Mathematical Reasoning on GPQA Diamond (average@8 score)
45.3Average@8 ScoreDeepTool
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepToolBackbone=Qwen2.5-7B, Training Stage=+RL2026.05 | 45.3 | |
| DeepToolBackbone=Qwen2.5-7B, Training Stage=SFT2026.05 | 44.8 | |
| DeepToolBackbone=Qwen3-4B, Training Stage=+RL2026.05 | 43.8 | |
| DeepToolBackbone=Qwen2.5-7B, Training Stage=w/o thinking2026.05 | 43.1 | |
| DeepToolBackbone=Qwen3-4B, Training Stage=SFT2026.05 | 37.4 | |
| DeepToolBackbone=Qwen3-4B, Training Stage=w/o thinking2026.05 | 34.5 | |
| Search-R1Backbone=Qwen2.5-7B, Training Stage=Search-based2026.05 | 33.3 | |
| Qwen3-4B-BaseBackbone=Qwen3-4B, Training Stage=Base2026.05 | 32.6 | |
| Qwen2.5-7B-BaseBackbone=Qwen2.5-7B, Training Stage=Base2026.05 | 32.4 | |
| Qwen2.5-7B-TIRBackbone=Qwen2.5-7B, Training Stage=TIR2026.05 | 28.5 |