Scientific Question Answering on SciEval Static
81.52AccuracyReason, Reward, Refine
Evaluation Results
| Method | Links | |
|---|---|---|
| Reason, Reward, RefineModel=LLaMA 3.2 3B2026.07 | 81.52 | |
| Reason, Reward, RefineModel=Qwen 2.5 1.5B2026.07 | 79.29 | |
| RAGModel=Qwen 2.5 1.5B2026.07 | 68.12 | |
| Reason, Reward, RefineModel=LLaMA 3.2 1B2026.07 | 68.09 | |
| RAGModel=LLaMA 3.2 3B2026.07 | 66.87 | |
| Reason, Reward, RefineModel=Phi 3.5 Mini 3.8B2026.07 | 65.19 | |
| CoTModel=Qwen 2.5 1.5B2026.07 | 62.73 | |
| CoTModel=Phi 3.5 Mini 3.8B2026.07 | 62.43 | |
| CoTModel=LLaMA 3.2 3B2026.07 | 62.26 | |
| SFTModel=LLaMA 3.2 3B2026.07 | 61.59 | |
| RAGModel=Phi 3.5 Mini 3.8B2026.07 | 61.01 | |
| DPOModel=Qwen 2.5 1.5B2026.07 | 59.32 | |
| SFTModel=Phi 3.5 Mini 3.8B2026.07 | 59.17 | |
| SFTModel=Qwen 2.5 1.5B2026.07 | 57.93 | |
| DPOModel=LLaMA 3.2 3B2026.07 | 57.93 | |
| RAGModel=LLaMA 3.2 1B2026.07 | 57.37 | |
| DPOModel=LLaMA 3.2 1B2026.07 | 56.93 | |
| DPOModel=Phi 3.5 Mini 3.8B2026.07 | 56.87 | |
| SFTModel=LLaMA 3.2 1B2026.07 | 55.49 | |
| CoTModel=LLaMA 3.2 1B2026.07 | 44.51 |