{
  "source_file": "ask_eval_A3B-window_d2.jsonl",
  "source_sha256": "50847bb6d3d226fa6ec3399aa055d777e0ae52b00ebbfe100866e64b3085551d",
  "questions": 50,
  "failed": 0,
  "model": "Qwen3.6-35B-A3B-MLX-8bit",
  "dial": 2,
  "seed": 20260910,
  "source_match": "Same episode; the retrieved chunk's time span (its start to the next chunk's start) overlaps the recorded answer span padded by 90 seconds on each side.",
  "source_match_start_rule": "The earlier rule, kept for comparison: same episode; the chunk's start timestamp within the padded answer span.",
  "source_shown": "The text the model saw shares a run of words with the recorded question and answer over 2x the longest run it shares with any other sampled question's text.",
  "metrics": {
    "source_in_floor": {
      "count": 35,
      "total": 50,
      "rate": 0.7,
      "wilson_95": [
        0.562493877226366,
        0.8089661091039063
      ]
    },
    "source_in_floor_start": {
      "count": 24,
      "total": 50,
      "rate": 0.48,
      "wilson_95": [
        0.3479691239550571,
        0.6148848774119157
      ]
    },
    "source_cited": {
      "count": 34,
      "total": 50,
      "rate": 0.68,
      "wilson_95": [
        0.5418944286372394,
        0.792419559060006
      ]
    },
    "surplus_seated": {
      "count": 46,
      "total": 50,
      "rate": 0.92,
      "wilson_95": [
        0.8116149745493629,
        0.9684509967442093
      ]
    },
    "source_shown": {
      "count": 35,
      "total": 50,
      "rate": 0.7,
      "wilson_95": [
        0.562493877226366,
        0.8089661091039063
      ]
    },
    "source_cited_given_in_floor": {
      "count": 34,
      "total": 35,
      "rate": 0.9714285714285714,
      "wilson_95": [
        0.8546659090672868,
        0.9949386386341466
      ]
    },
    "surplus_cited_given_seated": {
      "count": 2,
      "total": 46,
      "rate": 0.043478260869565216,
      "wilson_95": [
        0.012005193588830829,
        0.1453252291905342
      ]
    },
    "n_unsupported_flagged": {
      "count": 46,
      "total": 50,
      "rate": 0.92,
      "wilson_95": [
        0.8116149745493629,
        0.9684509967442093
      ]
    },
    "n_attribution_flagged": {
      "count": 44,
      "total": 50,
      "rate": 0.88,
      "wilson_95": [
        0.7619492111337642,
        0.9438247628937535
      ]
    }
  },
  "partial_rerank_labels": 50,
  "runs_over_300_s": 2,
  "timing_seconds": {
    "retrieve": {
      "mean": 13.484,
      "median": 12.9,
      "max": 16.1
    },
    "synthesize": {
      "mean": 13.379999999999999,
      "median": 13.350000000000001,
      "max": 28.9
    },
    "verify": {
      "mean": 52.85,
      "median": 16.95,
      "max": 747.7
    },
    "total": {
      "mean": 79.72800000000001,
      "median": 46.650000000000006,
      "max": 772.6
    }
  },
  "limitations": [
    "Span and text matching are source-recovery proxies, not semantic support or answer correctness.",
    "Verifier counts are model-reported flags, not independently confirmed repairs.",
    "Every run's reranking was labeled partial: the 12 s budget scored part of the candidate pool.",
    "No connection-off comparison or independent human accuracy assessment is included."
  ]
}