Multimodal Tool-use Reasoning on TRBench General sub-dataset
46Cultural Understanding ScoreGemini-2.5-pro
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-2.5-proCate.=Basic-COT2026.02 | 46 | 70.83 | 66.67 | 63.27 | |
| Qwen3-VL-235B-thinking + UCTCate.=Tool-based2026.02 | 46 | 87.5 | 65.31 | 65.71 | |
| Gemini-2.5-pro + UCTCate.=Tool-based2026.02 | 40 | 79.17 | 68.03 | 64.49 | |
| VanillaCate.=Tool-based2026.02 | 38 | 62.5 | 54.42 | 52.65 | |
| CRAFTCate.=Tool-based2026.02 | 36 | 83.33 | 48.98 | 48.98 | |
| Qwen3-VL-235B-thinkingCate.=Basic-COT2026.02 | 32 | 58.33 | 50.34 | 48.16 | |
| CREATORCate.=Tool-based2026.02 | 28 | 77.08 | 45.58 | 48.16 |