Reasoning Segmentation on FReasonSeg L3 1.0
80.48gIoUCR-Seg
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CR-SegTrain=SFT/LoRA2026.06 | 80.48 | 78.54 | |
| Dr. Seg (LoRA)Train=RL/LoRA2026.06 | 80.06 | 77.08 | |
| Dr. SegTrain=RL/Full FT2026.06 | 78.01 | 76.94 | |
| Seg-Zero-7BTrain=RL/Full FT2026.06 | 77.86 | 75.03 | |
| VisionReasonerTrain=RL/Full FT2026.06 | 76.84 | 73.15 | |
| LENSTrain=RL/Full FT2026.06 | 74.01 | 73.41 | |
| LISA-7BTrain=SFT/LoRA2026.06 | 70.33 | 70.98 | |
| Evol-SAM3Train=Agent2026.06 | 68.05 | 64.69 | |
| SAM3-Agent (reproduced)Train=Agent, Backbone=Qwen3-VL-4B and SAM32026.06 | 66.46 | 66.54 | |
| GSVATrain=SFT/LoRA2026.06 | 63.44 | 55.1 | |
| SAM3-AgentTrain=Agent2026.06 | 51.92 | 46.27 | |
| Evol-SAM3 (reproduced)Train=Agent, Backbone=Qwen3-VL-4B and SAM32026.06 | 51.91 | 40.23 |