Expected Calibration Error on BeerQA
23.28ECECDKC
Evaluation results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-7B-Instruct, Optimization Rounds=22026.02 | 23.28 | |
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 28.74 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 29.15 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 55.95 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 59.5 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 63.73 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 63.73 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 63.93 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 64.18 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 66.89 |