Captioning on RoboFine-Bench Hard setting
83.6Overall ScoreRoboFine-VLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RoboFine-VLM2026.05 | 83.6 | 81.9 | 75.3 | 93.7 | |
| GPT-5.42026.05 | 78.1 | 74.2 | 68.9 | 91.1 | |
| Gemini-3.1-Pro2026.05 | 77.2 | 77 | 61.3 | 93.4 | |
| Qwen3.5-Plus2026.05 | 72.5 | 70.9 | 56.8 | 89.7 | |
| Doubao-Seed-2.0-Pro2026.05 | 68.2 | 72.2 | 65.6 | 66.8 | |
| Qwen3-VL-Plus2026.05 | 65.1 | 68.7 | 57 | 69.6 |