Expected Calibration Error on 2Wiki
17.43ECECDKC
Evaluation Results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.02 | 17.43 | |
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 20.44 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 20.94 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 37.74 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 43.79 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 47.56 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 48.11 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 49.01 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 57.87 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 59.36 |