7.7 MB

Ctrl+K

1 contributor

History: 24 commits

siddhm11

Honest test: AUC=0.9053 without position leak

61371b7 verified 3 months ago

baseline_comparison.json

820 Bytes
Add baseline_comparison.json 3 months ago
eval_metrics.json

5.5 kB
Add eval_metrics.json 3 months ago
eval_metrics_exa_v1.json

2.97 kB
Exa model eval metrics 3 months ago
eval_metrics_final.json

2.54 kB
Final model metrics 3 months ago
eval_metrics_v2.json

4.91 kB
v2 model: production_model/eval_metrics_v2.json 3 months ago
eval_metrics_v3.json

2.23 kB
Add V3 eval metrics: nDCG@10=0.9494, hard_neg_auc=0.7380 3 months ago
eval_metrics_v4.json

3.06 kB
Add V4 eval metrics 3 months ago
eval_metrics_v5.json

2.46 kB
Add V5 eval metrics: nDCG=1.0, HN-AUC=0.837 3 months ago
eval_metrics_v6.json

2.88 kB
Add V6 eval results 3 months ago
eval_metrics_v6_production.json

2.89 kB
Add V6 production results 3 months ago
feature_importance.csv

1.43 kB
Add feature_importance.csv 3 months ago
feature_importance_v2.csv

1.55 kB
v2 model: production_model/feature_importance_v2.csv 3 months ago
feature_schema.json

1.12 kB
Add feature_schema.json 3 months ago
honest_test_no_position.json

123 Bytes
Honest test: AUC=0.9053 without position leak 3 months ago
reranker_exa_v1.txt

1.31 MB
Exa-distilled model: nDCG=0.9999, HN-AUC=0.9731, trained on 93K pairs 3 months ago
reranker_production_final.txt

1.39 MB
FINAL production model: intent-based labels, nDCG=0.8725, HN-AUC=0.8325 3 months ago
reranker_v1.txt

971 kB
Add reranker_v1.txt 3 months ago
reranker_v2.txt

361 kB
v2 model: production_model/reranker_v2.txt 3 months ago
reranker_v3.txt

10.2 kB
Add V3 LightGBM: trained on cross-survey authority labels, hard_neg_auc=0.738 3 months ago
reranker_v4.txt

17.8 kB
Add V4 LightGBM: 25 graph features, hard_neg_auc=0.778 (+5.4% over V3) 3 months ago
reranker_v5.txt

2.81 MB
Add V5: graph+metadata, Hard Neg AUC=0.837 (+7.6% over V4) 3 months ago
reranker_v6.txt

568 kB
Add V6: drop-in replacement, 37 features, cross-survey labels, HN-AUC=0.964 3 months ago
reranker_v6_production.txt

224 kB
Add V6 PRODUCTION MODEL: 37-feature schema, drops into app, Hard Neg AUC=0.758 3 months ago