Active Visual Search on SAT (real)
77.33AccuracySFT + RL (Ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| SFT + RL (Ours)Backbone=Qwen2.5-VL-7B, AVS Framework=SFT + RL2025.12 | 77.33 | |
| SFTBackbone=Qwen2.5-VL-7B, AVS Framework=SFT2025.12 | 67.33 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2025.12 | 60 | |
| RLBackbone=Qwen2.5-VL-7B, AVS Framework=RL2025.12 | 57.33 |