List your company
ResearchFormerly WizWand

Document Visual Question Answering on DocVQA v1.0 (test)

96.5ANLSQwen2-VL-72B
8.20431.12754.0576.973Apr 8, 2024May 18, 2024Jun 27, 2024Aug 7, 2024Sep 16, 2024Oct 26, 2024Dec 6, 2024
Updated 5mo ago

Evaluation results

MethodLinks
Qwen2-VL-72B2024.1296.5—
InternVL2.5-38B2024.1295.3—
Claude-3.5-Sonnet2024.1295.2—
InternVL2.5-78B2024.1295.1—
Qwen2-VL-7B2024.1294.5—
InternVL2-Llama3-76B2024.1294.1—
InternVL2.5-26B2024.1294—
InternVL2-40B2024.1293.9—
Molmo-72B2024.1293.5—
Gemini-1.5-Pro2024.1293.1—
InternVL2.5-8B2024.1293—
InternVL2-26B2024.1292.9—
GPT-4o-202405132024.1292.8—
NVLM-D-72B2024.1292.6—
InternVL-1.5-Plusparameters=26B, pretraining corpora=26, fine-tuning corpora=492024.1192.3—
Molmo-7B-D2024.1292.2—
InternVL2.5-4B2024.1291.6—
InternVL2-8B2024.1291.6—
LLaVA-OneVision-72B2024.1291.3—
InternVL-Chat-V1.52024.1290.9—
MiniCPM-V2.62024.1290.8—
Qwen2-VL-2B2024.1290.1—
Claude-3-Opus2024.1289.3—
InternVL2-4B2024.1289.2—
InternVL2.5-2B2024.1288.7—
GPT-4V2024.1288.4—
InternVL2-2B2024.1286.9—
Aquila-VL-2B2024.1285—
InternVL2.5-1B2024.1284.8—
LayoutLMv3Model Category=Fine-tuned PTM, Evaluation Protocol=Fine-tuned, Implementation=Original Paper2024.0483.37—
InternVL2-1B2024.1281.7—
Cambrian-34B2024.1275.5—
LayoutLLM-7BModel Category=Ours, Evaluation Protocol=Zero-shot, Backbone Initialization=Vicuna-1.5-7B2024.0474.27—
LayoutLLM-7BModel Category=Ours, Evaluation Protocol=Zero-shot, Backbone Initialization=Llama2-7B-chat2024.0474.25—
HVFAbackbone=mPLUG-Owl, pretraining dataset=1B, fine-tuning dataset=650K, OCR-free=true2024.1172.7—
LLaVA-OneVision-0.5B2024.1270—
Phi-3.5-Vision-4B2024.1269.3—
Vicuna-1.5-7BModel Category=LLM, Input Representation=Plain Text, Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0466.99—
Qwen-VL-7BModel Category=MLLM, Evaluation Protocol=Fine-tuned, Implementation=Original Paper2024.0465.1—
Llama2-7B-chatModel Category=LLM, Input Representation=Plain Text, Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0464.99—
mPLUG-DocOWL-7BModel Category=MLLM, Evaluation Protocol=Fine-tuned, Implementation=Original Paper2024.0462.2—
Vicuna-7BModel Category=LLM, Input Representation=Plain Text, Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0461.39—
Llama2-7BModel Category=LLM, Input Representation=Plain Text, Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0461.34—
Vicuna-1.5-7BModel Category=LLM, Input Representation=Layout Text (Text + Box), Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0456.81—
Llama2-7B-chatModel Category=LLM, Input Representation=Layout Text (Text + Box), Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0456.55—
Llama2-7BModel Category=LLM, Input Representation=Layout Text (Text + Box), Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0437.32—
Vicuna-7BModel Category=LLM, Input Representation=Layout Text (Text + Box), Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0437.21—
LLaVA-1.5-7BModel Category=MLLM, Evaluation Protocol=Zero-shot, Implementation=Re-implemented2024.0413.34—
LLaVAR-7BModel Category=MLLM, Evaluation Protocol=Zero-shot, Implementation=Original Paper2024.0411.6—
AdaReasoner-7BTool Setting=w/ IID Tools2026.06—89.4
AdaReasoner-7BTool Setting=w/ OOD Tools2026.06—92.3
Claude-Sonnet-4-6Tool Setting=w/ IID Tools2026.06—98.18
Claude-Sonnet-4-6Tool Setting=w/ OOD Tools2026.06—98.08
DeepEyesV2-7BTool Setting=w/ IID Tools2026.06—89.9
DeepEyesV2-7BTool Setting=w/ OOD Tools2026.06—93.05
Gemini-3.1-Pro-PreviewTool Setting=w/ IID Tools2026.06—96.95
Gemini-3.1-Pro-PreviewTool Setting=w/ OOD Tools2026.06—96.13
GPT-5.4Tool Setting=w/ IID Tools2026.06—96.85
GPT-5.4Tool Setting=w/ OOD Tools2026.06—96.55
InternVL3.5-14BTool Setting=w/ IID Tools2026.06—87.6
InternVL3.5-14BTool Setting=w/ OOD Tools2026.06—88.2
InternVL3.5-38BTool Setting=w/ IID Tools2026.06—91.8
InternVL3.5-38BTool Setting=w/ OOD Tools2026.06—92.1
LLaVA-1.5 13BVIG training=false2026.02—24.08
LLaVA-1.5 13B + VIG trainingVIG training=true2026.02—25.27
LLaVA-1.5 7BVIG training=false2026.02—22.31
LLaVA-1.5 7B + VIG trainingVIG training=true2026.02—22.51
MetaForgeTool Setting=w/ IID Tools2026.06—97.9
MetaForgeTool Setting=w/ OOD Tools2026.06—97.22
PixelReasoner-RL-v1-7BTool Setting=w/ IID Tools2026.06—92.3
PixelReasoner-RL-v1-7BTool Setting=w/ OOD Tools2026.06—91.85
PyVision-Image-7B-RLTool Setting=w/ IID Tools2026.06—69.95
PyVision-Image-7B-RLTool Setting=w/ OOD Tools2026.06—77
Qwen2.5-VL-7BTool Setting=w/ IID Tools2026.06—93.43
Qwen2.5-VL-7BTool Setting=w/ OOD Tools2026.06—93.63
Qwen3-VL-32BTool Setting=w/ IID Tools2026.06—94.7
Qwen3-VL-32BTool Setting=w/ OOD Tools2026.06—95.15
Qwen3-VL-8BTool Setting=w/ IID Tools2026.06—89.8
Qwen3-VL-8BTool Setting=w/ OOD Tools2026.06—90.2
R1-Onevision-7BTool Setting=w/ IID Tools2026.06—83.2
R1-Onevision-7BTool Setting=w/ OOD Tools2026.06—83.9
ShareGPT4V 7BVIG training=false2026.02—26.15
ShareGPT4V 7B + VIG trainingVIG training=true2026.02—28.23
TikArt-8BTool Setting=w/ IID Tools2026.06—96.7
TikArt-8BTool Setting=w/ OOD Tools2026.06—94.7
VTool-R1-7BTool Setting=w/ IID Tools2026.06—93.1
VTool-R1-7BTool Setting=w/ OOD Tools2026.06—93.25
ZwZ-8BTool Setting=w/ IID Tools2026.06—96.8
ZwZ-8BTool Setting=w/ OOD Tools2026.06—95.6