Expected Calibration Error on TriQA
8.58Expected Calibration ErrorCDKC
Evaluation results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 8.58 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 10.26 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 13.78 | |
| CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.02 | 14.57 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 31.43 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 34.96 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 35.45 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 35.71 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 36.49 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 37.32 |