Scientific Reasoning on SciKnowEval Material
71.48Mean@16GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOModel Size=Qwen3-4B, Training Domain=Physics2026.05 | 71.48 | |
| GRPOModel Size=Qwen3-8B, Training Domain=Chemistry2026.05 | 64.18 | |
| GRPOModel Size=Qwen3-8B, Training Domain=Physics2026.05 | 63.76 | |
| ROSDModel Size=Qwen3-4B, Training Domain=Physics2026.05 | 62.21 | |
| ROSDModel Size=Qwen3-8B, Training Domain=Physics2026.05 | 60.33 | |
| GRPOModel Size=Qwen3-8B, Training Domain=Biology2026.05 | 59.04 | |
| GRPOModel Size=Qwen3-8B, Training Domain=ToolUse2026.05 | 58.82 | |
| GRPOModel Size=Qwen3-4B, Training Domain=Biology2026.05 | 56.16 | |
| ROSDModel Size=Qwen3-8B, Training Domain=Biology2026.05 | 56.01 | |
| ROSDModel Size=Qwen3-8B, Training Domain=ToolUse2026.05 | 55.87 | |
| ROSDModel Size=Qwen3-4B, Training Domain=ToolUse2026.05 | 53.3 | |
| GRPOModel Size=Qwen3-4B, Training Domain=ToolUse2026.05 | 52.7 | |
| ROSDModel Size=Qwen3-4B, Training Domain=Biology2026.05 | 51.37 | |
| ROSDModel Size=Qwen3-8B, Training Domain=Chemistry2026.05 | 50.22 | |
| GRPOModel Size=Qwen3-4B, Training Domain=Chemistry2026.05 | 49.78 | |
| SDPOModel Size=Qwen3-8B, Training Domain=Physics2026.05 | 49.38 | |
| SDPOModel Size=Qwen3-8B, Training Domain=ToolUse2026.05 | 44.22 | |
| SDPOModel Size=Qwen3-4B, Training Domain=Physics2026.05 | 43.62 | |
| SDPOModel Size=Qwen3-4B, Training Domain=Biology2026.05 | 38.16 | |
| SDPOModel Size=Qwen3-8B, Training Domain=Biology2026.05 | 38.1 | |
| ROSDModel Size=Qwen3-4B, Training Domain=Chemistry2026.05 | 34.64 | |
| SDPOModel Size=Qwen3-8B, Training Domain=Chemistry2026.05 | 22.89 | |
| SDPOModel Size=Qwen3-4B, Training Domain=Chemistry2026.05 | 20.41 | |
| SDPOModel Size=Qwen3-4B, Training Domain=ToolUse2026.05 | 1.09 |