Multimodal Reasoning and Perception on RH-Bench
46.4Reasoning ScoreQwen2.5-VL-7B + ECRD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-VL-7B + ECRDECRD=true2026.02 | 46.4 | 57.1 | 58 | |
| Qwen2.5-VL-7B + supervisorsupervisor=true2026.02 | 42 | 53.3 | 54 | |
| Qwen2.5-VL-7Bbaseline=true2026.02 | 39.6 | 50.2 | 51 |