List your company
ResearchFormerly WizWand

Change Referring Segmentation on MTRefSeg-21K NS (val)

61.55mIoUGSVA-7B
0.221216.143132.06547.9869May 31, 2026
Updated 5mo ago

Evaluation results

MethodLinks
GSVA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Fine-tuned on MTRefSeg-21K2026.0561.55——66.9358.4250.543.1726.53
FIANetVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0561.32——70.1362.1652.2338.1513.48
RMSINVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0560.1——68.6760.7551.5337.0412.93
GLAMM-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Fine-tuned on MTRefSeg-21K2026.0560.08——64.1657.4350.340.425.94
GSVA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Protocol=Fine-tuned on MTRefSeg-21K2026.0559.91——62.3856.2450.343.1728.71
MTRefSeg-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Protocol=Fine-tuned on MTRefSeg-21K2026.0559.87——66.7261.1553.9343.0123.26
LISA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Protocol=Fine-tuned on MTRefSeg-21K2026.0558.1——6055.4549.3140.7928.12
LGCEVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0556.78——64.4656.0445.6132.3811.53
LAVTVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0556.46——63.4654.8945.3131.7812.03
LISA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Fine-tuned on MTRefSeg-21K2026.0555.59——58.0252.6744.5536.8324.36
RefSegformerVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0555.49——62.8654.0443.8129.629.22
SegEarth-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Protocol=Fine-tuned on MTRefSeg-21K2026.0554.95——60.3754.1147.4437.0719.04
RSRefSegVision Encoder=Swin-B, Text Encoder=BERT, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0550.02——54.5946.7237.8426.8214.34
UniChangeVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Fine-tuned on MTRefSeg-21K2026.0541.51——41.3935.8427.7219.610.89
CRIS-101Vision Encoder=ResNet-101, Text Encoder=ResNet-101, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0535.81——36.3926.7714.695.410.65
CRIS-50Vision Encoder=ResNet-50, Text Encoder=ResNet-50, Protocol=Specialist Method, Trained on MTRefSeg-21K2026.0535.24——35.0425.8114.996.721.3
GSVA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Protocol=Without Fine-tuning2026.0522.74——18.4214.659.75.940.59
GeoPixel-7BVision Encoder=CLIP-L, Text Encoder=InternLM-XComposer2.5, Protocol=Fine-tuned on MTRefSeg-21K2026.0517.47——10.186.373.611.90.55
GSVA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Without Fine-tuning2026.0516.6——11.297.135.944.361.78
UniGeoSegVision Encoder=Swin-B, Text Encoder=Phi-1.5, Protocol=Without Fine-tuning2026.0514.67——12.6310.487.675.362.71
GeoPixel-7BVision Encoder=CLIP-L, Text Encoder=InternLM-XComposer2.5, Protocol=Without Fine-tuning2026.0512.64——6.423.861.850.850.2
GLAMM-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Without Fine-tuning2026.0510.07——7.925.152.971.190.99
SegEarth-R1Vision Encoder=Swin-B, Text Encoder=Phi-1.5, Protocol=Without Fine-tuning2026.055.77——2.812.011.10.80.3
LISA-7BVision Encoder=CLIP-L, Text Encoder=Vicuna-7B, Protocol=Without Fine-tuning2026.054.88——1.780.790.40.20
LISA-13BVision Encoder=CLIP-L, Text Encoder=Vicuna-13B, Protocol=Without Fine-tuning2026.052.58——0.990.590.40.40