List your company
ResearchFormerly WizWand

Referring Segmentation on refCOCO (testA) using cIoU

87.5cIoUSegCompass-7B
37.47650.46363.4576.437Dec 23, 2024Mar 26, 2025Jun 28, 2025Sep 29, 2025Jan 1, 2026Apr 4, 2026Jul 7, 2026
Updated 3mo ago

Evaluation results

MethodLinks
SegCompass-7BBackbone=Qwen2.5-VL, Method Category=Interpretable Alignment Method2026.0587.5
SegCompass-13BBackbone=LLaVA-1.5, Method Category=Interpretable Alignment Method2026.0587.3
X-SAM-3.8BBackbone=Phi-3, Method Category=Latent Query Alignment Method2026.0587.1
TraceVision-7B2026.0286.8
ReChannelParameters=9B2026.0786.5
HiMTok-8BBackbone=InternVL-2.5, Method Category=Latent Query Alignment Method2026.0586.3
HyperSeg-1.5B2026.0285.7
ReChannelParameters=4B2026.0785.6
CoPRSParameters=7B2026.0785.3
PSALM-3B2026.0284.7
PSALMParameters=1.3B2026.0784.7
FCLMParameters=7B2026.0783.9
RAS-13BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0583.5
LIRA-8BBackbone=InternVL-2, Method Category=Latent Query Alignment Method2026.0583.4
GLaMM (Vicuna-7B)Additions=SAM/Pixel Decoder2026.0183.2
GLaMM-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0583.2
GLaMMParameters=7B2026.0783.2
UniPixel-7BBackbone=Qwen2.5-VL, Method Category=Latent Query Alignment Method2026.0583
SegCompass-7BBackbone=LLaVA-1.5, Method Category=Interpretable Alignment Method2026.0582.9
SAM4MLLM-7BBackbone=LLaVA-1.6, Method Category=Textual Localization Readout Method2026.0582.8
Text4Seg-13BBackbone=LLaVA-1.5, Method Category=Textual Localization Readout Method2026.0582.7
UniPixel (Qwne2.5-VL-3B)Additions=SAM Decoder2026.0182.6
UFO-8BBackbone=InternVL-2.5, Method Category=Latent Query Alignment Method2026.0582.6
Youtu-VL (4B)Additions=None2026.0182
Text4Seg-7BBackbone=LLaVA-1.5, Method Category=Textual Localization Readout Method2026.0581.9
UniRES-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0581.8
Text4SegParameters=8B2026.0781.7
UFO (InternVL2.5-8B)Additions=Mask Tokens2026.0181.6
SEGLLM2026.0581.5
SegLLM-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0581.5
RVG (ViT-B)Additions=MLP Decoder2026.0181.2
Ours (single-object)2026.0580.79
Seg-Zero-7BTraining protocol=training-based, Model size=7B2026.0580.3
Seg-Zero-7B2026.0580.3
OMG-LLaVA-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0580.3
Seg-Zero-7BBackbone=Qwen2.5-VL, Method Category=Textual Localization Readout Method2026.0580.3
OMG-LLaVAParameters=7B2026.0780.3
READ2024.1280.2
PixDLM2026.0480.2
Read-7B2026.0580.2
READParameters=7B2026.0780.2
Seg-Agent-7BTraining protocol=training-free, Model size=7B2026.0579.9
GroundHog-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0579.9
VisionLLM v2 (Swin-T)Additions=Deform-DETR2026.0179.3
Seg-Zero-3BTraining protocol=training-based, Model size=3B2026.0579.3
SAM-R1-7BBackbone=Qwen2.5-VL, Method Category=Textual Localization Readout Method2026.0579.2
LISA-7B2024.1279.1
LISA-7B2026.0279.1
LISA-7B2026.0579.1
LISA-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0579.1
PerceptionGPT-13BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0579.1
LISAParameters=7B2026.0779.1
Seg-Agent-3BTraining protocol=training-free, Model size=3B2026.0579
VisionReasoner-7BBackbone=Qwen2.5-VL, Method Category=Textual Localization Readout Method2026.0578.9
LISA-13BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0578.8
LaSagnA-7BModel Scale=7B2026.0478.7
Seg-R1-7B2026.0578.7
LaSagnA-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0578.7
Seg-R1-7BBackbone=Qwen2.5-VL, Method Category=Textual Localization Readout Method2026.0578.7
PerceptionGPT-7BTraining protocol=training-based, Model size=7B2026.0578.6
PerceptionGPT-7B2026.0578.6
PerceptionGPT-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0578.6
PixelLLM-7B2026.0278.5
Ours (multi-object)consistent-checkpoint=true2026.0578.04
Qwen2.5-VL-7Bre-implemented=true, consistent-checkpoint=true2026.0577.9
Qwen2.5-VL-7B + SAM2-LTraining protocol=training-free, Model size=7B, SAM model=SAM2-L2026.0577.8
SAM3Additions=DETR-like Decoder2026.0177.6
GSVA-7BModel Scale=7B2026.0477.4
VisionReasoner-7Bre-implemented=true, consistent-checkpoint=true2026.0577.4
PolyFormer2026.0776.6
ReLA2024.1276.5
PixelLM-7B2026.0276.5
LISA-7BModel Scale=7B2026.0476.5
PixelLM2026.0476.5
ReLA*Training protocol=training-based, Note=traditional approach2026.0576.5
LISA-7BTraining protocol=training-based, Model size=7B2026.0576.5
PixelLM-7BTraining protocol=training-based, Model size=7B2026.0576.5
PixelLM-7B2026.0576.5
ReLAMethod Category=Method without LLMs2026.0576.5
PixelLM-7BBackbone=LLaVA-1.5, Method Category=Latent Query Alignment Method2026.0576.5
VistaLLM (Vicuna-7B)Additions=None2026.0176
Qwen2.5-VL-3B + SAM2-LTraining protocol=training-free, Model size=3B, SAM model=SAM2-L2026.0575.9
LAVT2024.1275.8
LAVT*Training protocol=training-based, Note=traditional approach2026.0575.8
LAVTMethod Category=Method without LLMs2026.0575.8
CRIS2024.1273.2
CRIS*Training protocol=training-based, Note=traditional approach2026.0573.2
CRISMethod Category=Method without LLMs2026.0573.2
CRIS2026.0773.2
VLT2024.1270.5
VLTMethod Category=Method without LLMs2026.0570.5
MCN2024.1264.2
Qwen2-VL-7Bre-implemented=true, consistent-checkpoint=true2026.0558.9
LLaVA-OV-7B2026.0558.1
RegionVLM-4BZero-shot=true2026.0239.4