Multi-Turn Tool-Integrated Reasoning (TIR) on AIME24 (Peak avg@32 score)
41.46Peak avg@32 scoreOTB
Evaluation Results
| Method | Links | |
|---|---|---|
| OTBBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 41.46 | |
| SimpleTIRBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 37.91 | |
| OGBBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 30.63 | |
| OPOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 29.9 | |
| GRPOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 27.6 | |
| RLOOBackbone=Qwen2.5-7B, Reasoning Paradigm=TIR2026.02 | 25 |