{ "dataset": "hotpot", "questions": 600, "candidate_count_per_question": 10, "selection_rule": "same neural_evidence_graph_core.graph_arrays and graph_select as four new datasets: 0.4 * per-query minmax relevance + 0.5 * max selected-to-candidate directed title-body link + 0.2 * query-title match; no tuning, no query facets, no specialized presets", "embedding_source": "previously frozen/precomputed actual all-MiniLM-L6-v2 q and passage vectors; cosine recomputed", "crossencoder_source": "previously frozen/precomputed actual pretrained ms-marco-MiniLM-L6-v2 256-wordpiece query-passage logits, same ids", "complete_evidence_coverage_pct": { "dense_cosine": { "2": 33.5, "3": 50.67, "5": 68.0, "8": 85.67 }, "crossencoder": { "2": 45.83, "3": 60.83, "5": 74.67, "8": 86.83 }, "dense_graph": { "2": 61.67, "3": 74.83, "5": 84.83, "8": 94.17 }, "crossencoder_graph": { "2": 67.17, "3": 80.5, "5": 89.83, "8": 95.5 } }, "crossencoder_graph_vs_crossencoder_at3": { "diff_pp": 19.67, "ci95_pp": [ 16.17, 23.33 ], "wins": 135, "losses": 17, "ties": 448 }, "dense_graph_vs_dense_at3": { "diff_pp": 24.17, "ci95_pp": [ 20.33, 28.17 ], "wins": 163, "losses": 18, "ties": 419 }, "graph_link_present_pct": 87.83, "gold_support_count_distribution": { "2": 600 }, "timing": { "graph_feature_extraction_and_four_rankings_ms_per_question": 0.406, "graph_feature_extraction_and_four_rankings_total_s": 0.2437, "pretrained_crossencoder_scoring_ms_per_question_from_original_inference_run": 272.18, "cached_embedding_and_crossencoder_scores": true }, "note": "This is an actual rerun of ranking/selection and evidence coverage from cached neural scores; scoring step was previously executed, not repeated. Scores are on the same selected 600 held-out question IDs." }