{
  "audit": "passed",
  "queries": 300,
  "document_judgments": 6000,
  "saved_decision_responses": 6000,
  "errors": 0,
  "resolved_models": [
    "jev-1.13.0"
  ],
  "metrics_independently_recomputed": {
    "ndcg_at_10": 0.7676260618846087,
    "hit_at_1": 0.68,
    "hit_at_3": 0.8033333333333333,
    "hit_at_20": 0.87,
    "recall_at_20": 0.8569444444444444
  },
  "bm25_metrics_independently_recomputed": {
    "ndcg_at_10": 0.6863002612489066,
    "hit_at_1": 0.55,
    "hit_at_3": 0.7266666666666667,
    "hit_at_20": 0.87,
    "recall_at_20": 0.8569444444444444
  },
  "input_tokens": 4109959,
  "output_tokens": 132000,
  "estimated_total_usd": "0.172618278",
  "estimated_usd_per_1000_queries": "0.57539426000",
  "cost_rate_usd_per_input_token": "0.000000042",
  "output_cost_rate": 0,
  "pricing_is_estimate_not_invoice": true,
  "maximum_per_query_metric_difference": 0,
  "source_checksums_verified": true,
  "full_qrels_test_split_verified": true,
  "candidate_texts_queries_and_qrels_verified_against_dataset": true,
  "all_saved_rankings_verified_against_raw_predicate_probabilities": true,
  "sdk_head": "5aa0a8304cd5bfa095bbd07e77f6f8c367084595",
  "sdk_head_tracked_changes_recorded": false,
  "notes": [
    "6000 saved successful document judgments; no separate raw HTTP request log in full-run artifact. Runner disables local caching/retries and sends each pointwise document separately.",
    "Cost is recomputed from returned input tokens and the configured gateway model rate, not from 6000 independently recorded billing headers. Separate live smoke reconciled that rate against actual gateway headers.",
    "Single dataset and one run; no Perplexity or Cohere benchmark was measured in this artifact."
  ]
}
