Referring Segmentation on RefCOCOg (test)
81.9cIoUReChannel
Evaluation results
| Method | Links | |
|---|---|---|
| ReChannelParameters=9B2026.07 | 81.9 | |
| FCLMParameters=7B2026.07 | 80.5 | |
| TraceVision-7B2026.02 | 80.1 | |
| ReChannelParameters=4B2026.07 | 79.8 | |
| HyperSeg-1.5B2026.02 | 78.9 | |
| UniPixel (Qwne2.5-VL-3B)Additions=SAM Decoder2026.01 | 77 | |
| Youtu-VL (4B)Additions=None2026.01 | 76.6 | |
| UFO (InternVL2.5-8B)Additions=Mask Tokens2026.01 | 76.3 | |
| CoPRSParameters=7B2026.07 | 76.2 | |
| Text4SegParameters=8B2026.07 | 75.3 | |
| GLaMM (Vicuna-7B)Additions=SAM/Pixel Decoder2026.01 | 74.9 | |
| GLaMMParameters=7B2026.07 | 74.9 | |
| PSALM-3B2026.02 | 74.4 | |
| PSALMParameters=1.3B2026.07 | 74.4 | |
| SAM3Additions=DETR-like Decoder2026.01 | 74 | |
| OMG-LLaVAParameters=7B2026.07 | 72.9 | |
| PixDLM2026.04 | 72.8 | |
| Seg-Zero-7BTraining protocol=training-based, Model size=7B2026.05 | 72.6 | |
| PixelLLM-7B2026.02 | 72.4 | |
| Seg-Agent-7BTraining protocol=training-free, Model size=7B2026.05 | 72.2 | |
| RVG (ViT-B)Additions=MLP Decoder2026.01 | 72.1 | |
| GSVA-7BModel Scale=7B2026.04 | 72 | |
| LaSagnA-7BModel Scale=7B2026.04 | 71.9 | |
| PerceptionGPT-7BTraining protocol=training-based, Model size=7B2026.05 | 71.7 | |
| Seg-Zero-3BTraining protocol=training-based, Model size=3B2026.05 | 71.5 | |
| Seg-Agent-3BTraining protocol=training-free, Model size=3B2026.05 | 71.4 | |
| READParameters=7B2026.07 | 71.4 | |
| VisionLLM v2 (Swin-T)Additions=Deform-DETR2026.01 | 71.2 | |
| Qwen2.5-VL-7B + SAM2-LTraining protocol=training-free, Model size=7B, SAM model=SAM2-L2026.05 | 71.2 | |
| VistaLLM (Vicuna-7B)Additions=None2026.01 | 70.9 | |
| LISA-7B2026.02 | 70.6 | |
| LISAParameters=7B2026.07 | 70.6 | |
| PixelLM-7B2026.02 | 70.5 | |
| PixelLM2026.04 | 70.5 | |
| PixelLM-7BTraining protocol=training-based, Model size=7B2026.05 | 70.5 | |
| Qwen2.5-VL-3B + SAM2-LTraining protocol=training-free, Model size=3B, SAM model=SAM2-L2026.05 | 70.1 | |
| PolyFormer2026.07 | 69.1 | |
| LISA-7BTraining protocol=training-based, Model size=7B2026.05 | 68.5 | |
| LISA-7BModel size=7B2023.12 | 67.9 | |
| LISA-7BModel Scale=7B2026.04 | 66.7 | |
| SESAME2023.12 | 66.1 | |
| ReLA*Training protocol=training-based, Note=traditional approach2026.05 | 66 | |
| SEEM2023.12 | 65.7 | |
| ReLA2023.12 | 65 | |
| X-Decoder2023.12 | 64.6 | |
| LAVT*Training protocol=training-based, Note=traditional approach2026.05 | 62.1 | |
| LAVT2023.12 | 61.2 | |
| CRIS*Training protocol=training-based, Note=traditional approach2026.05 | 60.4 | |
| CRIS2026.07 | 60.4 | |
| CRIS2023.12 | 59.9 | |
| VLT2023.12 | 55 | |
| MCN2023.12 | 49.2 |