List your company
ResearchFormerly WizWand

Referring Segmentation on refCOCO+ (testA) using cIoU

0.842cIoUTraceVision-7B
0.319920.455460.5910.72654Dec 4, 2023May 9, 2024Oct 14, 2024Mar 21, 2025Aug 25, 2025Jan 30, 2026Jul 7, 2026
Updated 3mo ago

Evaluation results

MethodLinks
TraceVision-7B2026.020.842
HyperSeg-1.5B2026.020.835
ReChannelParameters=9B2026.070.83
ReChannelParameters=4B2026.070.816
FCLMParameters=7B2026.070.805
CoPRSParameters=7B2026.070.803
UFO (InternVL2.5-8B)Additions=Mask Tokens2026.010.799
Youtu-VL (4B)Additions=None2026.010.796
UniPixel (Qwne2.5-VL-3B)Additions=SAM Decoder2026.010.789
GLaMM (Vicuna-7B)Additions=SAM/Pixel Decoder2026.010.787
GLaMMParameters=7B2026.070.787
Text4SegParameters=8B2026.070.779
Ours (single-object)2026.050.7672
Seg-Zero-7BTraining protocol=training-based, Model size=7B2026.050.762
Seg-Zero-7B2026.050.762
Seg-Agent-7BTraining protocol=training-free, Model size=7B2026.050.76
RVG (ViT-B)Additions=MLP Decoder2026.010.757
PSALM-3B2026.020.755
PSALMParameters=1.3B2026.070.755
Qwen2.5-VL-7Bre-implemented=true, consistent-checkpoint=true2026.050.74
PerceptionGPT-7BTraining protocol=training-based, Model size=7B2026.050.739
PerceptionGPT-7B2026.050.739
READ2024.120.737
VistaLLM (Vicuna-7B)Additions=None2026.010.737
Seg-Zero-3BTraining protocol=training-based, Model size=3B2026.050.737
Read-7B2026.050.737
READParameters=7B2026.070.737
Qwen2.5-VL-7B + SAM2-LTraining protocol=training-free, Model size=7B, SAM model=SAM2-L2026.050.735
PixDLM2026.040.733
Ours (multi-object)consistent-checkpoint=true2026.050.733
Seg-Agent-3BTraining protocol=training-free, Model size=3B2026.050.732
OMG-LLaVAParameters=7B2026.070.731
SEGLLM2026.050.73
PolyFormer2026.070.729
PixelLLM-7B2026.020.721
PixelLMw/o SAM=true2023.120.717
PixelLM-7B2026.020.717
PixelLM2026.040.717
PixelLM-7BTraining protocol=training-based, Model size=7B2026.050.717
PixelLM-7B2026.050.717
Qwen2.5-VL-3B + SAM2-LTraining protocol=training-free, Model size=3B, SAM model=SAM2-L2026.050.715
SAM3Additions=DETR-like Decoder2026.010.711
VisionReasoner-7Bre-implemented=true, consistent-checkpoint=true2026.050.711
ReLAw/o SAM=true2023.120.71
ReLA2024.120.71
ReLA*Training protocol=training-based, Note=traditional approach2026.050.71
Seg-R1-7B2026.050.709
LISA-7B2024.120.708
LISA-7B2026.020.708
LISA-7B2026.050.708
LISAParameters=7B2026.070.708
LaSagnA-7BModel Scale=7B2026.040.706
VisionLLM v2 (Swin-T)Additions=Deform-DETR2026.010.698
LAVTw/o SAM=true2023.120.684
LAVT2024.120.684
LAVT*Training protocol=training-based, Note=traditional approach2026.050.684
CRISw/o SAM=true2023.120.681
CRIS2024.120.681
CRIS*Training protocol=training-based, Note=traditional approach2026.050.681
CRIS2026.070.681
GSVA-7BModel Scale=7B2026.040.677
LISAw/o SAM=false2023.120.674
LISA-7BModel Scale=7B2026.040.674
LISA-7BTraining protocol=training-based, Model size=7B2026.050.674
LISAaugw/o SAM=false2023.120.663
VLTw/o SAM=true2023.120.61
VLT2024.120.61
MCNw/o SAM=true2023.120.55
MCN2024.120.55
Qwen2-VL-7Bre-implemented=true, consistent-checkpoint=true2026.050.523
LLaVA-OV-7B2026.050.471
RegionVLM-4BZero-shot=true2026.020.34