Reasoning Segmentation on ReasonSeg
79.3gIoUVision Banana
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Vision BananaGemini-assisted=true2026.07 | 79.3 | — | |
| SenseNova-VisionGemini-assisted=false2026.07 | 63.2 | — | |
| InstructSeg2026.03 | 61.9 | 65.2 | |
| VIRST2026.03 | 60.8 | 65 | |
| HyperSeg-3BBackbone=Swin-B, Pre-trained SAM=false2024.11 | 59.2 | 56.7 | |
| HyperSeg2026.03 | 59.2 | 56.7 | |
| LISABackbone=LLaVA-7B2024.07 | 52.9 | 54 | |
| LISA-7BBackbone=ViT-H, Pre-trained SAM=true2024.11 | 52.9 | 54 | |
| LISA-7BModel size=7B2026.03 | 52.9 | 54 | |
| VISABackbone=Chat-UniVi-7B2024.07 | 52.7 | 57.8 | |
| VISA-7BBackbone=ViT-H, Pre-trained SAM=true2024.11 | 52.7 | 57.8 | |
| VISA-7BModel size=7B2026.03 | 52.7 | 57.8 | |
| VRS-HQ-7BModel size=7B2026.03 | 51.7 | 52.9 | |
| READ-7B + ROSERAG=true2026.04 | 50.3 | 54.9 | |
| READ-7BRAG=false2026.04 | 49.7 | 54.2 | |
| LaSagnA-7BBackbone=ViT-H, Pre-trained SAM=true2024.11 | 48.8 | 47.2 | |
| FAST2024.08 | 48.7 | 47.6 | |
| LISABackbone=LLaVA-7B, Note=trained via LISA's official GitHub repository2024.07 | 48.1 | 53.7 | |
| SAM4MLLM-7BBackbone=EfficientViT-SAM-XL1, Pre-trained SAM=false2024.11 | 46.7 | 48.1 | |
| LISA-7B2024.08 | 46 | 44.4 | |
| LISA-7BRAG=false2026.04 | 42.5 | 44.2 | |
| LISA-7B + ROSERAG=true2026.04 | 42.2 | 44.1 | |
| LLaVA w Seg AdapterArchitecture=LLaVA with Segmentation Adapter2024.08 | 41 | 43 | |
| SESAME-7B + ROSERAG=true2026.04 | 33.7 | 31.6 | |
| SEEMRAG=false2026.04 | 33.2 | 25.4 | |
| SESAME-7BRAG=false2026.04 | 30.9 | 28.7 | |
| SEEMBackbone=DaViT-L2024.07 | 25.5 | 21.2 | |
| X-DecoderBackbone=DaViT-L2024.07 | 22.6 | 17.9 | |
| SEEM2024.08 | 21.2 | 25.5 | |
| Grounded-SAMRAG=false2026.04 | 20.6 | 15.6 | |
| GRESRAG=false2026.04 | 20.3 | 17.7 | |
| GRES2024.08 | 19.9 | 22.4 | |
| OVSeg2024.08 | 18.6 | 28.5 | |
| CRISRAG=false2026.04 | 18.2 | 19.7 | |
| X-Decoder2024.08 | 17.9 | 22.6 |