Mathematical Reasoning on AIME24 (Pass@1, Pass@8, #Invoc.)
73.47Pass@1DART
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DARTBackbone=Qwen3-4B-Thinking-2507, Paradigm=RL, CoT Type=tool-integrated long2026.01 | 73.47 | 91.37 | 10.68 | |
| ToRL-ThinkBackbone=Qwen3-4B-Thinking-2507, Paradigm=RL, CoT Type=pure-text long2026.01 | 71.25 | 82.14 | 0 | |
| UTIRBackbone=Qwen3-4B-Thinking-2507, Paradigm=RL, CoT Type=tool-integrated long2026.01 | 70.56 | 87.61 | 10.08 | |
| DAPOBackbone=Qwen3-4B-Thinking-2507, Paradigm=RL, CoT Type=pure-text long2026.01 | 69.17 | 85.67 | 0 | |
| UTIRBackbone=Qwen3-8B, Paradigm=RL, CoT Type=tool-integrated long2026.01 | 69.12 | 84.59 | 11.97 | |
| DARTBackbone=Qwen3-8B, Paradigm=RL, CoT Type=tool-integrated long2026.01 | 68.89 | 86.81 | 12.82 | |
| DAPOBackbone=Qwen3-8B, Paradigm=RL, CoT Type=pure-text long2026.01 | 66.11 | 81.44 | 0 | |
| ToRL-ThinkBackbone=Qwen3-8B, Paradigm=RL, CoT Type=pure-text long2026.01 | 65.7 | 82.68 | 0 | |
| Original checkpointBackbone=Qwen3-8B, CoT Type=pure-text long2026.01 | 64.17 | 79.28 | 0 | |
| Original checkpointBackbone=Qwen3-4B-Thinking-2507, CoT Type=pure-text long2026.01 | 52.22 | 71.93 | 0 | |
| STARTBackbone=Qwen3-4B-Thinking-2507, Paradigm=SFT, CoT Type=tool-integrated long2026.01 | 44.31 | 65.15 | 13.3 | |
| ToRL-BaseBackbone=Qwen3-8B, Paradigm=RL, CoT Type=tool-integrated short2026.01 | 42.64 | 54.95 | 3.86 | |
| ToRL-BaseBackbone=Qwen3-4B-Thinking-2507, Paradigm=RL, CoT Type=tool-integrated short2026.01 | 40.49 | 50.73 | 3.12 | |
| STARTBackbone=Qwen3-8B, Paradigm=SFT, CoT Type=tool-integrated long2026.01 | 35.83 | 58.22 | 15.08 | |
| RetoolBackbone=Qwen3-8B, Paradigm=SFT+RL, CoT Type=tool-integrated short2026.01 | 25.41 | 46.73 | 1.67 | |
| RetoolBackbone=Qwen3-4B-Thinking-2507, Paradigm=SFT+RL, CoT Type=tool-integrated short2026.01 | 23.33 | 35.46 | 1.85 | |
| Retool (only SFT)Backbone=Qwen3-8B, Paradigm=SFT, CoT Type=tool-integrated short2026.01 | 22.78 | 43.46 | 2 | |
| Retool (only SFT)Backbone=Qwen3-4B-Thinking-2507, Paradigm=SFT, CoT Type=tool-integrated short2026.01 | 17.36 | 34.14 | 2.12 |