AI Paper Atlas

Research map / Benchmarks / MMStar

MMStar leaderboard

139 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MMStar. Best accuracy: 81.4 (L2-VMAS, 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.

MMStar · accuracy

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

#accuracy ↑MethodPaperTopic
181.4L2-VMAS
Qwen3-VL-8B-Thinking
Dual Latent Memory for Visual Multi-agent System
ICML 2026 · Table 1
LLM agents
275.77+ ReaGEN (4 iter)
Qwen3-VL-4B
ReaGEN: Adaptive Generation of Structured Chains-of-Thought for Efficient Multimodal Reasoning
CVPR 2026 · Table 1
Multimodal LLMs
375.7Ours (+sampling)
Qwen3-VL-4B-Thinking
When to Think and When to Look: Uncertainty-Guided Lookback
CVPR 2026 · Table 2
Multimodal LLMs
475.2Octopus-8B
Qwen3-VL-8B-Instruct
Learning Self-Correction in Vision–Language Models via Rollout Augmentation
ICML 2026 · Table 2
Multimodal LLMs
575.2MMFineReason-8B (SFT+RL)
Qwen3-VL-8B-Instruct
MMFineReason: Closing the Multimodal Reasoning Gap via Open Data-Centric Methods
NeurIPS 2026 · Table 3
Multimodal LLMs
673.1R-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
773.07Staged training
Qwen3-VL-8B
From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models
ICML 2026 · Table 2
Multimodal LLMs
872.3VaLR-M
Qwen2.5-VL-7B
Vision-aligned Latent Reasoning for Multi-modal Large Language Model
ICML 2026 · Table 2
Multimodal LLMs
972.13DPE (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
1072.1MiniCPM-V 4.5MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
CVPR 2026 · Table 1
Multimodal LLMs
94 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

MMStar · score

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

#score ↑MethodPaperTopic
171.1DenseMLLM
DenseMLLM-4B
DenseMLLM: Standard Multimodal LLMs for Dense Prediction
ICML 2026 · Table 2
Multimodal LLMs
266.5Qwen2.5VL-7B + IPA
Qwen2.5VL-7B
Instruction-Oriented Preference Alignment for Enhancing Multi-Modal Comprehension Capability of MLLMs
ICCV 2025 · Table 1
Multimodal LLMs
365.6MiCo-7B
Qwen2.5VL-7B
MiCo: Multi-image Contrast for Reinforcement Visual Reasoning
NeurIPS 2025 · Table: Quantitative results on
Multimodal LLMs
463.5CoS
InternVL2.5-MPO 8B
Unveiling Chain of Step Reasoning for Vision-Language Models with Fine-grained Rewards
NeurIPS 2025 · Table 1
Multimodal LLMs
563.1InternVL-Spatial-8B
InternVL2.5-8B
InternSpatial: A Comprehensive Dataset for Spatial Reasoning in Vision-Language Models
ICLR 2026 · Table 4
Multimodal LLMs
662.7Qwen2.5-VL +Cold Start +Multimodal RL
Qwen2.5-VL
Open Vision Reasoner: Transferring Linguistic Cognitive Behavior for Visual Reasoning
NeurIPS 2025 · Perception benchmark table
Multimodal LLMs
762.4FeedQuill-7B
LLaVA-Onevision-7B
Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning
ICLR 2025 · Table, Main Results
Multimodal LLMs
862.3BlueLM-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
961.5Insight-V
Our Base Model
Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models
CVPR 2025 · Table 1
Multimodal LLMs
1061.4HaploVL-7B-Pro
Qwen2.5-7B
HaploVL: A Single-Transformer Baseline for Multi-Modal Understanding
ICML 2025 · Main results table
Language, safety and interpretability
25 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.