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.
hub Mixed citations
Improved Baselines with Visual Instruction Tuning
Mixed citation behavior. Most common role is background (69%).
97
Pith papers citing it
Background
69% of classified citations
abstract
Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this note, we show that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient. With simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~1 day on a single 8-A100 node. We hope this can make state-of-the-art LMM research more accessible. Code and model will be publicly available.
hub tools
citation-role summary
background 19
method 5
baseline 4
dataset 1
citation-polarity summary
claims ledger
- abstract Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this note, we show that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient. With simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~1
co-cited works
representative citing papers
MMMU provides 11.5K heterogeneous college-level multimodal questions that current models solve at 56-59% accuracy, establishing a new standard for expert multimodal evaluation.
ALMs unify pretrained atomistic encoder, LLM, and denoising diffusion via continuous projectors and staged training to reach SOTA on text-conditioned crystal prediction and de novo generation.
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
Introduces MMBU benchmark for VLMs in biomedicine and demonstrates that established benchmarks mask perception deficiencies in evaluated models.
FindIt is the first comprehensive benchmark for evaluating generalist MLLMs on promptable object detection, referring expression detection, instance-level detection, and video detection with standardized parsable outputs.
PixelRAG shows that operating RAG entirely over web screenshots outperforms text-based retrieval on NQ, SimpleQA, MMSearch, LiveVQA, and MoNaCo, with up to 18.1% accuracy gains and 3x token savings via image compression.
EvoCut is a training-free visual token compression technique that identifies important tokens via multi-layer evolution deviation, retaining 11.1% tokens with 94.4% average performance preserved on LLaVA-1.5-7B.
VLMs exhibit consistent vertical-distance entanglement in embeddings from perspective bias in natural images, producing accuracy gaps that a new synthetic benchmark SpatialTunnel exposes as model-intrinsic.
CAS mitigates object hallucinations in MLLMs by extracting two context preference vectors from designed conflict samples and applying signed residual injection at mid-early MLP layers without retraining or added latency.
VASA is a vision-guided agent for open ad-hoc segmentation that creates and validates masks through planning, tool use, and error recovery, outperforming baselines on the new PARS benchmark and RefCOCOm.
GRASP is a large-scale dataset and benchmark for social reasoning grounded in gaze and gesture events in multi-person videos, with Social Grounding Reward (SGR) proposed to improve model performance on GRASP-Bench.
The paper creates FISD, a controlled benchmark for composed image retrieval that removes query ambiguity via generative models, and proposes a multi-round agentic evaluation to assess models in interactive settings.
Instruction-tuned vision-language model PaveGPT, trained on a large unified pavement dataset, achieves substantial gains over general models in comprehensive, standard-compliant pavement condition assessment.
Panorama-Language Models with a sparse attention module and PanoVQA dataset deliver superior holistic reasoning on 360° adverse omni-scenes compared to stitched pinhole views.
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.
Introduces YesBut benchmark showing state-of-the-art multimodal models lag humans on interpreting humorous contradictions in comics.
VisualWebArena benchmark demonstrates that state-of-the-art multimodal agents still exhibit significant limitations on visually grounded web tasks.
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
HPP decouples perception from reasoning in long-video VLMs by having an LLM run iterative programmatic probes on hierarchically segmented video, reporting gains on LongVideoBench, EgoSchema, VideoMME, and MLVU.
EventDrive supplies a multi-task benchmark and EventDrive-VLM architecture that fuses event data, RGB, and language supervision, reporting gains in temporal precision and motion awareness for driving intelligence.
Qwen-RobotWorld is a language-conditioned video world model using Double-Stream MMDiT, an 8.6M-frame embodied corpus, and progressive curriculum training that ranks first on EWMBench and DreamGen Bench.
Social gaze consistency between interacting people is proposed as a new semantic cue orthogonal to low-level artifacts for detecting AI-generated images, with reported accuracy gains on vision and vision-language models.
SMART unlocks latent multi-vector capabilities in single-vector embedding models by applying late interaction to frozen hidden states shaped by contrastive training, yielding consistent gains on MMEB-V2 and visual document retrieval.
citing papers explorer
-
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.
-
MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI
MMMU provides 11.5K heterogeneous college-level multimodal questions that current models solve at 56-59% accuracy, establishing a new standard for expert multimodal evaluation.
-
Atomistic Language Models Understand and Generate Materials
ALMs unify pretrained atomistic encoder, LLM, and denoising diffusion via continuous projectors and staged training to reach SOTA on text-conditioned crystal prediction and de novo generation.
-
Fine-tuning Multi-modal LLMs with ART: Art-based Reinforcement Training
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
-
MMBU: A Massive Multi-modal Biomedical Understanding Benchmark to Probe the Perception Capabilities of Vision-Language Models
Introduces MMBU benchmark for VLMs in biomedicine and demonstrates that established benchmarks mask perception deficiencies in evaluated models.
-
FindIt: A Format-Informed Visual Detection Benchmark for Generalist Multimodal LLMs
FindIt is the first comprehensive benchmark for evaluating generalist MLLMs on promptable object detection, referring expression detection, instance-level detection, and video detection with standardized parsable outputs.
-
PIXELRAG: Web Screenshots Beat Text for Retrieval-Augmented Generation
PixelRAG shows that operating RAG entirely over web screenshots outperforms text-based retrieval on NQ, SimpleQA, MMSearch, LiveVQA, and MoNaCo, with up to 18.1% accuracy gains and 3x token savings via image compression.
-
EvoCut: Multi-Layer Evolution-Aware Visual Token Compression for Efficient Large Vision-Language Models
EvoCut is a training-free visual token compression technique that identifies important tokens via multi-layer evolution deviation, retaining 11.1% tokens with 94.4% average performance preserved on LLaVA-1.5-7B.
-
Why Far Looks Up: Probing Spatial Representation in Vision-Language Models
VLMs exhibit consistent vertical-distance entanglement in embeddings from perspective bias in natural images, producing accuracy gaps that a new synthetic benchmark SpatialTunnel exposes as model-intrinsic.
-
Rethinking Visual Neglect: Steering via Context-Preference for MLLM Hallucination Mitigation
CAS mitigates object hallucinations in MLLMs by extracting two context preference vectors from designed conflict samples and applying signed residual injection at mid-early MLP layers without retraining or added latency.
-
Vision Harnessing Agent for Open Ad-hoc Segmentation
VASA is a vision-guided agent for open ad-hoc segmentation that creates and validates masks through planning, tool use, and error recovery, outperforming baselines on the new PARS benchmark and RefCOCOm.
-
GRASP: Learning to Ground Social Reasoning in Multi-Person Non-Verbal Interactions
GRASP is a large-scale dataset and benchmark for social reasoning grounded in gaze and gesture events in multi-person videos, with Social Grounding Reward (SGR) proposed to improve model performance on GRASP-Bench.
-
A Sanity Check on Composed Image Retrieval
The paper creates FISD, a controlled benchmark for composed image retrieval that removes query ambiguity via generative models, and proposes a multi-round agentic evaluation to assess models in interactive settings.
-
Vision-Language Foundation Models for Comprehensive Automated Pavement Condition Assessment
Instruction-tuned vision-language model PaveGPT, trained on a large unified pavement dataset, achieves substantial gains over general models in comprehensive, standard-compliant pavement condition assessment.
-
More than the Sum: Panorama-Language Models for Adverse Omni-Scenes
Panorama-Language Models with a sparse attention module and PanoVQA dataset deliver superior holistic reasoning on 360° adverse omni-scenes compared to stitched pinhole views.
-
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.
-
Cracking the Code of Juxtaposition: Can AI Models Understand the Humorous Contradictions
Introduces YesBut benchmark showing state-of-the-art multimodal models lag humans on interpreting humorous contradictions in comics.
-
VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks
VisualWebArena benchmark demonstrates that state-of-the-art multimodal agents still exhibit significant limitations on visually grounded web tasks.
-
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
-
HPP: Hierarchical Programmatic Probing for Long Video Understanding by Decoupling Perception and Reasoning
HPP decouples perception from reasoning in long-video VLMs by having an LLM run iterative programmatic probes on hierarchically segmented video, reporting gains on LongVideoBench, EgoSchema, VideoMME, and MLVU.
-
EventDrive: Event Cameras for Vision-Language Driving Intelligence
EventDrive supplies a multi-task benchmark and EventDrive-VLM architecture that fuses event data, RGB, and language supervision, reporting gains in temporal precision and motion awareness for driving intelligence.
-
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
Qwen-RobotWorld is a language-conditioned video world model using Double-Stream MMDiT, an 8.6M-frame embodied corpus, and progressive curriculum training that ranks first on EWMBench and DreamGen Bench.
-
When Eyes Betray AI: Social Gaze Consistency as a Semantic Cue for AI-Generated Image Detection
Social gaze consistency between interacting people is proposed as a new semantic cue orthogonal to low-level artifacts for detecting AI-generated images, with reported accuracy gains on vision and vision-language models.
-
Your Embedding Model is SMARTer Than You Think
SMART unlocks latent multi-vector capabilities in single-vector embedding models by applying late interaction to frozen hidden states shaped by contrastive training, yielding consistent gains on MMEB-V2 and visual document retrieval.
-
Pareto-Enhanced Portrait Generation: Vision-Aligned Text Supervision for Alignment, Realism, and Aesthetics
A feature supervision approach using SigLIP 2 extracts multi-granularity vision-aligned text representations to supervise MM-DiT image branches, pushing the Pareto frontier for portrait generation across alignment, realism, and aesthetics.
-
From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models
Staged post-training that first solidifies visual perception before visual and textual reasoning improves VLM accuracy and shortens reasoning traces on visual math and perception benchmarks.
-
Revealing Interpretable Failure Modes of VLMs
REVELIO uncovers interpretable failure modes in VLMs by searching combinatorial concept spaces with diversity-aware beam search and Gaussian-process Thompson sampling, revealing vulnerabilities in autonomous driving and indoor robotics.
-
When Looking Is Not Enough: Visual Attention Structure Reveals Hallucination in MLLMs
Layer-wise Laplacian energy of visual attention reveals hallucination emergence in MLLMs and enables LaSCD, a closed-form logit remapping strategy that mitigates hallucinations while preserving general performance.
-
Text-Conditional JEPA for Learning Semantically Rich Visual Representations
TC-JEPA conditions masked feature prediction on text captions via sparse cross-attention to produce more semantically rich visual representations and outperforms contrastive methods on fine-grained tasks.
-
Catching the Infection Before It Spreads: Foresight-Guided Defense in Multi-Agent Systems
FLP uses multi-persona foresight simulation to detect infections via response diversity and applies local purification to reduce maximum cumulative infection rates in multi-agent systems from over 95% to below 5.47%.
-
PhysNote: Self-Knowledge Notes for Evolvable Physical Reasoning in Vision-Language Model
PhysNote lets VLMs externalize physical knowledge into hierarchical self-generated notes, stabilizing spatio-temporal reasoning and yielding 56.68% accuracy on PhysBench with a 4.96% gain over the best multi-agent baseline.
-
HypEHR: Hyperbolic Modeling of Electronic Health Records for Efficient Question Answering
HypEHR is a hyperbolic embedding model for EHR data that uses Lorentzian geometry and hierarchy-aware pretraining to answer clinical questions nearly as well as large language models but with much smaller size.
-
Every Picture Tells a Dangerous Story: Memory-Augmented Multi-Agent Jailbreak Attacks on VLMs
MemJack achieves 71.48% attack success rate on unmodified COCO val2017 images against Qwen3-VL-Plus by coordinating agents to map visual entities to malicious intents, apply multi-angle camouflage, and filter refusals via iterative nullspace projection while transferring strategies through a shared
-
Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs
DACO curates a 15,000-concept dictionary from 400K image-caption pairs and uses it to initialize an SAE that enables granular, concept-specific steering of MLLM activations, raising safety scores on MM-SafetyBench and JailBreakV while preserving general capabilities.
-
Entropy-Gradient Grounding: Training-Free Evidence Retrieval in Vision-Language Models
Entropy-gradient grounding uses model uncertainty to retrieve evidence regions in VLMs, improving performance on detail-critical and compositional tasks across multiple architectures.
-
Chat-Scene++: Exploiting Context-Rich Object Identification for 3D LLM
Chat-Scene++ improves 3D scene understanding in multimodal LLMs by representing scenes as context-rich object sequences with identifier tokens and grounded chain-of-thought reasoning, reaching state-of-the-art on five benchmarks using pre-trained encoders.
-
Adaptive Residual-Update Steering for Low-Overhead Hallucination Mitigation in Large Vision Language Models
RUDDER creates a persistent visual anchor by extracting CARD from prefill residuals and modulating its injection via an adaptive Beta Gate, cutting CHAIR_S by 24.4% and CHAIR_i by 23.6% on average across LLaVA, Idefics2, InstructBLIP and Qwen2.5-VL with >96% throughput.
-
World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training
World-Env replaces physical robot interactions with a world model-based virtual environment and VLM-guided rewards to enable efficient RL post-training for VLA models, showing gains with only five demonstrations per task.
-
Fourier Compressor: Frequency-Domain Visual Token Compression for Vision-Language Models
Fourier Compressor uses FFT to remove frequency-domain redundancy from visual tokens in VLMs, retaining over 96% accuracy with up to 83.8% FLOP reduction.
-
PRISM: Programmatic Reasoning with Image Sequence Manipulation for LVLM Jailbreaking
PRISM decomposes harmful instructions into benign visual gadgets and directs LVLMs via prompts to compose them through reasoning into harmful outputs, achieving ASR over 0.90 on SafeBench.
-
ReGATE: Learning Faster and Better with Fewer Tokens in MLLMs
ReGATE introduces a teacher-student adaptive token elision method that reduces training tokens to 38% while matching or exceeding baseline accuracy on multimodal benchmarks.
-
When 'YES' Meets 'BUT': Can Large Models Comprehend Contradictory Humor Through Comparative Reasoning?
Presents YesBut (V2) benchmark and shows state-of-the-art VLMs significantly underperform humans on tasks requiring comparative reasoning for contradictory humor in comics.
-
SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training
Reinforcement learning post-training enables generalization to unseen textual rule variants and visual changes in foundation models, while supervised fine-tuning primarily leads to memorization.
-
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.
-
Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization
Mixed Preference Optimization with the MMPR dataset boosts multimodal CoT reasoning, lifting InternVL2-8B to 67.0 accuracy on MathVista (+8.7 points) and matching the 76B model.
-
ChatSearch: a Dataset and a Generative Retrieval Model for General Conversational Image Retrieval
Presents ChatSearch dataset and ChatSearcher generative model for conversational image retrieval on open-domain images, claiming superior performance on the new dataset and competitive results elsewhere.
-
OpenVLA: An Open-Source Vision-Language-Action Model
OpenVLA achieves 16.5% higher task success than the 55B RT-2-X model across 29 tasks with 7x fewer parameters while enabling effective fine-tuning and quantization without performance loss.
-
Chameleon: Mixed-Modal Early-Fusion Foundation Models
Chameleon is an early-fusion token model that handles mixed image-text sequences for understanding and generation, achieving competitive or superior performance to larger models like Llama-2, Mixtral, and Gemini-Pro on captioning, VQA, text, and image tasks.
-
Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding
Hunyuan-DiT is a new multi-resolution diffusion transformer that achieves state-of-the-art Chinese text-to-image generation through custom architecture, data pipelines, and multimodal caption refinement.
-
SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generation
SEED-X is a unified multimodal foundation model that handles multi-granularity visual semantics for both comprehension and generation across arbitrary image sizes and ratios.