,
      "success_pct": 43.0,
      "retrieval_missed_needed_tool": 38
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 11,
      "success_pct": 47.8,
      "retrieval_missed_needed_tool": 9
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 32,
      "success_pct": 41.6,
      "retrieval_missed_needed_tool": 29
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 25,
      "success_pct": 59.5,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 18,
      "success_pct": 31.0,
      "retrieval_missed_needed_tool": 26
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 7,
      "success_pct": 46.7,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 4,
      "success_pct": 50.0,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 18,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 7
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 14,
      "success_pct": 28.0,
      "retrieval_missed_needed_tool": 22
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 39,
      "success_pct": 39.0,
      "retrieval_missed_needed_tool": 49
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 11,
      "success_pct": 47.8,
      "retrieval_missed_needed_tool": 9
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 28,
      "success_pct": 36.4,
      "retrieval_missed_needed_tool": 40
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 27,
      "success_pct": 64.3,
      "retrieval_missed_needed_tool": 11
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 12,
      "success_pct": 20.7,
      "retrieval_missed_needed_tool": 38
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 9,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 2,
      "success_pct": 25.0,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 18,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 7
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 10,
      "success_pct": 20.0,
      "retrieval_missed_needed_tool": 33
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "agent-search-kw",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 61,
      "success_pct": 61.0,
      "retrieval_missed_needed_tool": 19
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 17,
      "success_pct": 73.9,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 44,
      "success_pct": 57.1,
      "retrieval_missed_needed_tool": 17
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 30,
      "success_pct": 71.4,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 31,
      "success_pct": 53.4,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 20,
      "success_pct": 74.1,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 24,
      "success_pct": 48.0,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "gpt-oss-20b",
      "strategy": "hybrid",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 72,
      "success_pct": 72.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 19,
      "success_pct": 82.6,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 53,
      "success_pct": 68.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 34,
      "success_pct": 81.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 38,
      "success_pct": 65.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 12,
      "success_pct": 80.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 22,
      "success_pct": 81.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 31,
      "success_pct": 62.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "all",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 54,
      "success_pct": 54.0,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 15,
      "success_pct": 65.2,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 39,
      "success_pct": 50.6,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 26,
      "success_pct": 61.9,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 28,
      "success_pct": 48.3,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 16,
      "success_pct": 59.3,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 23,
      "success_pct": 46.0,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 41,
      "success_pct": 41.0,
      "retrieval_missed_needed_tool": 45
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 15,
      "success_pct": 65.2,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 26,
      "success_pct": 33.8,
      "retrieval_missed_needed_tool": 41
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 19,
      "success_pct": 45.2,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 22,
      "success_pct": 37.9,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 8,
      "success_pct": 53.3,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 11,
      "success_pct": 40.7,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 15,
      "success_pct": 30.0,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "category-laya",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 21,
      "success_pct": 21.0,
      "retrieval_missed_needed_tool": 76
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 7,
      "success_pct": 30.4,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 14,
      "success_pct": 18.2,
      "retrieval_missed_needed_tool": 60
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 14,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 25
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 7,
      "success_pct": 12.1,
      "retrieval_missed_needed_tool": 51
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 5,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 2,
      "success_pct": 25.0,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 9,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 5,
      "success_pct": 10.0,
      "retrieval_missed_needed_tool": 45
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 47,
      "success_pct": 47.0,
      "retrieval_missed_needed_tool": 38
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 31,
      "success_pct": 40.3,
      "retrieval_missed_needed_tool": 34
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 28,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 19,
      "success_pct": 32.8,
      "retrieval_missed_needed_tool": 34
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 9,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 19,
      "success_pct": 70.4,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 12,
      "success_pct": 24.0,
      "retrieval_missed_needed_tool": 33
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "keyword@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 55,
      "success_pct": 55.0,
      "retrieval_missed_needed_tool": 26
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 17,
      "success_pct": 73.9,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 38,
      "success_pct": 49.4,
      "retrieval_missed_needed_tool": 23
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 29,
      "success_pct": 69.0,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 26,
      "success_pct": 44.8,
      "retrieval_missed_needed_tool": 22
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 11,
      "success_pct": 73.3,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 18,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 20,
      "success_pct": 40.0,
      "retrieval_missed_needed_tool": 21
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "semantic-nim@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 68,
      "success_pct": 68.0,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 19,
      "success_pct": 82.6,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 49,
      "success_pct": 63.6,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 33,
      "success_pct": 78.6,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 35,
      "success_pct": 60.3,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 12,
      "success_pct": 80.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 21,
      "success_pct": 77.8,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 28,
      "success_pct": 56.0,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 56,
      "success_pct": 56.0,
      "retrieval_missed_needed_tool": 24
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 40,
      "success_pct": 51.9,
      "retrieval_missed_needed_tool": 21
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 29,
      "success_pct": 69.0,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 27,
      "success_pct": 46.6,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 19,
      "success_pct": 70.4,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 21,
      "success_pct": 42.0,
      "retrieval_missed_needed_tool": 17
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "agent-search-kw",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 62,
      "success_pct": 62.0,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 18,
      "success_pct": 78.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 44,
      "success_pct": 57.1,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 30,
      "success_pct": 71.4,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 32,
      "success_pct": 55.2,
      "retrieval_missed_needed_tool": 13
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 11,
      "success_pct": 73.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 7,
      "success_pct": 87.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 19,
      "success_pct": 70.4,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 25,
      "success_pct": 50.0,
      "retrieval_missed_needed_tool": 13
    },
    {
      "model": "nemotron-3.5-lightning-30b-a3b",
      "strategy": "hybrid",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 66,
      "success_pct": 66.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 17,
      "success_pct": 73.9,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 49,
      "success_pct": 63.6,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 27,
      "success_pct": 64.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 39,
      "success_pct": 67.2,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 12,
      "success_pct": 80.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 15,
      "success_pct": 55.6,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 34,
      "success_pct": 68.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "all",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 32,
      "success_pct": 32.0,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 11,
      "success_pct": 47.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 21,
      "success_pct": 27.3,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 15,
      "success_pct": 35.7,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 17,
      "success_pct": 29.3,
      "retrieval_missed_needed_tool": 9
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 7,
      "success_pct": 46.7,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 4,
      "success_pct": 50.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 8,
      "success_pct": 29.6,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 13,
      "success_pct": 26.0,
      "retrieval_missed_needed_tool": 9
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 45,
      "success_pct": 45.0,
      "retrieval_missed_needed_tool": 46
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 14,
      "success_pct": 60.9,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 31,
      "success_pct": 40.3,
      "retrieval_missed_needed_tool": 42
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 19,
      "success_pct": 45.2,
      "retrieval_missed_needed_tool": 19
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 26,
      "success_pct": 44.8,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 9,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 10,
      "success_pct": 37.0,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 21,
      "success_pct": 42.0,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "category-laya",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 23,
      "success_pct": 23.0,
      "retrieval_missed_needed_tool": 73
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 7,
      "success_pct": 30.4,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 16,
      "success_pct": 20.8,
      "retrieval_missed_needed_tool": 58
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 13,
      "success_pct": 31.0,
      "retrieval_missed_needed_tool": 26
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 10,
      "success_pct": 17.2,
      "retrieval_missed_needed_tool": 47
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 4,
      "success_pct": 26.7,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 3,
      "success_pct": 37.5,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 9,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 7,
      "success_pct": 14.0,
      "retrieval_missed_needed_tool": 42
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 40,
      "success_pct": 40.0,
      "retrieval_missed_needed_tool": 41
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 14,
      "success_pct": 60.9,
      "retrieval_missed_needed_tool": 7
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 26,
      "success_pct": 33.8,
      "retrieval_missed_needed_tool": 34
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 24,
      "success_pct": 57.1,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 16,
      "success_pct": 27.6,
      "retrieval_missed_needed_tool": 36
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 9,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 15,
      "success_pct": 55.6,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 11,
      "success_pct": 22.0,
      "retrieval_missed_needed_tool": 33
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "keyword@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 47,
      "success_pct": 47.0,
      "retrieval_missed_needed_tool": 29
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 31,
      "success_pct": 40.3,
      "retrieval_missed_needed_tool": 27
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 24,
      "success_pct": 57.1,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 23,
      "success_pct": 39.7,
      "retrieval_missed_needed_tool": 25
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 11,
      "success_pct": 73.3,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 5,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 13,
      "success_pct": 48.1,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 18,
      "success_pct": 36.0,
      "retrieval_missed_needed_tool": 25
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "semantic-nim@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 61,
      "success_pct": 61.0,
      "retrieval_missed_needed_tool": 11
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 18,
      "success_pct": 78.3,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 43,
      "success_pct": 55.8,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 27,
      "success_pct": 64.3,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 34,
      "success_pct": 58.6,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 12,
      "success_pct": 80.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 15,
      "success_pct": 55.6,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 28,
      "success_pct": 56.0,
      "retrieval_missed_needed_tool": 9
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 60,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 16,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 44,
      "success_pct": 57.1,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 26,
      "success_pct": 61.9,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 34,
      "success_pct": 58.6,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 10,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 16,
      "success_pct": 59.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 28,
      "success_pct": 56.0,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "agent-search-kw",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 100,
      "completed": 63,
      "success_pct": 63.0,
      "retrieval_missed_needed_tool": 14
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 23,
      "completed": 18,
      "success_pct": 78.3,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 77,
      "completed": 45,
      "success_pct": 58.4,
      "retrieval_missed_needed_tool": 13
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 42,
      "completed": 27,
      "success_pct": 64.3,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 58,
      "completed": 36,
      "success_pct": 62.1,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 15,
      "completed": 12,
      "success_pct": 80.0,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 8,
      "completed": 6,
      "success_pct": 75.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 27,
      "completed": 15,
      "success_pct": 55.6,
      "retrieval_missed_needed_tool": 1
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 50,
      "completed": 30,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "nemotron-3-super-120b-a12b",
      "strategy": "hybrid",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 207,
      "success_pct": 69.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 53,
      "success_pct": 76.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 154,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 93,
      "success_pct": 73.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 114,
      "success_pct": 65.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 34,
      "success_pct": 75.6,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 19,
      "success_pct": 79.2,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 59,
      "success_pct": 72.8,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 95,
      "success_pct": 63.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "all",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 140,
      "success_pct": 46.7,
      "retrieval_missed_needed_tool": 51
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 42,
      "success_pct": 60.9,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 98,
      "success_pct": 42.4,
      "retrieval_missed_needed_tool": 51
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 68,
      "success_pct": 54.0,
      "retrieval_missed_needed_tool": 17
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 72,
      "success_pct": 41.4,
      "retrieval_missed_needed_tool": 34
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 27,
      "success_pct": 60.0,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 15,
      "success_pct": 62.5,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 41,
      "success_pct": 50.6,
      "retrieval_missed_needed_tool": 17
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 57,
      "success_pct": 38.0,
      "retrieval_missed_needed_tool": 34
    },
    {
      "model": "all-models",
      "strategy": "category",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 127,
      "success_pct": 42.3,
      "retrieval_missed_needed_tool": 138
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 43,
      "success_pct": 62.3,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 84,
      "success_pct": 36.4,
      "retrieval_missed_needed_tool": 126
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 55,
      "success_pct": 43.7,
      "retrieval_missed_needed_tool": 57
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 72,
      "success_pct": 41.4,
      "retrieval_missed_needed_tool": 81
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 24,
      "success_pct": 53.3,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 19,
      "success_pct": 79.2,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 31,
      "success_pct": 38.3,
      "retrieval_missed_needed_tool": 45
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 53,
      "success_pct": 35.3,
      "retrieval_missed_needed_tool": 81
    },
    {
      "model": "all-models",
      "strategy": "category-laya",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 60,
      "success_pct": 20.0,
      "retrieval_missed_needed_tool": 233
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 18,
      "success_pct": 26.1,
      "retrieval_missed_needed_tool": 50
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 42,
      "success_pct": 18.2,
      "retrieval_missed_needed_tool": 183
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 40,
      "success_pct": 31.7,
      "retrieval_missed_needed_tool": 80
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 20,
      "success_pct": 11.5,
      "retrieval_missed_needed_tool": 153
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 13,
      "success_pct": 28.9,
      "retrieval_missed_needed_tool": 31
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 5,
      "success_pct": 20.8,
      "retrieval_missed_needed_tool": 19
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 27,
      "success_pct": 33.3,
      "retrieval_missed_needed_tool": 49
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 15,
      "success_pct": 10.0,
      "retrieval_missed_needed_tool": 134
    },
    {
      "model": "all-models",
      "strategy": "semantic@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 130,
      "success_pct": 43.3,
      "retrieval_missed_needed_tool": 123
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 43,
      "success_pct": 62.3,
      "retrieval_missed_needed_tool": 18
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 87,
      "success_pct": 37.7,
      "retrieval_missed_needed_tool": 105
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 80,
      "success_pct": 63.5,
      "retrieval_missed_needed_tool": 15
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 50,
      "success_pct": 28.7,
      "retrieval_missed_needed_tool": 108
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 26,
      "success_pct": 57.8,
      "retrieval_missed_needed_tool": 11
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 17,
      "success_pct": 70.8,
      "retrieval_missed_needed_tool": 7
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 54,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 4
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 33,
      "success_pct": 22.0,
      "retrieval_missed_needed_tool": 101
    },
    {
      "model": "all-models",
      "strategy": "keyword@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 157,
      "success_pct": 52.3,
      "retrieval_missed_needed_tool": 84
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 49,
      "success_pct": 71.0,
      "retrieval_missed_needed_tool": 8
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 108,
      "success_pct": 46.8,
      "retrieval_missed_needed_tool": 76
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 85,
      "success_pct": 67.5,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 72,
      "success_pct": 41.4,
      "retrieval_missed_needed_tool": 72
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 32,
      "success_pct": 71.1,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 17,
      "success_pct": 70.8,
      "retrieval_missed_needed_tool": 2
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 53,
      "success_pct": 65.4,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 55,
      "success_pct": 36.7,
      "retrieval_missed_needed_tool": 70
    },
    {
      "model": "all-models",
      "strategy": "semantic-nim@10",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 172,
      "success_pct": 57.3,
      "retrieval_missed_needed_tool": 67
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 48,
      "success_pct": 69.6,
      "retrieval_missed_needed_tool": 12
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 124,
      "success_pct": 53.7,
      "retrieval_missed_needed_tool": 55
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 85,
      "success_pct": 67.5,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 87,
      "success_pct": 50.0,
      "retrieval_missed_needed_tool": 51
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 31,
      "success_pct": 68.9,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 17,
      "success_pct": 70.8,
      "retrieval_missed_needed_tool": 6
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 54,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 10
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 70,
      "success_pct": 46.7,
      "retrieval_missed_needed_tool": 45
    },
    {
      "model": "all-models",
      "strategy": "agent-search",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 155,
      "success_pct": 51.7,
      "retrieval_missed_needed_tool": 91
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 43,
      "success_pct": 62.3,
      "retrieval_missed_needed_tool": 16
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 112,
      "success_pct": 48.5,
      "retrieval_missed_needed_tool": 75
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 82,
      "success_pct": 65.1,
      "retrieval_missed_needed_tool": 19
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 73,
      "success_pct": 42.0,
      "retrieval_missed_needed_tool": 72
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 29,
      "success_pct": 64.4,
      "retrieval_missed_needed_tool": 8
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 14,
      "success_pct": 58.3,
      "retrieval_missed_needed_tool": 8
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 53,
      "success_pct": 65.4,
      "retrieval_missed_needed_tool": 11
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 59,
      "success_pct": 39.3,
      "retrieval_missed_needed_tool": 64
    },
    {
      "model": "all-models",
      "strategy": "agent-search-kw",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "all_multi_step",
      "subgroup": "all",
      "count": 300,
      "completed": 186,
      "success_pct": 62.0,
      "retrieval_missed_needed_tool": 47
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "1_category",
      "count": 69,
      "completed": 53,
      "success_pct": 76.8,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories",
      "subgroup": "2+_categories",
      "count": 231,
      "completed": 133,
      "success_pct": 57.6,
      "retrieval_missed_needed_tool": 44
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "2_ops",
      "count": 126,
      "completed": 87,
      "success_pct": 69.0,
      "retrieval_missed_needed_tool": 8
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "operations",
      "subgroup": "3+_ops",
      "count": 174,
      "completed": 99,
      "success_pct": 56.9,
      "retrieval_missed_needed_tool": 39
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|2_ops",
      "count": 45,
      "completed": 33,
      "success_pct": 73.3,
      "retrieval_missed_needed_tool": 3
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "1_category|3+_ops",
      "count": 24,
      "completed": 20,
      "success_pct": 83.3,
      "retrieval_missed_needed_tool": 0
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|2_ops",
      "count": 81,
      "completed": 54,
      "success_pct": 66.7,
      "retrieval_missed_needed_tool": 5
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "categories_x_operations",
      "subgroup": "2+_categories|3+_ops",
      "count": 150,
      "completed": 79,
      "success_pct": 52.7,
      "retrieval_missed_needed_tool": 39
    },
    {
      "model": "all-models",
      "strategy": "hybrid",
      "dimension": "classification_reliability",
      "subgroup": "flagged_tasks",
      "count": 0,
      "completed": 0,
      "success_pct": null,
      "retrieval_missed_needed_tool": 0
    }
  ],
  "total": {
    "canonical_input_tokens": 67279958,
    "canonical_output_tokens": 4216823,
    "canonical_model_tokens": 71496781
  }
}
