Review

[논문리뷰] When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling

arXiv에 게시된 'When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling' 논문에 대한 자세한 리뷰입니다.

#Review #LLM Ensembling #Token-level Ensembling #Speculative Decoding #Tokenization Mismatch #Probability Sharpening #Long-form Generation #KV Cache Management

2025년 10월 21일

[논문리뷰] Visual Autoregressive Models Beat Diffusion Models on Inference Time Scaling

Dim P. Papadopoulos이 arXiv에 게시한 'Visual Autoregressive Models Beat Diffusion Models on Inference Time Scaling' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Autoregressive Models #Diffusion Models #Inference Time Scaling #Beam Search #Image Generation #Text-to-Image Synthesis #Discrete Latent Space

2025년 10월 21일

[논문리뷰] Uniworld-V2: Reinforce Image Editing with Diffusion Negative-aware Finetuning and MLLM Implicit Feedback

arXiv에 게시된 'Uniworld-V2: Reinforce Image Editing with Diffusion Negative-aware Finetuning and MLLM Implicit Feedback' 논문에 대한 자세한 리뷰입니다.

#Review #Image Editing #Diffusion Models #Reinforcement Learning #MLLM #Policy Optimization #Finetuning #Reward Modeling #Human Alignment

2025년 10월 21일

[논문리뷰] UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action

arXiv에 게시된 'UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Action' 논문에 대한 자세한 리뷰입니다.

#Review #Computer Use Agents #Hybrid Action #Foundation Models #Reinforcement Learning #Supervised Fine-tuning #Synthetic Data Generation #Tool Learning #GUI Automation

2025년 10월 21일

[논문리뷰] Towards Mixed-Modal Retrieval for Universal Retrieval-Augmented Generation

arXiv에 게시된 'Towards Mixed-Modal Retrieval for Universal Retrieval-Augmented Generation' 논문에 대한 자세한 리뷰입니다.

#Review #Universal RAG #Multimodal Retrieval #Mixed-Modal Data Generation #Vision-Language Models #Contrastive Learning #Matryoshka Representation Learning

2025년 10월 21일

[논문리뷰] RL makes MLLMs see better than SFT

arXiv에 게시된 'RL makes MLLMs see better than SFT' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Language Models #Reinforcement Learning #Supervised Finetuning #Vision Encoder #Visual Representations #Direct Preference Optimization #Preference Alignment #PIVOT

2025년 10월 21일

[논문리뷰] QueST: Incentivizing LLMs to Generate Difficult Problems

arXiv에 게시된 'QueST: Incentivizing LLMs to Generate Difficult Problems' 논문에 대한 자세한 리뷰입니다.

#Review #LLM #Problem Generation #Competitive Programming #Synthetic Data #Difficulty Estimation #Rejection Fine-tuning #Graph Sampling

2025년 10월 21일

[논문리뷰] PICABench: How Far Are We from Physically Realistic Image Editing?

Kaiwen Zhu이 arXiv에 게시한 'PICABench: How Far Are We from Physically Realistic Image Editing?' 논문에 대한 자세한 리뷰입니다.

#Review #Image Editing #Physical Realism #Benchmark #VLM-as-a-Judge #Synthetic Data #Physics-Aware AI #Diffusion Models #Evaluation Metrics

2025년 10월 21일

[논문리뷰] On Non-interactive Evaluation of Animal Communication Translators

Adam Tauman Kalai이 arXiv에 게시한 'On Non-interactive Evaluation of Animal Communication Translators' 논문에 대한 자세한 리뷰입니다.

#Review #Machine Translation Quality Evaluation #Reference-Free Evaluation #Animal Communication #Language Models #Shuffle Test #Conlangs #Non-interactive Evaluation

2025년 10월 21일

[논문리뷰] MultiVerse: A Multi-Turn Conversation Benchmark for Evaluating Large Vision and Language Models

arXiv에 게시된 'MultiVerse: A Multi-Turn Conversation Benchmark for Evaluating Large Vision and Language Models' 논문에 대한 자세한 리뷰입니다.

#Review #Multi-Turn Conversation #VLM Evaluation #Benchmark #Vision and Language Models #Contextual Understanding #Checklist-based Evaluation #Interactive AI

2025년 10월 21일

[논문리뷰] Knowledge-based Visual Question Answer with Multimodal Processing, Retrieval and Filtering

arXiv에 게시된 'Knowledge-based Visual Question Answer with Multimodal Processing, Retrieval and Filtering' 논문에 대한 자세한 리뷰입니다.

#Review #Visual Question Answering #Retrieval-Augmented Generation #Multimodal AI #Reinforcement Learning #Knowledge Base #Tool Learning #Information Filtering

2025년 10월 21일

[논문리뷰] GuideFlow3D: Optimization-Guided Rectified Flow For Appearance Transfer

arXiv에 게시된 'GuideFlow3D: Optimization-Guided Rectified Flow For Appearance Transfer' 논문에 대한 자세한 리뷰입니다.

#Review #3D Appearance Transfer #Rectified Flow #Generative Models #Optimization-Guided Sampling #Neural Latent Representations #Training-Free #GPT-Based Evaluation

2025년 10월 21일

[논문리뷰] Glyph: Scaling Context Windows via Visual-Text Compression

Wenyi Hong이 arXiv에 게시한 'Glyph: Scaling Context Windows via Visual-Text Compression' 논문에 대한 자세한 리뷰입니다.

#Review #Long-Context Modeling #Visual Compression #Vision-Language Models #Token Efficiency #Genetic Algorithms #Multimodal AI #LLM Scaling

2025년 10월 21일

[논문리뷰] FineVision: Open Data Is All You Need

arXiv에 게시된 'FineVision: Open Data Is All You Need' 논문에 대한 자세한 리뷰입니다.

#Review #Multimodal Datasets #VLM #Data Curation #Data Hygiene #De-duplication #Human-in-the-loop #GUI Automation #Test-set Decontamination

2025년 10월 21일

[논문리뷰] Executable Knowledge Graphs for Replicating AI Research

arXiv에 게시된 'Executable Knowledge Graphs for Replicating AI Research' 논문에 대한 자세한 리뷰입니다.

#Review #AI Research Replication #Large Language Models (LLMs)#Knowledge Graphs (KGs)#Executable Code Generation #Retrieval-Augmented Generation (RAG)#PaperBench #Automated AI Research

2025년 10월 21일

[논문리뷰] Enterprise Deep Research: Steerable Multi-Agent Deep Research for Enterprise Analytics

arXiv에 게시된 'Enterprise Deep Research: Steerable Multi-Agent Deep Research for Enterprise Analytics' 논문에 대한 자세한 리뷰입니다.

#Review #Multi-Agent Systems #Deep Research #Enterprise AI #Human-in-the-Loop #Steerable AI #LLM Agents #Context Engineering #Enterprise Analytics

2025년 10월 21일

[논문리뷰] Embody 3D: A Large-scale Multimodal Motion and Behavior Dataset

arXiv에 게시된 'Embody 3D: A Large-scale Multimodal Motion and Behavior Dataset' 논문에 대한 자세한 리뷰입니다.

#Review #3D Motion Dataset #Multimodal Data #Human Behavior #Pose Tracking #Hand Tracking #Audio-Visual Data #Large-scale Dataset #SMPL-X

2025년 10월 21일

[논문리뷰] Distractor Injection Attacks on Large Reasoning Models: Characterization and Defense

arXiv에 게시된 'Distractor Injection Attacks on Large Reasoning Models: Characterization and Defense' 논문에 대한 자세한 리뷰입니다.

#Review #Large Reasoning Models (LRMs)#Prompt Injection #Adversarial Attack #Reasoning Distraction #Chain-of-Thought #Robustness #Supervised Fine-Tuning (SFT)#Reinforcement Learning (RL)

2025년 10월 21일

[논문리뷰] DeepAnalyze: Agentic Large Language Models for Autonomous Data Science

arXiv에 게시된 'DeepAnalyze: Agentic Large Language Models for Autonomous Data Science' 논문에 대한 자세한 리뷰입니다.

#Review #Autonomous Data Science #Agentic LLM #Curriculum Learning #Reinforcement Learning #Data Agents #End-to-end Data Science

2025년 10월 21일

[논문리뷰] Deep Self-Evolving Reasoning

arXiv에 게시된 'Deep Self-Evolving Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Deep Self-Evolving Reasoning #LLMs #Iterative Reasoning #Markov Chain #Self-Verification #Self-Refinement #Mathematical Reasoning #AIME Benchmark

2025년 10월 21일