Multimodal Reasoning on HR-4K
72.4AccuracyMVH-RL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MVH-RLSetting=Our Model2026.05 | 72.4 | 69.7 | |
| MVH-SFTSetting=Our Model2026.05 | 71.5 | 67.8 | |
| Monet-RLSetting=Open-Source Baseline2026.05 | 71 | 67.6 | |
| Monet-SFTSetting=Open-Source Baseline2026.05 | 70.4 | 66.5 | |
| ILVR-Stage2Setting=Open-Source Baseline2026.05 | 70.3 | 67 | |
| Qwen2.5-VL-7B + SFT + GRPOSetting=Open-Source Baseline2026.05 | 70 | 66.9 | |
| ILVR-Stage1Setting=Open-Source Baseline2026.05 | 70 | 66.6 | |
| Qwen2.5-VL-7B + vanilla SFTSetting=Open-Source Baseline2026.05 | 68.4 | 65.8 | |
| Qwen2.5-VL-7BSetting=Open-Source Baseline2026.05 | 68 | 63.5 | |
| GPT-4oSetting=Proprietary Model2026.05 | 59 | 58.5 |