MMMU-Pro is a stricter multimodal benchmark that removes text-only solvable questions, augments options, and requires reading text from images, yielding substantially lower model scores of 16.8-26.9%.
hub Baseline reference
BLINK: Multimodal Large Language Models Can See but Not Perceive
Baseline reference. 62% of citing Pith papers use this work as a benchmark or comparison.
32
Pith papers citing it
Baseline
62% of classified citations
abstract
We introduce Blink, a new benchmark for multimodal language models (LLMs) that focuses on core visual perception abilities not found in other evaluations. Most of the Blink tasks can be solved by humans "within a blink" (e.g., relative depth estimation, visual correspondence, forensics detection, and multi-view reasoning). However, we find these perception-demanding tasks cast significant challenges for current multimodal LLMs because they resist mediation through natural language. Blink reformats 14 classic computer vision tasks into 3,807 multiple-choice questions, paired with single or multiple images and visual prompting. While humans get 95.70% accuracy on average, Blink is surprisingly challenging for existing multimodal LLMs: even the best-performing GPT-4V and Gemini achieve accuracies of 51.26% and 45.72%, only 13.17% and 7.63% higher than random guessing, indicating that such perception abilities have not "emerged" yet in recent multimodal LLMs. Our analysis also highlights that specialist CV models could solve these problems much better, suggesting potential pathways for future improvements. We believe Blink will stimulate the community to help multimodal LLMs catch up with human-level visual perception.
hub tools
citation-role summary
dataset 7
background 5
baseline 1
citation-polarity summary
representative citing papers
MME-RealWorld is the largest manually annotated high-resolution benchmark for MLLMs, where even the best models achieve less than 60% accuracy on challenging real-world tasks.
Perceval is a perception-centric PRM that detects token-level perceptual errors in VLMs, supporting token-advantage RL training and iterative test-time scaling for improved reasoning.
EmbodiedMidtrain mid-trains VLMs on curated VLA-aligned data subsets to improve downstream performance on robot manipulation benchmarks.
Omni-R1 unifies multimodal reasoning by generating intermediate images during the process in a SFT-plus-RL framework, with an Omni-R1-Zero variant that matches or exceeds it using only text data.
SandboxVLM enhances VLMs' spatial intelligence by encoding 3D geometry with abstract bounding boxes in a four-stage zero-shot pipeline, yielding an 8.3% improvement on SAT Real benchmark.
LLaVA-NeXT-Interleave unifies multi-image, video, and 3D capabilities in large multimodal models via a new 1.18M-sample interleaved dataset and benchmark, achieving leading results across those tasks while preserving single-image performance.
Cambrian-1 is a vision-centric multimodal LLM family that evaluates over 20 vision encoders, introduces CV-Bench and the Spatial Vision Aggregator, and releases open models, code, and data achieving strong performance on visual grounding tasks.
MuirBench is a new benchmark showing that top multimodal LLMs struggle with robust multi-image understanding, with GPT-4o at 68% and open-source models below 33% accuracy.
Semantic Generative Tuning uses image segmentation as a generative proxy to align misaligned representation spaces in unified multimodal models and improve both perception and generative layout fidelity.
Video MLLMs show an audio-visual Clever Hans effect relying on visual-acoustic correlations rather than audio verification; Thud interventions diagnose it and a 10K-sample preference alignment improves intervention performance by 28 points.
Data curation alone raises VLM accuracy by more than 11 points on average across many benchmarks while cutting required training compute by up to 87 times.
Visual latents in MLLMs are systematically silenced by autoregressive training but can be unsilenced at inference via query-guided contrastive alignment followed by a confidence-progression reward.
RetentiveKV uses entropy to drive state-space model transitions that retain and reactivate low-attention visual tokens in a continuous memory instead of pruning them, delivering 5x KV cache compression and 1.5x faster decoding.
Multimodal LLMs significantly underperform humans at spotting objects that break 3D consistency in multi-view image pairs.
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
Multimodal foundation models achieve respectable but sub-specialist performance on semantic vision tasks and weaker results on geometric tasks when evaluated through prompt chaining on established benchmarks.
GLM-4.5V reaches state-of-the-art results on 42 multimodal benchmarks among open-source models of similar size by applying reinforcement learning with curriculum sampling to a strong vision foundation model.
ViGoRL introduces visually grounded RL that anchors reasoning steps to image coordinates and uses multi-turn zooming to outperform standard RL and supervised baselines on spatial and GUI reasoning benchmarks.
InternVL3-78B sets a new open-source SOTA of 72.2 on MMMU via native joint multimodal pre-training, V2PE, MPO, and test-time scaling while remaining competitive with proprietary models.
VPiT enables pretrained LLMs to perform both visual understanding and generation by predicting discrete text tokens and continuous visual tokens, with understanding data proving more effective than generation-specific data.
InternVL 2.5 is the first open-source MLLM to surpass 70% on the MMMU benchmark via model, data, and test-time scaling, with a 3.7-point gain from chain-of-thought reasoning.
Depth Anything V2 delivers finer, more robust monocular depth predictions by replacing real labeled images with synthetic data, scaling the teacher model, and using large-scale pseudo-labeled real images for student training.
Phi-3-mini (3.8B params, 3.3T tokens) reaches 69% MMLU and 8.38 MT-bench, matching larger models, with scaled-up 7B/14B variants and phi-3.5 extensions for multilingual, MoE, and vision capabilities.
citing papers explorer
-
MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark
MMMU-Pro is a stricter multimodal benchmark that removes text-only solvable questions, augments options, and requires reading text from images, yielding substantially lower model scores of 16.8-26.9%.
-
MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?
MME-RealWorld is the largest manually annotated high-resolution benchmark for MLLMs, where even the best models achieve less than 60% accuracy on challenging real-world tasks.
-
Improving Vision-language Models with Perception-centric Process Reward Models
Perceval is a perception-centric PRM that detects token-level perceptual errors in VLMs, supporting token-advantage RL training and iterative test-time scaling for improved reasoning.
-
EmbodiedMidtrain: Bridging the Gap between Vision-Language Models and Vision-Language-Action Models via Mid-training
EmbodiedMidtrain mid-trains VLMs on curated VLA-aligned data subsets to improve downstream performance on robot manipulation benchmarks.
-
Omni-R1: Towards the Unified Generative Paradigm for Multimodal Reasoning
Omni-R1 unifies multimodal reasoning by generating intermediate images during the process in a SFT-plus-RL framework, with an Omni-R1-Zero variant that matches or exceeds it using only text data.
-
Abstract 3D Perception for Spatial Intelligence in Vision-Language Models
SandboxVLM enhances VLMs' spatial intelligence by encoding 3D geometry with abstract bounding boxes in a four-stage zero-shot pipeline, yielding an 8.3% improvement on SAT Real benchmark.
-
LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models
LLaVA-NeXT-Interleave unifies multi-image, video, and 3D capabilities in large multimodal models via a new 1.18M-sample interleaved dataset and benchmark, achieving leading results across those tasks while preserving single-image performance.
-
Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs
Cambrian-1 is a vision-centric multimodal LLM family that evaluates over 20 vision encoders, introduces CV-Bench and the Spatial Vision Aggregator, and releases open models, code, and data achieving strong performance on visual grounding tasks.
-
MuirBench: A Comprehensive Benchmark for Robust Multi-image Understanding
MuirBench is a new benchmark showing that top multimodal LLMs struggle with robust multi-image understanding, with GPT-4o at 68% and open-source models below 33% accuracy.
-
Semantic Generative Tuning for Unified Multimodal Models
Semantic Generative Tuning uses image segmentation as a generative proxy to align misaligned representation spaces in unified multimodal models and improve both perception and generative layout fidelity.
-
When Vision Speaks for Sound
Video MLLMs show an audio-visual Clever Hans effect relying on visual-acoustic correlations rather than audio verification; Thud interventions diagnose it and a 10K-sample preference alignment improves intervention performance by 28 points.
-
20/20 Vision Language Models: A Prescription for Better VLMs through Data Curation Alone
Data curation alone raises VLM accuracy by more than 11 points on average across many benchmarks while cutting required training compute by up to 87 times.
-
Visual Latents Know More Than They Say: Unsilencing Latent Reasoning in MLLMs
Visual latents in MLLMs are systematically silenced by autoregressive training but can be unsilenced at inference via query-guided contrastive alignment followed by a confidence-progression reward.
-
RetentiveKV: State-Space Memory for Uncertainty-Aware Multimodal KV Cache Eviction
RetentiveKV uses entropy to drive state-space model transitions that retain and reactivate low-attention visual tokens in a continuous memory instead of pruning them, delivering 5x KV cache compression and 1.5x faster decoding.
-
Multimodal Language Models Cannot Spot Spatial Inconsistencies
Multimodal LLMs significantly underperform humans at spotting objects that break 3D consistency in multi-view image pairs.
-
InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
-
How Well Does GPT-4o Understand Vision? Evaluating Multimodal Foundation Models on Standard Computer Vision Tasks
Multimodal foundation models achieve respectable but sub-specialist performance on semantic vision tasks and weaker results on geometric tasks when evaluated through prompt chaining on established benchmarks.
-
GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
GLM-4.5V reaches state-of-the-art results on 42 multimodal benchmarks among open-source models of similar size by applying reinforcement learning with curriculum sampling to a strong vision foundation model.
-
Grounded Reinforcement Learning for Visual Reasoning
ViGoRL introduces visually grounded RL that anchors reasoning steps to image coordinates and uses multi-turn zooming to outperform standard RL and supervised baselines on spatial and GUI reasoning benchmarks.
-
InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
InternVL3-78B sets a new open-source SOTA of 72.2 on MMMU via native joint multimodal pre-training, V2PE, MPO, and test-time scaling while remaining competitive with proprietary models.
-
MetaMorph: Multimodal Understanding and Generation via Instruction Tuning
VPiT enables pretrained LLMs to perform both visual understanding and generation by predicting discrete text tokens and continuous visual tokens, with understanding data proving more effective than generation-specific data.
-
Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
InternVL 2.5 is the first open-source MLLM to surpass 70% on the MMMU benchmark via model, data, and test-time scaling, with a 3.7-point gain from chain-of-thought reasoning.
-
Depth Anything V2
Depth Anything V2 delivers finer, more robust monocular depth predictions by replacing real labeled images with synthetic data, scaling the teacher model, and using large-scale pseudo-labeled real images for student training.
-
Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
Phi-3-mini (3.8B params, 3.3T tokens) reaches 69% MMLU and 8.38 MT-bench, matching larger models, with scaled-up 7B/14B variants and phi-3.5 extensions for multilingual, MoE, and vision capabilities.
-
What's Holding Back Latent Visual Reasoning?
Latent visual reasoning fails in current models because standard datasets make oracle latents uninformative and inference-time latents collapse away from useful representations.
-
Kimi K2.5: Visual Agentic Intelligence
Kimi K2.5 combines joint text-vision training with an Agent Swarm parallel orchestration framework to reach claimed state-of-the-art results on coding, vision, reasoning, and agent tasks while cutting latency up to 4.5 times.
-
Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs
Phi-4-Mini achieves strong math and coding performance with only 3.8B parameters via high-quality synthetic data, while Phi-4-Multimodal uses Mixture-of-LoRAs to integrate modalities and top speech recognition leaderboards.
-
LLaVA-OneVision: Easy Visual Task Transfer
LLaVA-OneVision is the first single open LMM to simultaneously achieve strong performance in single-image, multi-image, and video scenarios with cross-scenario transfer capabilities.
-
Gemma 3 Technical Report
Gemma 3 introduces multimodal open models with architectural changes for efficient long context, trained via distillation and a new post-training recipe that makes the 4B version competitive with prior 27B models and the 27B version comparable to Gemini-1.5-Pro.
- ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison
- ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop
- The Cartesian Shortcut: Re-evaluate Vision Reasoning in Polar Coordinate Space