AI Paper Atlas

Research map / Benchmarks / MVBench

MVBench leaderboard

100 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MVBench. Best accuracy: 90 (MACD, 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.

MVBench · accuracy

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

#accuracy ↑MethodPaperTopic
190MACD
Qwen2-VL-7B
MACD: Model-Aware Contrastive Decoding via Counterfactual Data for Video-LLMs
ICML 2026 · Table 1
Video-language understanding
283.23GAM-Agent
InternVL3-14B
GAM-Agent: Game-Theoretic and Uncertainty-Aware Collaboration for Complex Visual Reasoning
NeurIPS 2025 · Table: Image and Video Underst
LLM agents
378.5InternVL3.5-8B + G2F-RAG
InternVL3.5-8B
Graph-to-Frame RAG: Visual-Space Knowledge Fusion for Training-Free and Auditable Video Reasoning
CVPR 2026 · Table 1
Video-language understanding
477.1CoF-InternVL3-8BChain-of-Frames: Advancing Video Understanding in Multimodal LLMs via Frame-Aware Reasoning
CVPR 2026 · Table 1
Video-language understanding
576.6AdaCodec (comparable token budget)
Qwen3-VL-8B
AdaCodec: A Predictive Visual Code for Video MLLMs
NeurIPS 2026 · Table 1
Video-language understanding
675.9Molmo2-8B
Qwen3-8B
Molmo2: Open Weights and Data for Vision-Language Models with Video Understanding and Grounding
CVPR 2026 · Table 2
Video-language understanding
775.8CMMCoT (Ours)
Qwen2.5-VL-7B
CMMCoT: Enhancing Complex Multi-Image Comprehension via Multi-Modal Chain-of-Thought and Memory Augmentation
AAAI 2026 · Table 1
Multimodal LLMs
875.6PPLLaVA
InternVL3
PPLLaVA: Varied Video Sequence Understanding With Prompt Guidance
ICLR 2026 · Table 2
Video-language understanding
974.5STAVEQ2 72B
Qwen2-VL 72B
Enhancing Temporal Understanding in Video-LLMs through Stacked Temporal Attention in Vision Encoders
NeurIPS 2025 · Table: Accuracy on video under
Video-language understanding
1074VideoChat-Flash@448
Qwen2-7B
VideoChat-Flash: Hierarchical Compression for Long-Context Video Modeling
ICLR 2026 · Table 1
Video-language understanding
76 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

MVBench · score

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

#score ↑MethodPaperTopic
174.8Eagle2.5-8B
Qwen2.5-7B
Eagle 2.5: Boosting Long-Context Post-Training for Frontier Vision-Language Models
NeurIPS 2025 · Video benchmarks table
Multimodal LLMs
270.1Oryx-1.5-32B
Qwen-2.5-32B
Oryx MLLM: On-Demand Spatial-Temporal Understanding at Arbitrary Resolution
ICLR 2025 · General Temporal Understanding
Video-language understanding
370EgoThinker
Qwen2VL
EgoThinker: Unveiling Egocentric Reasoning with Spatio-Temporal CoT
NeurIPS 2025 · Table: general video benchmark
Video-language understanding
463.6MIA-DPO
InternLM-XC2.5
MIA-DPO: Multi-Image Augmented Direct Preference Optimization For Large Vision-Language Models
ICLR 2025 · Main results on multi-image be
Multimodal LLMs
562.3UniARUnified Multimodal Autoregressive Modeling with Shared Context—Visual Tokenizer is Key to Unification
ICML 2026 · Table 4
Human and embodied modeling
662LLaVA-Video-RRPO
LLaVA-Video-Qwen-7B
Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization
NeurIPS 2025 · Table: Comparison with off-the
Video-language understanding
761.4GRPO + ReAgent-V
Qwen2.5-VL-7B
ReAgent-V: A Reward-Driven Multi-Agent Framework for Video Understanding
NeurIPS 2025 · Table 2
Video-language understanding
859.6VideoRefer
VideoLLaMA2.1
VideoRefer Suite: Advancing Spatial-Temporal Object Understanding with Video LLM
CVPR 2025 · Table 4
Video-language understanding
957.58Distinct and balanced positional embeddingsMASH-VLM: Mitigating Action-Scene Hallucination in Video-LLMs through Disentangled Spatial-Temporal Representations
CVPR 2025 · Table 7
Video-language understanding
1057.3HoliTom
LLaVA-OneVision-7B
HoliTom: Holistic Token Merging for Fast Video Large Language Models
NeurIPS 2025 · Table 1
Video-language understanding
4 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.