List your company

Referring Expression Segmentation on RefCOCOg Google (val)

67.2gIoUTarot-SAM3
35.37643.63851.960.162Apr 9, 2026

Evaluation results

MethodLinks
Tarot-SAM3Version=Qwen2.5-VL 7B, Trained on RES=false2026.0467.2—
EVOL-SAM3Version=Qwen2.5-VL 7B, Trained on RES=false2026.0465.9—
Tarot-SAM3Version=Qwen2.5-VL 3B, Trained on RES=false2026.0465.1—
EVOL-SAM3Version=Qwen2.5-VL 3B, Trained on RES=false2026.0464—
SAM 3 AgentVersion=Qwen2.5-VL 7B, Trained on RES=false2026.0459.7—
LGD+DINOVersion=ViT-B/32, Trained on RES=false2026.0452.5—
TASVersion=ResNet-50, Trained on RES=false2026.0448.7—
TASVersion=ViT-B/32, Trained on RES=false2026.0448.1—
Pseudo-RISVersion=ETRIS, Trained on RES=false2026.0446.8—
IterRPrimeEVersion=N/A, Trained on RES=false2026.0445.8—
Ref-DiffVersion=VAE, Trained on RES=false2026.0444.3—
GL-CLIPVersion=ResNet-50, Trained on RES=false2026.0444—
Pseudo-RISVersion=CRIS, Trained on RES=false2026.0443.8—
GL-CLIPVersion=ViT-B/32, Trained on RES=false2026.0442.7—
CaRVersion=ViT-B/16, Trained on RES=false2026.0436.6—
PolyFormerLabel-Rate Mask=100%, Label-Rate BBox=100%, Vis. Backbone=Swin-B, Supervision Type=Fully Supervised, Extra Training Data=true2024.07—69.36
ResTRLabel-Rate Mask=100%, Label-Rate BBox=100%, Vis. Backbone=ViT-B, Supervision Type=Fully Supervised2024.07—54.48
SAFARILabel-Rate Mask=100%, Label-Rate BBox=100%, Vis. Backbone=Swin-B, Supervision Type=Fully Supervised, Extra Training Data=true2024.07—70.48
VLTLabel-Rate Mask=100%, Label-Rate BBox=100%, Vis. Backbone=DN53, Supervision Type=Fully Supervised2024.07—49.76