List your company
ResearchFormerly WizWand

Image-to-text on COCO (CHAIRs/BERT Metrics)

0.12CHAIRsFrozen
0.00560.03530.0650.0947May 29, 2026
Updated 5mo ago

Evaluation results

MethodLinks
FrozenBackbone=GPT-5.52026.050.120.070.668
L1 Naive fbBackbone=GPT-5.52026.050.120.0750.672
FrozenBackbone=GPT-5.5, Type=Proprietary2026.050.12—0.668
L1 Naive fbBackbone=Claude Haiku 4.52026.050.1050.0370.66
L2 Text fbBackbone=GPT-5.52026.050.10.0650.669
DeGFBackbone=GPT-5.52026.050.090.0580.673
VolcanoBackbone=GPT-5.52026.050.090.0530.672
WoodpeckerBackbone=GPT-5.52026.050.080.050.668
FrozenBackbone=Claude Haiku 4.52026.050.080.0360.653
VolcanoBackbone=Claude Haiku 4.52026.050.080.0460.641
FrozenBackbone=Claude Haiku 4.5, Type=Proprietary2026.050.08—0.653
DeGFBackbone=Claude Haiku 4.52026.050.070.0310.661
FrozenBackbone=Qwen2.5-Omni-7B, Type=Open2026.050.07—0.588
WoodpeckerBackbone=Claude Haiku 4.52026.050.060.0310.665
FrozenBackbone=LLaVA-1.5-7B, Type=Open2026.050.06—0.74
L2 Text fbBackbone=Claude Haiku 4.52026.050.0510.0270.663
TIGERBackbone=GPT-5.52026.050.050.030.686
TIGERBackbone=Claude Haiku 4.52026.050.050.0250.669
TIGERBackbone=Qwen2.5-Omni-7B, Type=Open2026.050.05—0.643
TIGERBackbone=GPT-5.5, Type=Proprietary2026.050.05—0.686
TIGERBackbone=Claude Haiku 4.5, Type=Proprietary2026.050.05—0.669
FrozenBackbone=Gemini 3.5 Flash2026.050.030.0150.497
TIGERBackbone=LLaVA-1.5-7B, Type=Open2026.050.03—0.772
FrozenBackbone=Gemini 3.5 Flash, Type=Proprietary2026.050.03—0.497
VolcanoBackbone=Gemini 3.5 Flash2026.050.0250.0190.55
DeGFBackbone=Gemini 3.5 Flash2026.050.020.0150.535
L1 Naive fbBackbone=Gemini 3.5 Flash2026.050.0170.010.542
L2 Text fbBackbone=Gemini 3.5 Flash2026.050.0150.0060.459
WoodpeckerBackbone=Gemini 3.5 Flash2026.050.0110.0130.581
TIGERBackbone=Gemini 3.5 Flash2026.050.010.0050.63
TIGERBackbone=Gemini 3.5 Flash, Type=Proprietary2026.050.01—0.63