Expected Calibration Error on MusQ
48.08ECECDKC
Evaluation Results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.02 | 48.08 | |
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 53.01 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 54.79 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 76.15 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 78.61 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 80.34 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 80.68 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 80.83 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 81.61 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 85.05 |