Question Answering on HotQA
48.5AccuracyCDKC
Evaluation Results
| Method | Links | |
|---|---|---|
| CDKCBackbone=Qwen2.5-14B-Instruct2026.02 | 48.5 | |
| GRPOBackbone=Qwen2.5-14B-Instruct2026.02 | 48.17 | |
| CDKCBackbone=Qwen2.5-3B-Instruct2026.02 | 40.07 | |
| GRPOBackbone=Qwen2.5-3B-Instruct2026.02 | 38.67 | |
| RAGBackbone=Qwen2.5-14B-Instruct2026.02 | 38.6 | |
| CGKEBackbone=Qwen2.5-14B-Instruct2026.02 | 33.27 | |
| CoTBackbone=Qwen2.5-14B-Instruct2026.02 | 32.9 | |
| RAGBackbone=Qwen2.5-3B-Instruct2026.02 | 32.57 | |
| Vanilla SFTBackbone=Qwen2.5-14B-Instruct2026.02 | 28 | |
| Vanilla LLMBackbone=Qwen2.5-14B-Instruct2026.02 | 26.4 | |
| CGKEBackbone=Qwen2.5-3B-Instruct2026.02 | 25.47 | |
| CoTBackbone=Qwen2.5-3B-Instruct2026.02 | 21.43 | |
| Vanilla SFTBackbone=Qwen2.5-3B-Instruct2026.02 | 19.9 | |
| Vanilla LLMBackbone=Qwen2.5-3B-Instruct2026.02 | 16.9 |