Reasoning Segmentation on ReasonSeg-X (val)
62.6gIoUStAR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| StARBackbone=Qwen 3, Model Components=+ MV2026.05 | 62.6 | 60.1 | |
| StARBackbone=Qwen 32026.05 | 60.8 | 57.2 | |
| B-GRTO2026.05 | 52.1 | 53.7 | |
| StARBackbone=Qwen 2.5, Model Components=+ MV2026.05 | 51.3 | 48.4 | |
| GRPO2026.05 | 51.3 | 49.9 | |
| GRTO2026.05 | 51.2 | 51.3 | |
| StARBackbone=Qwen 2.52026.05 | 50.5 | 48 | |
| StARBackbone=Qwen 2.5, Training Stage=stage-12026.05 | 48.5 | 42.7 | |
| GRPOTraining Stage=stage-12026.05 | 44.5 | 39 | |
| VisionReasoner2026.05 | 44.1 | 37.6 | |
| SAM 3 AgentBackbone=7B2026.05 | 34.1 | 28 | |
| LISA2026.05 | 23.8 | 24.3 |