Question Answering on BeerQA
44.27AccuracyCDKC
Evaluation Results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-14B-Instruct2026.02 | 44.27 | |
| GRPOBackbone=Qwen2.5-14B-Instruct2026.02 | 43.5 | |
| RAGBackbone=Qwen2.5-14B-Instruct2026.02 | 40.03 | |
| RAGBackbone=Qwen2.5-3B-Instruct2026.02 | 34.77 | |
| CDKCBackbone=Qwen2.5-3B-Instruct2026.02 | 34.47 | |
| GRPOBackbone=Qwen2.5-3B-Instruct2026.02 | 33.2 | |
| CGKEBackbone=Qwen2.5-14B-Instruct2026.02 | 29.07 | |
| CoTBackbone=Qwen2.5-14B-Instruct2026.02 | 27.6 | |
| Vanilla SFTBackbone=Qwen2.5-14B-Instruct2026.02 | 25.2 | |
| Vanilla LLMBackbone=Qwen2.5-14B-Instruct2026.02 | 23.7 | |
| CGKEBackbone=Qwen2.5-3B-Instruct2026.02 | 22.7 | |
| CoTBackbone=Qwen2.5-3B-Instruct2026.02 | 18.63 | |
| Vanilla SFTBackbone=Qwen2.5-3B-Instruct2026.02 | 18.07 | |
| Vanilla LLMBackbone=Qwen2.5-3B-Instruct2026.02 | 15.27 |