AI Paper Atlas

Research map / Benchmarks / MM-Vet

MM-Vet leaderboard

138 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MM-Vet. Best score: 79.6 (NaViL-9B, NeurIPS 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.

MM-Vet · score

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

#score ↑MethodPaperTopic
179.6NaViL-9B
Qwen3-8B
NaViL: Rethinking Scaling Properties of Native Multimodal Large Language Models under Data Constraints
NeurIPS 2025 · Appendix table: Comparison bet
Multimodal LLMs
277.3Qwen2.5-VL-7B w. RIL
Qwen2.5-VL-7B
Unified Reinforcement and Imitation Learning for Vision-Language Models
NeurIPS 2025 · Table 1
Multimodal LLMs
375.5MiniCPM-V 4.5MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
CVPR 2026 · Table 1
Multimodal LLMs
475.2VLsI-7B
Qwen2-VL-7B
VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models
CVPR 2025 · Table 1
Multimodal LLMs
571MuNG
Qwen2.5-VL-7B
Explore How to Inject Beneficial Noise in MLLMs
AAAI 2026 · Table 2
Multimodal LLMs
669.9Game-R1 7B
Qwen2.5-VL-7B
Game Ground Bench: Probing the Limits of LVLMs in Complex Semantic Grounding Across Game Universes
AAAI 2026 · Table 3
Multimodal LLMs
769.7AD-Loop (Ours)
BAGEL-7B
Synergizing Understanding and Generation with Interleaved Analyzing-Drafting Thinking
ICLR 2026 · Table 1
Multimodal LLMs
868.5VisionThink
Qwen2.5-VL-7B-Instruct
VisionThink: Smart and Efficient Vision Language Model via Reinforcement Learning
NeurIPS 2025 · Table: Comparison with Traditi
Multimodal LLMs
967.1LightFusionLightFusion: A Light-weighted, Double Fusion Framework for Unified Multimodal Understanding and Generation
ECCV 2026 · Table 1
Multimodal LLMs
1066.6UniVideo
Qwen2.5-VL-7B (MLLM backbone)
VOGUE: Unified Understanding, Generation, and Editing for Videos
ICLR 2026 · Table 1
Video generation
107 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

MM-Vet · accuracy

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

#accuracy ↑MethodPaperTopic
183.9Bee-8B-RL
Qwen3-8B
Bee: A High-Quality Corpus and Full-Stack Suite to Unlock Advanced Fully Open MLLMs
ICLR 2026 · Table 2
Multimodal LLMs
274.77DLMR (SFT)
Qwen2.5-VL-7B
Dual-Latent Memory Routing for Vision-Language Reasoning
ICML 2026 · Table 1
Multimodal LLMs
374.2LaCoT-Qwen-7B
Qwen2.5-VL-7B
Latent Chain-of-Thought for Visual Reasoning
NeurIPS 2025 · Table: Test accuracy on visual
Reasoning and chain of thought
473.4PointThinker-7B
Qwen2.5-VL-7B
PointThinker: Point-Incentivized Parallel Thinking for Multimodal Large Language Model
CVPR 2026 · Table 1
Multimodal LLMs
573.2InternVL2.5-8B-GenRecal (InternVL2.5-78B)
InternVL2.5-8B
GenRecal: Generation after Recalibration from Large to Small Vision-Language Models
ECCV 2026 · Table 1
Multimodal LLMs
671.4Lyra-Pro (74B)Lyra: An Efficient and Speech-Centric Framework for Omni-Cognition
ICCV 2025 · Table 3
Multimodal LLMs
771.2FAST-7B
Qwen2.5-VL-7B
Fast-Slow Thinking GRPO for Large Vision-Language Model Reasoning
NeurIPS 2025 · Main results table
Multimodal LLMs
868.35DPE (Iter 3)
Qwen2.5-VL-7B-Instruct
From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models
ICML 2026 · Table 1
RL from feedback
964.5ORION (Ours)
Qwen2.5-VL-7B
ORION: Decoupling and Alignment for Unified Autoregressive Understanding and Generation
ICLR 2026 · Table 3
Multimodal LLMs
1060.3LLaVA-CoT
Llama-3.2-11B-Vision-Instruct
LLaVA-CoT: Let Vision Language Models Reason Step-by-Step
ICCV 2025 · Table 2
Multimodal LLMs
12 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.