Reasoning Segmentation on ReasonSeg-R (test)
74.9gIoUStAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| StARBackbone=Qwen 3, Model Components=+ MV, Zero-shot=true2026.05 | 74.9 | 69.7 | |
| StARBackbone=Qwen 3, Zero-shot=true2026.05 | 73.8 | 69.2 | |
| B-GRTOZero-shot=true2026.05 | 72 | 66.6 | |
| GRPOZero-shot=true2026.05 | 70.9 | 68.1 | |
| GRTOZero-shot=true2026.05 | 70.9 | 63 | |
| StARBackbone=Qwen 2.5, Model Components=+ MV, Zero-shot=true2026.05 | 70.7 | 67.2 | |
| StARBackbone=Qwen 2.5, Zero-shot=true2026.05 | 69.7 | 66.2 | |
| StARBackbone=Qwen 2.5, Training Stage=stage-1, Zero-shot=true2026.05 | 69 | 65.6 | |
| GRPOTraining Stage=stage-1, Zero-shot=true2026.05 | 69 | 64.7 | |
| VisionReasonerZero-shot=true2026.05 | 64.8 | 56.8 | |
| SAM 3 AgentBackbone=7B, Zero-shot=true2026.05 | 63.1 | 58 | |
| LISAZero-shot=true2026.05 | 52.5 | 53.3 |