#MLLMs

29개의 포스트

[논문리뷰] CurveStream: Boosting Streaming Video Understanding in MLLMs via Curvature-Aware Hierarchical Visual Memory Management

Tao Chen이 arXiv에 게시한 'CurveStream: Boosting Streaming Video Understanding in MLLMs via Curvature-Aware Hierarchical Visual Memory Management' 논문에 대한 자세한 리뷰입니다.

#Review #Streaming Video Understanding #MLLMs #Memory Management #Curvature Score #Hierarchical Visual Memory #Catastrophic Forgetting

2026년 3월 22일

[논문리뷰] TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering

Hao Feng이 arXiv에 게시한 'TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Text Rendering #Reinforcement Learning #Structural Anomaly Perception #Reward Modeling #Text-to-Image Generation #OCR #MLLMs #Data Augmentation

2026년 2월 24일

[논문리뷰] BiManiBench: A Hierarchical Benchmark for Evaluating Bimanual Coordination of Multimodal Large Language Models

arXiv에 게시된 'BiManiBench: A Hierarchical Benchmark for Evaluating Bimanual Coordination of Multimodal Large Language Models' 논문에 대한 자세한 리뷰입니다.

#Review #Bimanual Manipulation #MLLMs #Robotics Benchmark #Spatial Reasoning #Action Planning #End-Effector Control #Embodied AI #Multimodal LLMs

2026년 2월 18일

[논문리뷰] Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception

arXiv에 게시된 'Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Perception #Fine-Grained Analysis #Knowledge Distillation #Region-to-Image #MLLMs #ZoomBench #Reinforcement Learning

2026년 2월 15일

[논문리뷰] Thinking with Drafting: Optical Decompression via Logical Reconstruction

arXiv에 게시된 'Thinking with Drafting: Optical Decompression via Logical Reconstruction' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Reasoning #Visual Algebra #Domain-Specific Language #Optical Decompression #Logical Reconstruction #Bar Model #MLLMs #Verification

2026년 2월 12일

[논문리뷰] MetaphorStar: Image Metaphor Understanding and Reasoning with End-to-End Visual Reinforcement Learning

Hongsheng Li이 arXiv에 게시한 'MetaphorStar: Image Metaphor Understanding and Reasoning with End-to-End Visual Reinforcement Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Image Metaphor Understanding #Visual Reasoning #Reinforcement Learning #MLLMs #TFQ-GRPO #End-to-End Learning #Cognitive AI

2026년 2월 12일

[논문리뷰] HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding

arXiv에 게시된 'HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding' 논문에 대한 자세한 리뷰입니다.

#Review #Streaming Video Understanding #KV Cache Management #Hierarchical Memory #MLLMs #Low Latency #Training-free #Memory Efficiency

2026년 1월 22일

[논문리뷰] Taming Hallucinations: Boosting MLLMs' Video Understanding via Counterfactual Video Generation

arXiv에 게시된 'Taming Hallucinations: Boosting MLLMs' Video Understanding via Counterfactual Video Generation' 논문에 대한 자세한 리뷰입니다.

#Review #MLLMs #Video Understanding #Hallucinations #Counterfactual Generation #Diffusion Models #Reinforcement Learning #QA Dataset #DNA-Train

2026년 1월 4일

[논문리뷰] GroundingME: Exposing the Visual Grounding Gap in MLLMs through Multi-Dimensional Evaluation

arXiv에 게시된 'GroundingME: Exposing the Visual Grounding Gap in MLLMs through Multi-Dimensional Evaluation' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Grounding #MLLMs #Benchmark #Multi-Dimensional Evaluation #Rejection Capability #Test-Time Scaling #Data Mixture Training

2025년 12월 21일

[논문리뷰] Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task

arXiv에 게시된 'Tool-Augmented Spatiotemporal Reasoning for Streamlining Video Question Answering Task' 논문에 대한 자세한 리뷰입니다.

#Review #VideoQA #MLLMs #Tool Learning #Spatiotemporal Reasoning #Video Toolkit #Agentic AI

2025년 12월 11일

[논문리뷰] WiseEdit: Benchmarking Cognition- and Creativity-Informed Image Editing

Wendong Bu이 arXiv에 게시한 'WiseEdit: Benchmarking Cognition- and Creativity-Informed Image Editing' 논문에 대한 자세한 리뷰입니다.

#Review #Image Editing #Benchmarking #Cognitive AI #Creativity #Multimodal AI #Knowledge-based Reasoning #Diffusion Models #MLLMs

2025년 12월 1일

[논문리뷰] StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos

arXiv에 게시된 'StreamGaze: Gaze-Guided Temporal Reasoning and Proactive Understanding in Streaming Videos' 논문에 대한 자세한 리뷰입니다.

#Review #Streaming Video Understanding #Gaze-Guided AI #Temporal Reasoning #Proactive AI #MLLMs #Eye Tracking #Benchmark #Human-Computer Interaction

2025년 12월 1일

[논문리뷰] VideoSSR: Video Self-Supervised Reinforcement Learning

arXiv에 게시된 'VideoSSR: Video Self-Supervised Reinforcement Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Video Understanding #Self-Supervised Learning #Reinforcement Learning #MLLMs #Pretext Tasks #Verifiable Rewards #Data Generation #Temporal Grounding

2025년 11월 11일

[논문리뷰] When Visualizing is the First Step to Reasoning: MIRA, a Benchmark for Visual Chain-of-Thought

arXiv에 게시된 'When Visualizing is the First Step to Reasoning: MIRA, a Benchmark for Visual Chain-of-Thought' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal AI #Visual Reasoning #Chain-of-Thought (CoT)#Benchmark #Image Generation #MLLMs #Visual-CoT

2025년 11월 9일

[논문리뷰] UME-R1: Exploring Reasoning-Driven Generative Multimodal Embeddings

Jinsong Su이 arXiv에 게시한 'UME-R1: Exploring Reasoning-Driven Generative Multimodal Embeddings' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Embeddings #Generative AI #Reasoning #Reinforcement Learning #MLLMs #Supervised Fine-tuning #Information Retrieval #Unified Embeddings

2025년 11월 9일

[논문리뷰] GUI-AIMA: Aligning Intrinsic Multimodal Attention with a Context Anchor for GUI Grounding

Wanrong Zhu이 arXiv에 게시한 'GUI-AIMA: Aligning Intrinsic Multimodal Attention with a Context Anchor for GUI Grounding' 논문에 대한 자세한 리뷰입니다.

#Review #GUI Grounding #Multimodal Attention #MLLMs #Coordinate-Free #Visual Grounding #Attention Weighting #Anchor Token

2025년 11월 9일

[논문리뷰] VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations

Jiayi Zhang이 arXiv에 게시한 'VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations' 논문에 대한 자세한 리뷰입니다.

#Review #Visualization Quality Assessment #MLLMs #Benchmark #Aesthetics #Fidelity #Expressiveness #Fine-tuning #Reinforcement Learning

2025년 10월 29일

[논문리뷰] IF-VidCap: Can Video Caption Models Follow Instructions?

arXiv에 게시된 'IF-VidCap: Can Video Caption Models Follow Instructions?' 논문에 대한 자세한 리뷰입니다.

#Review #Video Captioning #Instruction Following #MLLMs #Benchmark #Controllable Generation #Multimodal Evaluation #Fine-tuning

2025년 10월 22일

[논문리뷰] Scaling Language-Centric Omnimodal Representation Learning

arXiv에 게시된 'Scaling Language-Centric Omnimodal Representation Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Embeddings #MLLMs #Contrastive Learning #Cross-modal Alignment #Generative Pretraining #Representation Learning #Scaling Laws

2025년 10월 15일

[논문리뷰] Multimodal Prompt Optimization: Why Not Leverage Multiple Modalities for MLLMs

arXiv에 게시된 'Multimodal Prompt Optimization: Why Not Leverage Multiple Modalities for MLLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal AI #Prompt Optimization #MLLMs #Bayesian Optimization #Cross-modal Alignment #Prompt Engineering #Generative AI #Exploration-Exploitation

2025년 10월 13일

[논문리뷰] Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods

Yiyu Wang이 arXiv에 게시한 'Are We Using the Right Benchmark: An Evaluation Framework for Visual Token Compression Methods' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Token Compression #MLLMs #Evaluation Framework #Benchmarking #Downsampling #Data Filtering #Model Efficiency

2025년 10월 9일

[논문리뷰] Visual Jigsaw Post-Training Improves MLLMs

Lewei Lu이 arXiv에 게시한 'Visual Jigsaw Post-Training Improves MLLMs' 논문에 대한 자세한 리뷰입니다.

#Review #MLLMs #Post-training #Self-supervised Learning #Visual Understanding #Jigsaw Puzzles #RLVR #Multimodal Perception #Spatial Reasoning

2025년 9월 30일

[논문리뷰] Video-MTR: Reinforced Multi-Turn Reasoning for Long Video Understanding

Lionel Ni이 arXiv에 게시한 'Video-MTR: Reinforced Multi-Turn Reasoning for Long Video Understanding' 논문에 대한 자세한 리뷰입니다.

#Review #Long Video Understanding #Reinforcement Learning #Multi-Turn Reasoning #MLLMs #Video Segment Selection #Bi-level Reward #Question Answering

2025년 9월 5일

[논문리뷰] From reactive to cognitive: brain-inspired spatial intelligence for embodied agents

Songming Liu이 arXiv에 게시한 'From reactive to cognitive: brain-inspired spatial intelligence for embodied agents' 논문에 대한 자세한 리뷰입니다.

#Review #Spatial Cognition #Embodied Agents #Brain-inspired AI #Cognitive Map #Spatial Memory #MLLMs #Navigation

2025년 9월 2일

[논문리뷰] We-Math 2.0: A Versatile MathBook System for Incentivizing Visual Mathematical Reasoning

Xiaowan Wang이 arXiv에 게시한 'We-Math 2.0: A Versatile MathBook System for Incentivizing Visual Mathematical Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Mathematical Reasoning #MLLMs #Knowledge System #Reinforcement Learning #Curriculum Learning #Dataset Construction #Mathematical Benchmark

2025년 8월 15일

[논문리뷰] VisR-Bench: An Empirical Study on Visual Retrieval-Augmented Generation for Multilingual Long Document Understanding

Tong Yu이 arXiv에 게시한 'VisR-Bench: An Empirical Study on Visual Retrieval-Augmented Generation for Multilingual Long Document Understanding' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Retrieval #Retrieval-Augmented Generation #Long Document Understanding #Multilingual NLP #Visual QA #Benchmark #MLLMs #Table Understanding

2025년 8월 12일

[논문리뷰] UI-AGILE: Advancing GUI Agents with Effective Reinforcement Learning and Precise Inference-Time Grounding

Bingqi Chen이 arXiv에 게시한 'UI-AGILE: Advancing GUI Agents with Effective Reinforcement Learning and Precise Inference-Time Grounding' 논문에 대한 자세한 리뷰입니다.

#Review #GUI Agents #Reinforcement Learning #Grounding #MLLMs #Reward Function #Resampling #Visual Noise Reduction

2025년 8월 11일

[논문리뷰] InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization

Pengxiang Li이 arXiv에 게시한 'InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization' 논문에 대한 자세한 리뷰입니다.

#Review #GUI Grounding #MLLMs #Reinforcement Learning #Policy Optimization #Exploration Strategy #Semantic Alignment #Adaptive Exploration Reward #Human-Computer Interaction

2025년 8월 11일

[논문리뷰] TARS: MinMax Token-Adaptive Preference Strategy for Hallucination Reduction in MLLMs

Jiasheng Tang이 arXiv에 게시한 'TARS: MinMax Token-Adaptive Preference Strategy for Hallucination Reduction in MLLMs' 논문에 대한 자세한 리뷰입니다.

#Review #MLLMs #Hallucination Reduction #Preference Optimization #Min-Max Optimization #Token-Adaptive Strategy #Spectral Regularization #Visual Grounding

2025년 8월 2일