List your company
ResearchFormerly WizWand

Scene Text-Centric Visual Question Answering on STVQA

0.759AccuracyVisual-SR1
-0.00540.193050.39150.58995Jul 23, 2024Oct 2, 2024Dec 13, 2024Feb 22, 2025May 5, 2025Jul 15, 2025Sep 25, 2025
Updated 5mo ago

Evaluation results

MethodLinks
Visual-SR1Backbone=Qwen2.5-VL-7B2025.090.759
ViCropBackbone=LLaVA-1.5 (Vicuna-7B)2025.090.724
DeFactoBackbone=Qwen2.5-VL-7B2025.090.712
Qwen2.5-VLBackbone=Qwen2.5-VL-7B2025.090.679
GRITBackbone=Qwen2.5-VL-3B2025.090.647
InternVLModel Generation Capability=Text only2024.070.622
InternLM-XComposer2Model Generation Capability=Text only2024.070.596
MonkeyModel Generation Capability=Text only2024.070.547
TextHarmony-ChatModel Generation Capability=Text and image, Slide-LoRA=true2024.070.513
mPLUG-Owl2Model Generation Capability=Text only2024.070.498
TextHarmonyModel Generation Capability=Text and image, Slide-LoRA=true2024.070.497
TextHarmony*Model Generation Capability=Text and image, Slide-LoRA=false2024.070.472
DeepEyesBackbone=Qwen2.5-VL-7B2025.090.459
DocPediaModel Generation Capability=Text only2024.070.455
LLaVARModel Generation Capability=Text only2024.070.392
LLaVA1.5-7BModel Generation Capability=Text only2024.070.381
UniDocModel Generation Capability=Text only2024.070.352
MM-InterleavedModel Generation Capability=Text and image2024.070.264
SEED-LLaMA-14BModel Generation Capability=Text and image2024.070.201
MiniGPT5Model Generation Capability=Text and image2024.070.024