Calibration on TREC
0.0086CB-ECENLE (conditional)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NLE (conditional)Explanation Type=conditional, LLM=LLaMA2-13B-Chat2024.02 | 0.0086 | 0.428 | 0.832 | — | — | |
| NLE (literal)Explanation Type=literal, LLM=LLaMA2-13B-Chat2024.02 | 0.01 | 0.067 | 0.602 | — | — | |
| PRCalibration type=Platt Ranking2024.02 | 0.01 | 1.246 | 2.137 | — | — | |
| FC (monoBERT)Calibration type=Full Calibration, Backbone=monoBERT2024.02 | 0.0109 | 0.221 | 0.721 | — | — | |
| PCCalibration type=Post-hoc Calibration2024.02 | 0.0114 | 0.125 | 0.684 | — | — | |
| PLCalibration type=Platt Linear2024.02 | 0.0117 | 0.145 | 0.673 | — | — | |
| NCCalibration type=No Calibration2024.02 | 0.0121 | 0.32 | 0.773 | — | — | |
| FC (BERT)Calibration type=Full Calibration, Backbone=BERT2024.02 | 0.0125 | 0.266 | 0.785 | — | — | |
| AD-KDTeacher Training=Fine-tuned on Banking77, Evaluation Context=Wrong predictions only2026.02 | — | — | — | 0.95 | 1.9 | |
| CKDTeacher Training=Fine-tuned on Banking77, Evaluation Context=Wrong predictions only2026.02 | — | — | — | 0.87 | 1.675 | |
| CUDTeacher Training=Fine-tuned on Banking77, Evaluation Context=Wrong predictions only2026.02 | — | — | — | 0.566 | 0.993 | |
| TinyBERTTeacher Training=Fine-tuned on Banking77, Evaluation Context=Wrong predictions only2026.02 | — | — | — | 0.929 | 1.851 |