Generalized Referring Image Segmentation on gRefCOCO (val)
78.4gIoUSETCON
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SETCON2026.05 | 78.4 | 72 | — | — | — | — | |
| MLLMSeg2026.05 | 75.1 | 71.6 | — | — | — | — | |
| Text4Seg2026.05 | 74.4 | 69.1 | — | — | — | — | |
| SAM4MLLM2026.05 | 71.9 | 67.8 | — | — | — | — | |
| SSP-SAMModel architecture=SAM based, Input resolution=3362026.03 | 69.09 | 62.56 | 66.12 | — | — | — | |
| SSP-SAMResolution=3362026.03 | 69.09 | 62.56 | — | 44.77 | 61.54 | 67.33 | |
| SSP-SAMModel architecture=SAM based, Input resolution=2242026.03 | 68.87 | 61.28 | 69.37 | — | — | — | |
| SSP-SAMResolution=2242026.03 | 68.87 | 61.28 | — | 44 | 60.55 | 66.49 | |
| GSVA2026.05 | 66.5 | 63.3 | — | — | — | — | |
| GSVA-Vicuna-7BModel architecture=SAM based, Fine-tuning=true2026.03 | 66.47 | 63.29 | 62.43 | — | — | — | |
| ReLAModel architecture=Non-SAM based2026.03 | 63.6 | 62.42 | 56.37 | — | — | — | |
| ReLA2026.03 | 63.6 | 62.42 | — | 24.68 | 68.33 | 74.2 | |
| LISA-Llama2-13BBackbone=Llama2-13B2026.05 | 63.5 | 63 | — | — | — | — | |
| GSVA-Vicuna-7BModel architecture=SAM based, Fine-tuning=false2026.03 | 63.32 | 61.7 | 56.45 | — | — | — | |
| CGFormerModel architecture=Non-SAM based2026.03 | 63.01 | 62.28 | — | — | — | — | |
| CGFormer2026.03 | 63.01 | 62.28 | — | 22.43 | 56.57 | 68.93 | |
| LISA-Vicuna-7BModel architecture=SAM based, Fine-tuning=true2026.03 | 61.63 | 61.76 | 54.67 | — | — | — | |
| LISA-7BBackbone=7B2026.05 | 61.6 | 61.8 | — | — | — | — | |
| SAM3 Agent2026.05 | 59.2 | 49.1 | — | — | — | — | |
| LAVTModel architecture=Non-SAM based2026.03 | 58.4 | 57.64 | 49.32 | — | — | — | |
| CRISModel architecture=Non-SAM based2026.03 | 56.27 | 55.34 | — | — | — | — | |
| LTSModel architecture=Non-SAM based2026.03 | 52.7 | 52.3 | — | — | — | — | |
| VLTModel architecture=Non-SAM based2026.03 | 52 | 52.51 | 47.17 | — | — | — | |
| MattNetModel architecture=Non-SAM based2026.03 | 48.24 | 47.51 | 41.15 | — | — | — | |
| Visionreasoner2026.05 | 41.5 | 48.3 | — | — | — | — | |
| LISA-Vicuna-7BModel architecture=SAM based, Fine-tuning=false2026.03 | 32.21 | 38.72 | 2.71 | — | — | — |