Hugging Face's logo

nhnk
/
FactCascade

English
factuality-verification
routing
cascading
gradient-boosting
numpy
factcascade
Model card Files
xet
 Community
FactCascade
File size: 2,929 Bytes
08bd580
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
{
  "schema": "FACTCASCADE_HF_ROUTER_CONFIG_V1",
  "name": "FactCascade",
  "checkpoint": "router.json",
  "feature_count": 31,
  "feature_order": [
    "ft5_score",
    "ft5_abs_distance_to_calibrated_threshold",
    "b192_score",
    "ft5_b192_delta",
    "ft5_b192_disagreement",
    "table_or_mixed_indicator",
    "ft5_margin_x_table",
    "ft5_full_score",
    "selected_full_score_delta",
    "selected_full_disagreement",
    "log_ft5_selected_chunks",
    "log_ft5_full_chunks",
    "log_claim_word_count",
    "log_selected_evidence_words",
    "log_full_evidence_words",
    "log_selected_source_count",
    "selected_source_coverage_ratio",
    "ft5_selected_margin_to_050",
    "ft5_selected_margin_to_0763",
    "ft5_full_minus_selected",
    "b192_minus_ft5_selected",
    "b192_ft5_decision_disagreement",
    "full_only_support_signal",
    "b192_disagreement_x_table_or_mixed",
    "selected_full_discrepancy_x_evidence_length_ratio",
    "claim_negation_proxy_count",
    "claim_comparator_quantifier_count",
    "claim_numeral_unit_count",
    "claim_titlecase_entity_proxy_count",
    "claim_selected_number_overlap_ratio",
    "claim_entity_selected_minus_full_overlap"
  ],
  "learner": {
    "max_depth": 2,
    "max_iter": 100,
    "learning_rate": 0.05,
    "min_samples_leaf": 30,
    "max_bins": 255,
    "random_state": 0
  },
  "training_target": "unweighted signed replacement utility: correctness(Bespoke)-correctness(fallback)",
  "training_rows": 3337,
  "training_groups": 1397,
  "policy": {
    "utility_weight": 0.5,
    "disagreement_weight": 0.5,
    "rank": "empirical midrank within evaluation component",
    "disagreement": "abs(FT5_selected_score-B192_score)",
    "quota_numerator": 920,
    "quota_denominator": 3337,
    "quota_rounding": "integer half-up per component",
    "ties": "ascending existing row key"
  },
  "thresholds": {
    "FT5_fallback": {
      "operator": ">",
      "value": 0.763
    },
    "FT5_standalone": {
      "operator": ">",
      "value": 0.5
    },
    "B192": {
      "operator": ">=",
      "value": 0.398
    },
    "Bespoke": {
      "operator": ">",
      "value": 0.5
    }
  },
  "training_metric_alignment": "Per-row squared-loss surrogate; does not directly optimize component-macro balanced accuracy",
  "selection_role": "Final fusion chosen after observed DEV and external results; retrospective for those exposed cohorts",
  "verifier_dependencies": [
    {
      "repo": "lytang/MiniCheck-Flan-T5-Large",
      "revision": "a496016e7b493686ed6e1c52250b9b9d39b0dcb2"
    },
    {
      "repo": "lytang/MiniCheck-DeBERTa-v3-Large",
      "revision": "2f2d01a54fa022a7ffadb76260e1ea8bc88c82bb"
    },
    {
      "repo": "bespokelabs/Bespoke-MiniCheck-7B",
      "revision": "17b48556b0a54218e8fc2f0dbf0c50fa65a0cfba"
    }
  ],
  "verifier_weights_included": false,
  "requires_native_verifier_input_geometry_for_full_pipeline": true
}