Multi-image Understanding on Blink
61.18AccuracyOurs
Evaluation Results
| Method | Links | |
|---|---|---|
| OursTraining=RL-enhanced2025.07 | 61.18 | |
| Qwen2.5-VL-7BModel Size=7B2025.07 | 58.07 | |
| Qwen2.5-VL-7B + SFTTraining=SFT2025.07 | 56.56 | |
| LLaVA2026.03 | 40.4 | |
| VISOR2026.03 | 39 |
Loading the SOTA2 catalog…