Reasoning Segmentation on FReasonSeg 1.0 (L1)
61.55gIoUCR-Seg
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CR-SegTrain=SFT/LoRA2026.06 | 61.55 | 51.69 | |
| Dr. Seg (LoRA)Train=RL/LoRA2026.06 | 61.04 | 58.74 | |
| LENSTrain=RL/Full FT2026.06 | 59.8 | 51.51 | |
| GSVATrain=SFT/LoRA2026.06 | 59.25 | 55.96 | |
| Seg-Zero-7BTrain=RL/Full FT2026.06 | 59.22 | 47.38 | |
| Evol-SAM3 (reproduced)Train=Agent, Backbone=Qwen3-VL-4B and SAM32026.06 | 58 | 50.32 | |
| SAM3-AgentTrain=Agent2026.06 | 56.65 | 44.97 | |
| Dr. SegTrain=RL/Full FT2026.06 | 55.64 | 45.46 | |
| VisionReasonerTrain=RL/Full FT2026.06 | 55.48 | 44.24 | |
| Evol-SAM3Train=Agent2026.06 | 54.46 | 52.52 | |
| SAM3-Agent (reproduced)Train=Agent, Backbone=Qwen3-VL-4B and SAM32026.06 | 51.02 | 42.54 | |
| LISA-7BTrain=SFT/LoRA2026.06 | 44.1 | 48.17 |