Active Visual Search on SAT (synthetic)
69.33AccuracySFT
Evaluation Results
| Method | Links | |
|---|---|---|
| SFTBackbone=Qwen2.5-VL-7B, AVS Framework=SFT2025.12 | 69.33 | |
| SFT + RL (Ours)Backbone=Qwen2.5-VL-7B, AVS Framework=SFT + RL2025.12 | 69.33 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2025.12 | 59.11 | |
| RLBackbone=Qwen2.5-VL-7B, AVS Framework=RL2025.12 | 57.31 |