List your company
ResearchFormerly WizWand

Referring Expression Grounding on RefCOCO+

88.77AccuracySingle Task SoTA
43.550855.290467.0378.7696Dec 28, 2023May 24, 2024Oct 19, 2024Mar 17, 2025Aug 12, 2025Jan 7, 2026Jun 5, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Single Task SoTA2023.1288.77
Qwen-VLModel Size=9.6B2023.1283.1
UIO-2XXLModel Size=6.8B2023.1283.1
ShikraModel Size=7.2B2023.1281.6
FerretModel Size=7.2B2023.1280.8
UIO-2XLModel Size=3.2B2023.1279.8
UIO-2LModel Size=1.1B2023.1271.7
SWUDImode=Full fine-tuning, backbone=InternVL2.5-1B2026.0670.18
WUDI Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0670.14
SWUDI-Amode=Full fine-tuning, backbone=InternVL2.5-1B2026.0669.99
OptMergemode=Full fine-tuning, backbone=InternVL2.5-1B2026.0669.82
TIES w/ DAREmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0668.48
TIES Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0668.45
Task Arithmeticmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0668.15
TA w/ DAREmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0668.07
Iso-Cmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0666.56
Mixture Trainingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0665.93
InternVL2.5-Instructmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0665.41
TSV Mergingmode=Full fine-tuning, backbone=InternVL2.5-1B2026.0658.51
Weight Averagemode=Full fine-tuning, backbone=InternVL2.5-1B2026.0645.29