{
  "passed": true,
  "source_run": "reports/random20-seed20261015",
  "checked_cases": 320,
  "groups": [
    {
      "model": "google/gemini-3.8-flash",
      "backend": "lancedb",
      "method": "bm25",
      "scored_cases": 20,
      "mean_answer_seconds": 8.516314306200048,
      "mean_retrieval_seconds": 0.01933659629989961
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "lancedb",
      "method": "vector",
      "scored_cases": 20,
      "mean_answer_seconds": 7.367952985200145,
      "mean_retrieval_seconds": 0.02209555005006223
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "lancedb",
      "method": "hybrid",
      "scored_cases": 20,
      "mean_answer_seconds": 10.311845652200054,
      "mean_retrieval_seconds": 0.03563654764993771
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "lancedb",
      "method": "agentic",
      "scored_cases": 20,
      "mean_answer_seconds": 27.033250321050083,
      "mean_retrieval_seconds": 18.38452256985024
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "qdrant",
      "method": "bm25",
      "scored_cases": 20,
      "mean_answer_seconds": 8.230327031099979,
      "mean_retrieval_seconds": 0.007952243850013474
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "qdrant",
      "method": "vector",
      "scored_cases": 20,
      "mean_answer_seconds": 5.810627831200145,
      "mean_retrieval_seconds": 0.055463029450129396
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "qdrant",
      "method": "hybrid",
      "scored_cases": 20,
      "mean_answer_seconds": 7.030186159099958,
      "mean_retrieval_seconds": 0.03435491230020489
    },
    {
      "model": "google/gemini-3.8-flash",
      "backend": "qdrant",
      "method": "agentic",
      "scored_cases": 20,
      "mean_answer_seconds": 35.36996802374988,
      "mean_retrieval_seconds": 23.454901457300092
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "lancedb",
      "method": "bm25",
      "scored_cases": 20,
      "mean_answer_seconds": 4.102670281450128,
      "mean_retrieval_seconds": 0.01933659629989961
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "lancedb",
      "method": "vector",
      "scored_cases": 20,
      "mean_answer_seconds": 6.166311452000082,
      "mean_retrieval_seconds": 0.02209555005006223
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "lancedb",
      "method": "hybrid",
      "scored_cases": 20,
      "mean_answer_seconds": 4.261235679399943,
      "mean_retrieval_seconds": 0.03563654764993771
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "lancedb",
      "method": "agentic",
      "scored_cases": 20,
      "mean_answer_seconds": 17.05526517119988,
      "mean_retrieval_seconds": 18.242840221600044
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "qdrant",
      "method": "bm25",
      "scored_cases": 20,
      "mean_answer_seconds": 5.738852074900114,
      "mean_retrieval_seconds": 0.007952243850013474
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "qdrant",
      "method": "vector",
      "scored_cases": 20,
      "mean_answer_seconds": 5.7358261237499395,
      "mean_retrieval_seconds": 0.055463029450129396
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "qdrant",
      "method": "hybrid",
      "scored_cases": 20,
      "mean_answer_seconds": 5.947256503049994,
      "mean_retrieval_seconds": 0.03435491230020489
    },
    {
      "model": "openai/gpt-6-luna",
      "backend": "qdrant",
      "method": "agentic",
      "scored_cases": 20,
      "mean_answer_seconds": 5.671115662700004,
      "mean_retrieval_seconds": 8.683418138049955
    }
  ],
  "definitions": {
    "answer": "Recorded answer.seconds; for agentic cases, sum of model-call durations.",
    "retrieval": "Recorded retrieval.seconds; cached search latency for BM25/Vector/Hybrid, cumulative tool duration for Agentic Search.",
    "exclusions": "Judge duration and embedding-preparation time are excluded from these stage timings.",
    "shared_cache": "Non-agentic retrieval was precomputed once per question/database/method and shared by both answer models."
  }
}
