AI Paper Atlas

Research map / Benchmarks / DocVQA

DocVQA leaderboard

93 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on DocVQA. Best accuracy: 96.4 (AdaThinking-E (ATK), ECCV 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

DocVQA · accuracy

40 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
196.4AdaThinking-E (ATK)
Qwen3-VL-8B
AdaThinking-E: One-Token Entropy Regulation for Adaptive Thinking
ECCV 2026 · Table 1
Reasoning and chain of thought
294.7MiniCPM-V 4.5MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
CVPR 2026 · Table 1
Multimodal LLMs
394.3Manzano-30BMANZANO: A Simple and Scalable Unified Multimodal Model with a Hybrid Vision Tokenizer
ICLR 2026 · Table 2
Multimodal LLMs
494DeFacto
Qwen2.5-VL-7B
DeFacto: Counterfactual Thinking with Images for Enforcing Evidence-Grounded and Faithful Reasoning
ICML 2026 · Table 2
Multimodal LLMs
593.5Molmo-72B
Qwen2 72B
Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models
CVPR 2025 · Table 1
Multimodal LLMs
693.5OURO-7B
Qwen2-VL-7B-Instruct
OURO: A Self-Bootstrapped Framework for Enhancing Multimodal Scene Understanding
ICCV 2025 · Table 2
Multimodal LLMs
793.38Text-Printed Image (TPI)
Qwen2.5 VL 7B
Text-Printed Image: Bridging the Image-Text Modality Gap for Text-centric Training of Large Vision-Language Models
CVPR 2026 · Table 2
Multimodal LLMs
892.8LLaVA-UHD-v3
Qwen2-7B
NaVLM-PVC: Progressive Visual Compression for Efficient Native-Resolution Encoding in MLLMs
ECCV 2026 · Table 3
Multimodal LLMs
992.5PVCInternVL2-8B
InternVL2-8B
PVC: Progressive Visual Token Compression for Unified Image and Video Processing in Large Vision-Language Models
CVPR 2025 · Table 2
Video-language understanding
1091.7ROSE-7B
Qwen2.5-7B
ROSE: Rotate Your Large Language Model to See
CVPR 2026 · Table 1
Multimodal LLMs
30 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

DocVQA · ANLS

28 papers, each with its best reported number for its own method; higher is better.

#ANLS ↑MethodPaperTopic
194.2TokenVL
8B
A Token-level Text Image Foundation Model for Document Understanding
ICCV 2025 · Table 5
Multimodal LLMs
292Docopilot-8B
InternVL2-8B
Docopilot: Improving Multimodal Models for Document-Level Understanding
CVPR 2025 · Table: Results on single-page
Multimodal LLMs
391.7DOGRDOGR: Towards Versatile Visual Document Grounding and Referring
ICCV 2025 · Table 3
Multimodal LLMs
491.4DocVAL
Gemma3-12B
DocVAL: Validated Chain-of-Thought Distillation for Grounded Document VQA
ICML 2026 · Table 1
Multimodal LLMs
591.2DocVLM (Qwen2-VL, 320 tokens)
Qwen2-VL 7B
DocVLM: Make Your VLM an Efficient Reader
CVPR 2025 · Table 1
Multimodal LLMs
691.16+ Entropy-gradient grounding (Ours)
Qwen2.5-VL 7B
Entropy-Gradient Grounding: Training-Free Evidence Retrieval in Vision-Language Models
ECCV 2026 · Table 1
Multimodal LLMs
787.84DocFormerv2-large†DocFormerv2: Local Features for Document Understanding
AAAI 2024 · Table: Comparison on Document
Human and embodied modeling
887.4Mini-Monkey
InternVL2-2B
Mini-Monkey: Alleviating the Semantic Sawtooth Effect for Lightweight MLLMs via Complementary Image Pyramid
ICLR 2025 · Table: Comparison to state-of-
Multimodal LLMs
986.8PaLI-X (with OCR pipeline input)On Scaling Up a Multilingual Vision and Language Model
CVPR 2024 · Table 2
Visual question answering
1086.7GRAM (large)
DocFormerv2 large
GRAM: Global Reasoning for Multi-Page VQA
CVPR 2024 · Table 2
Visual question answering
18 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

DocVQA · score

26 papers, each with its best reported number for its own method; higher is better.

#score ↑MethodPaperTopic
196.8VLM-R³
Qwen2.5-VL 7B
VLM-R³: Region Recognition, Reasoning, and Refinement for Enhanced Multimodal Chain-of-Thought
NeurIPS 2025 · Main Result table
Multimodal LLMs
294.1Eagle2.5-8B
Qwen2.5-7B
Eagle 2.5: Boosting Long-Context Post-Training for Frontier Vision-Language Models
NeurIPS 2025 · Image benchmarks table
Multimodal LLMs
392.7Oryx-1.5-32B
Qwen-2.5-32B
Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution
ICLR 2025 · Image Understanding table
Video-language understanding
492Marten
InternViT-300M + InternLM2 7B
Marten: Visual Question Answering with Mask Generation for Multi-modal Document Understanding
CVPR 2025 · Table 3
Visual question answering
591.5Insight-V
Our Base Model
Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models
CVPR 2025 · Table 2
Multimodal LLMs
691.4UniARUnified Multimodal Autoregressive Modeling with Shared Context—Visual Tokenizer is Key to Unification
ICML 2026 · Table 4
Human and embodied modeling
790.6NaViL-9B
Qwen3-8B
NaViL: Rethinking Scaling Properties of Native Multimodal Large Language Models under Data Constraints
NeurIPS 2025 · Appendix table: Comparison bet
Multimodal LLMs
890.3Llama-3.2-11B + Lavender LoRA-FT
LM32-11B
Diffusion Instruction Tuning
ICML 2025 · Table
Multimodal LLMs
990IXC2-4KHD
InternLM2-7B
InternLM-XComposer2-4KHD: A Pioneering Large Vision-Language Model Handling Resolutions from 336 Pixels to 4K HD
NeurIPS 2024 · Table: Comparison with closed-
Multimodal LLMs
1087.8BlueLM-V-3B
BlueLM-3B
BlueLM-V-3B: Algorithm and System Co-Design for Multimodal Large Language Models on Mobile Devices
CVPR 2025 · Table 4
Quantization
16 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.