AI Paper Atlas

Research map / Benchmarks / MMMU

MMMU leaderboard

202 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MMMU. Best accuracy: 85.79 (DREAM-R, ICML 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.

MMMU · accuracy

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

#accuracy ↑MethodPaperTopic
185.79DREAM-R
Target Qwen3-VL-32B, draft Qwen3-VL-2B
DREAM-R: Multimodal Speculative Reasoning with RL-Based Refined Drafting, Precise Verification, and Fully Parallel Execution
ICML 2026 · Table 2
Reasoning and chain of thought
278.33RLVR w/ Multi-Hop (HopChain)
Qwen3.5-35B-A3B
HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning
NeurIPS 2026 · Table 1
Multimodal LLMs
375.01GAM-Agent
InternVL3-78B
GAM-Agent: Game-Theoretic and Uncertainty-Aware Collaboration for Complex Visual Reasoning
NeurIPS 2025 · Table: Image and Video Underst
LLM agents
474.6CW-DPO (Ours)
Qwen2.5-VL-72B
Dynamics-Aware Preference Optimization for Vision-Language Models
CVPR 2026 · Table 2
Multimodal LLMs
571.4ReMoT-4b-CoT
Qwen3-VL-4B
ReMoT: Reinforcement Learning with Motion Contrast Triplets
CVPR 2026 · Table 4
Video-language understanding
670.6OneThinker-8B
Qwen3-VL-8B-Instruct
OneThinker: All-in-one Reasoning Model for Image and Video
CVPR 2026 · Table 1
Multimodal LLMs
770.1R1-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
870.04Ours (base Qwen2.5)
Qwen2.5-14B controller
Training-Free Multimodal Large Language Model Orchestration
ICML 2026 · Table 1
Cross-modal learning
969.4PulseMind-72B
Qwen2.5-VL-72B
PulseMind: A Multi-Modal Medical Model for Real-World Clinical Diagnosis
AAAI 2026 · Table 1
Cross-modal learning
1069.11DPE (Iter 3)
Qwen3-VL-8B-Instruct
From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models
ICML 2026 · Table 1
RL from feedback
84 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

MMMU · score

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

#score ↑MethodPaperTopic
174.9Wan-WeaverWan-Weaver: Interleaved Multi-modal Generation via Decoupled Training
CVPR 2026 · Table 2
Multimodal LLMs
274.3UniDFlow (4B)Best of Both Worlds: Multimodal Reasoning and Generation via Unified Discrete Flow Matching
ICML 2026 · Table 1
Multimodal LLMs
369.3VLsI-7B
Qwen2-VL-7B
VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models
CVPR 2025 · Table 1
Multimodal LLMs
462.2VLM-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
561.8Qwen2.5-VL-7B w. RIL
Qwen2.5-VL-7B
Unified Reinforcement and Imitation Learning for Vision-Language Models
NeurIPS 2025 · Table 1
Multimodal LLMs
656.1Oryx-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
755.5CoS
InternVL2.5-MPO 8B
Unveiling Chain of Step Reasoning for Vision-Language Models with Fine-grained Rewards
NeurIPS 2025 · Table 1
Multimodal LLMs
855Mulberry-7B
Qwen2-VL-7B
Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Search
NeurIPS 2025 · Main Results table
Agents, code and math
954.77MiCo-7B
Qwen2.5VL-7B
MiCo: Multi-image Contrast for Reinforcement Visual Reasoning
NeurIPS 2025 · Table: Quantitative results on
Multimodal LLMs
1054.7NaViL-9B
Qwen3-8B
NaViL: Rethinking Scaling Properties of Native Multimodal Large Language Models under Data Constraints
NeurIPS 2025 · Appendix table: Comparison bet
Multimodal LLMs
43 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

MMMU · accuracy (val)

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

#accuracy ↑MethodPaperTopic
179.2AgoraWhy Keep Your Doubts to Yourself? Trading Visual Uncertainties in Multi-Agent Bandit Systems
ICLR 2026 · Table 1
LLM agents
273.4Agent0-VL-8B
Qwen3-VL-8B
Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning
ICML 2026 · Table 1
Multimodal LLMs
372.4Octopus-8B
Qwen3-VL-8B-Instruct
Learning Self-Correction in Vision–Language Models via Rollout Augmentation
ICML 2026 · Table 2
Multimodal LLMs
471.3MMFineReason-8B (SFT+RL)
Qwen3-VL-8B-Instruct
MMFineReason: Closing the Multimodal Reasoning Gap via Open Data-Centric Methods
NeurIPS 2026 · Table 3
Multimodal LLMs
571CARE
Qwen3-VL-8B-Instruct
CARE What Fails: Contrastive Anchored-REflection for Verifiable Multimodal Reasoning
CVPR 2026 · Table 1
Reasoning and chain of thought
668.1R-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
763.8AVAR-Thinker
Qwen2.5-VL-7B
From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal Reasoning
ICLR 2026 · Table 1
Multimodal LLMs
862DACO
InternVL3.5-8B-Instruct
Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs
CVPR 2026 · Table 1
Multimodal LLMs
960.8Perception-R1-7B
Qwen2.5-VL-7B-IT
Perception-R1: Advancing Multimodal Reasoning Capabilities of MLLMs via Visual Perception Reward
ICLR 2026 · Table 1
Multimodal LLMs
1059.1RLOO + SCS
Qwen2.5-VL-7B-Instruct
Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling
NeurIPS 2025 · Table 1
RL from feedback
32 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

MMMU · score (val)

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

#score ↑MethodPaperTopic
155.8Eagle2.5-8B
Qwen2.5-7B
Eagle 2.5: Boosting Long-Context Post-Training for Frontier Vision-Language Models
NeurIPS 2025 · Image benchmarks table
Multimodal LLMs
254.6Qwen2VL-7B + IPA
Qwen2VL-7B
Instruction-Oriented Preference Alignment for Enhancing Multi-Modal Comprehension Capability of MLLMs
ICCV 2025 · Table 1
Multimodal LLMs
354.22ThinkLite-VL + Image Jigsaw
ThinkLite-VL
Visual Jigsaw Post-Training Improves MLLMs
ICLR 2026 · Table 4
Multimodal LLMs
452.6VLM-UniDDT
Qwen3-VL-4B
UniDDT: Unifying Multimodal Understanding and Generation with Decoupled Diffusion Transformer
ECCV 2026 · Table 2
Human and embodied modeling
551.2VisionThink
Qwen2.5-VL-7B-Instruct
VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning
NeurIPS 2025 · Table: Effective Performance C
Multimodal LLMs
649.1Llama-3.2-11B + Lavender Full-FT
LM32-11B
Diffusion Instruction Tuning
ICML 2025 · Table
Multimodal LLMs
748.9Show-o2
Qwen2.5-7B-Instruct
Show-o2: Improved Native Unified Multimodal Models
NeurIPS 2025 · Multimodal understanding table
Human and embodied modeling
843.21MODEL-SAT
Wings
Capability Instruction Tuning
AAAI 2025 · Table 3
Instruction and fine-tuning
941.9InternLM-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
1039.8LLaVA-NeXT + RoboSpatial
LLaVA-NeXT
RoboSpatial: Teaching Spatial Understanding to 2D and 3D Vision-Language Models for Robotics
CVPR 2025 · Table: General-purpose benchma
Multimodal LLMs
3 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.