Scientific Reasoning on SciKnowEval (test)
65.38Biology AccuracyVPD
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VPDBackbone=QWEN3-8B, Feedback Source=autonomous self-critique2026.05 | 65.38 | 83.52 | 84.2 | 79.45 | 78.14 | |
| VPDBackbone=Qwen3-1.7B, Feedback Source=autonomous self-critique2026.05 | 64.75 | 79.43 | 73.54 | 70.31 | 72.01 | |
| SDPOBackbone=Qwen3-1.7B, Feedback Source=autonomous self-critique2026.05 | 62.12 | 78.66 | 71.21 | 58.13 | 67.53 | |
| SDPOBackbone=QWEN3-8B, Feedback Source=autonomous self-critique2026.05 | 61.38 | 82.35 | 83.71 | 72.03 | 74.87 |