AI Paper Atlas

Research map / Benchmarks / AI2D

AI2D leaderboard

86 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on AI2D. Best accuracy: 96.3 (Molmo-72B, CVPR 2025).

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.

AI2D · accuracy

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

#accuracy ↑MethodPaperTopic
196.3Molmo-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
290SoT
Qwen2.5-VL-32B
State of Thought Enables Endogenous Reasoning
NeurIPS 2026 · Table 5
Reasoning and chain of thought
387.9MMFineReason-8B (SFT+RL)
Qwen3-VL-8B-Instruct
MMFineReason: Closing the Multimodal Reasoning Gap via Open Data-Centric Methods
NeurIPS 2026 · Table 3
Multimodal LLMs
486.5MiniCPM-V 4.5MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
CVPR 2026 · Table 1
Multimodal LLMs
586.2R-4B-RL (A-T)
R-4B
R-4B: Incentivizing General-Purpose Auto-Thinking in MLLMs via Bi-Mode Annealing and Reinforce Learning
CVPR 2026 · Table 1
Reasoning and chain of thought
686.2DocMark-8B
InternLM-8B
Adaptive Markup Language Generation for Contextually-Grounded Visual Document Understanding
CVPR 2025 · Table: Performance comparisons
Multimodal LLMs
786.2R1-ShareVL-32B
Qwen2.5-VL-32B
R1-ShareVL: Incentivizing Reasoning Capabilities of Multimodal Large Language Models via Share-GRPO
NeurIPS 2025 · Main Results table
RL from feedback
886Manzano-30BMANZANO: A Simple and Scalable Unified Multimodal Model with a Hybrid Vision Tokenizer
ICLR 2026 · Table 2
Multimodal LLMs
985.6TVI-CoT
Qwen3-VL-8B
TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding
ICML 2026 · Table 1
Multimodal LLMs
1085.4CoLT
Qwen3-VL-8B
CoLT: Teaching Multi-Modal Models to Think with Chain of Latent Thoughts
ECCV 2026 · Table 1
Multimodal LLMs
40 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

AI2D · score

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

#score ↑MethodPaperTopic
194.1NVILA 15B
Qwen2
NVILA: Efficient Frontier Visual Language Models
CVPR 2025 · Image benchmarks table
Multimodal LLMs
290.77Twin-T-7B
Qwen2.5-VL-7B
Twin-T & TwintVQA: A Reliable Structure-Detail Separating VLM and a Comprehensive Benchmark for Chart and Table Tasks
CVPR 2026 · Table 1
Multimodal LLMs
389VLsI-2B
Qwen2-VL-2B
VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models
CVPR 2025 · Table 2
Multimodal LLMs
486.7Qwen2.5-VL-7B w. RIL
Qwen2.5-VL-7B
Unified Reinforcement and Imitation Learning for Vision-Language Models
NeurIPS 2025 · Table 1
Multimodal LLMs
585.3BlueLM-V-3B
BlueLM-3B
BlueLM-V-3B: Algorithm and System Co-Design for Multimodal Large Language Models on Mobile Devices
CVPR 2025 · Table 3
Quantization
684.9CoS
InternVL2.5-MPO 8B
Unveiling Chain of Step Reasoning for Vision-Language Models with Fine-grained Rewards
NeurIPS 2025 · Table 1
Multimodal LLMs
783.2Oryx-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
881.2PACT
LLaVA-OneVision-7B
PACT: Pruning and Clustering-Based Token Reduction for Faster Visual Language Models
CVPR 2025 · Table 2
Attention and transformers
981.2InternLM-XC2 + MMDU-45k
InternLM-XC2
MMDU: A Multi-Turn Multi-Image Dialog Understanding Benchmark and Instruction-Tuning Dataset for LVLMs
NeurIPS 2024 · Fine-tuning Results, Table
Multimodal LLMs
1080.9IXC2-4KHD
InternLM2-7B
InternLM-XComposer2-4KHD: A Pioneering Large Vision-Language Model Handling Resolutions from 336 Pixels to 4K HD
NeurIPS 2024 · Table: Comparison with open-so
Multimodal LLMs
26 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.