Fact Verification on FEVER (Accuracy and AVG)
73.73AccuracyCDKC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CDKCBackbone=Qwen2.5-7B-Instruct, Number of Rounds=22026.02 | 73.73 | 52.43 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 72.2 | 38.45 | |
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 69.13 | 47.52 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 69.03 | 46.53 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 62.13 | 30.4 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 61.23 | 27.1 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 61.2 | 35.12 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 60.93 | 30.99 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 59.43 | 24.22 | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.02 | 59.3 | 48.83 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 58.7 | 32.08 |