Multi-Turn Tool-Integrated Reasoning (TIR) on AIME25 (Peak avg@32 score)
28.13Peak avg@32 ScoreOTB
Evaluation Results
| Method | Links | |
|---|---|---|
| OTBBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 28.13 | |
| SimpleTIRBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 26.67 | |
| OGBBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 21.15 | |
| OPOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 20.83 | |
| RLOOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 20.1 | |
| GRPOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 18.54 |