List your company
ResearchFormerly WizWand

Document Information Extraction on DeepForm (test)

80.061F1InternVL2-8B-CoB
2.6215622.7260342.830562.93497Nov 8, 2024Dec 10, 2024Jan 12, 2025Feb 14, 2025Mar 19, 2025Apr 21, 2025May 24, 2025
Updated 5mo ago

Evaluation results

MethodLinks
InternVL2-8B-CoBDoc-CoB=true, Backbone=InternVL2-8B2025.0580.061
DocOwl 1.5-CoBDoc-CoB=true, Backbone=DocOwl 1.52025.0579.332
Qwen2.5-VL-CoBDoc-CoB=true, Backbone=Qwen2.5-VL2025.0579.2
InternVL2-2B-CoBDoc-CoB=true, Backbone=InternVL2-2B2025.0578.57
DocOwl 1.5-RReasoning tasks=true, Backbone=DocOwl 1.52025.0577.86
DocOwl 1.5*SFT baseline=true, Backbone=DocOwl 1.52025.0574.98
DonutModel Params=143M, Trainable Params=143M, Pre-training Data=13M, Fine-tuning Data=Task-specific2024.1161.6
mPLUG-Owl-7B + OursModel Params=7.2B, Trainable Params=96M, Pre-training Data=1.1B, Fine-tuning Data=650K2024.1153
mPLUG-Owl-7B + UReaderModel Params=7.2B, Trainable Params=86M, Pre-training Data=1.1B, Fine-tuning Data=650K2024.1149.5
InternVL2-2B-RReasoning tasks=true, Backbone=InternVL2-2B2025.0548.49
InternVL2-8B-RReasoning tasks=true, Backbone=InternVL2-8B2025.0548.49
Visual CoTMethod Class=Visual Prompting Methods2025.0544.76
GPT-4oSize=Much Larger2025.0544.75
InternVL2-40BSize=Much Larger2025.0543.71
TextCoTMethod Class=Visual Prompting Methods2025.0542.45
Qwen2.5-VL-RReasoning tasks=true, Backbone=Qwen2.5-VL2025.0542.11
MonkeyModel Params=9.8B, Trainable Params=207M, Pre-training Data=1.4B+76.8M, Fine-tuning Data=1.44M2024.1140.6
Qwen2.5-VL*SFT baseline=true, Backbone=Qwen2.5-VL2025.0538.2
InternVL2-8B*SFT baseline=true, Backbone=InternVL2-8B2025.0536.33
InternVL2-2B*SFT baseline=true, Backbone=InternVL2-2B2025.0534.15
BLIP-2-OPT-2.7B + OursModel Params=3.8B, Trainable Params=14M, Pre-training Data=129M, Fine-tuning Data=650K2024.1114.6
BLIP-2-OPT-2.7B + UReaderModel Params=3.8B, Trainable Params=8M, Pre-training Data=129M, Fine-tuning Data=650K2024.115.6