Multimodal Tool-use Reasoning on TRBench Total
83.21AccuracyQwen3-VL-235B-thinking + UCT
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-VL-235B-thinking + UCTCate.=Tool-based2026.02 | 83.21 | |
| Gemini-2.5-pro + UCTCate.=Tool-based2026.02 | 73.41 | |
| CRAFTCate.=Tool-based2026.02 | 69.13 | |
| CREATORCate.=Tool-based2026.02 | 67.47 | |
| Qwen3-VL-235B-thinkingCate.=Basic-COT2026.02 | 60.17 | |
| Gemini-2.5-proCate.=Basic-COT2026.02 | 52.35 | |
| VanillaCate.=Tool-based2026.02 | 49.22 |