#LLMs

116개의 포스트

[논문리뷰] QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization

Li Ding이 arXiv에 게시한 'QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization' 논문에 대한 자세한 리뷰입니다.

#Review #Program Repair #Over-editing #LLMs #Reinforcement Learning #EA-GRPO #Speculative Edits #Code Maintainability

2026년 4월 7일

[논문리뷰] Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Decoder Translation Models for Extensible Multilinguality

Yang Feng이 arXiv에 게시한 'Language on Demand, Knowledge at Core: Composing LLMs with Encoder-Decoder Translation Models for Extensible Multilinguality' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Multilinguality #Encoder-Decoder #Optimal Transport #Cross-Model Mapping #Language-on-Demand #NMT

2026년 3월 22일

[논문리뷰] Efficient Reasoning on the Edge

arXiv에 게시된 'Efficient Reasoning on the Edge' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Edge Computing #LoRA #Quantization #Budget Forcing #Switcher Module #Parallel Reasoning

2026년 3월 17일

[논문리뷰] Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs

arXiv에 게시된 'Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Reasoning #Parametric Knowledge #Factual Recall #Hallucination #Computational Buffer #Factual Priming #Chain-of-Thought

2026년 3월 10일

[논문리뷰] The Reasoning Trap -- Logical Reasoning as a Mechanistic Pathway to Situational Awareness

Divya Chaudhary이 arXiv에 게시한 'The Reasoning Trap -- Logical Reasoning as a Mechanistic Pathway to Situational Awareness' 논문에 대한 자세한 리뷰입니다.

#Review #Logical Reasoning #Situational Awareness #LLMs #Deceptive Alignment #AI Safety #RAISE Framework #Self-Modeling #Deduction #Induction #Abduction

2026년 3월 10일

[논문리뷰] Recursive Think-Answer Process for LLMs and VLMs

Yong Man Ro이 arXiv에 게시한 'Recursive Think-Answer Process for LLMs and VLMs' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #VLMs #Reasoning #Self-Correction #Reinforcement Learning #Confidence Estimation #Iterative Refinement #Think-Answer

2026년 3월 2일

[논문리뷰] LaSER: Internalizing Explicit Reasoning into Latent Space for Dense Retrieval

arXiv에 게시된 'LaSER: Internalizing Explicit Reasoning into Latent Space for Dense Retrieval' 논문에 대한 자세한 리뷰입니다.

#Review #Dense Retrieval #LLMs #Reasoning #Knowledge Distillation #Latent Space #Self-Distillation #Chain-of-Thought

2026년 3월 2일

[논문리뷰] Efficient RLVR Training via Weighted Mutual Information Data Selection

arXiv에 게시된 'Efficient RLVR Training via Weighted Mutual Information Data Selection' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #Data Selection #Mutual Information #Epistemic Uncertainty #LLMs #RLVR #Training Efficiency

2026년 3월 2일

[논문리뷰] Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking

arXiv에 게시된 'Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking' 논문에 대한 자세한 리뷰입니다.

#Review #Context Parallelism #Memory Efficiency #Headwise Chunking #Transformer Training #DeepSpeed Ulysses #LLMs #Activation Memory #Flash Attention

2026년 2월 24일

[논문리뷰] Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?

Ivan Oseledets이 arXiv에 게시한 'Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?' 논문에 대한 자세한 리뷰입니다.

#Review #Sparse Autoencoders #Interpretability #Neural Network Internals #Evaluation Baselines #Feature Decomposition #LLMs #Mechanistic Interpretability

2026년 2월 17일

[논문리뷰] Less is Enough: Synthesizing Diverse Data in Feature Space of LLMs

Ninghao Liu이 arXiv에 게시한 'Less is Enough: Synthesizing Diverse Data in Feature Space of LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Data Synthesis #LLMs #Feature Space #Sparse Autoencoders #Diversity Metrics #Post-Training #Instruction Tuning #Feature Activation Coverage

2026년 2월 15일

[논문리뷰] Adapting Vision-Language Models for E-commerce Understanding at Scale

arXiv에 게시된 'Adapting Vision-Language Models for E-commerce Understanding at Scale' 논문에 대한 자세한 리뷰입니다.

#Review #E-commerce #Vision-Language Models #Multimodal Understanding #Instruction Tuning #Attribute Extraction #Fine-tuning #Benchmarking #LLMs

2026년 2월 12일

[논문리뷰] FeatureBench: Benchmarking Agentic Coding for Complex Feature Development

Jiahe Wang이 arXiv에 게시한 'FeatureBench: Benchmarking Agentic Coding for Complex Feature Development' 논문에 대한 자세한 리뷰입니다.

#Review #Agentic Coding #Benchmarking #LLMs #Feature Development #Software Engineering #Test-Driven Development #Scalability

2026년 2월 11일

[논문리뷰] Blockwise Advantage Estimation for Multi-Objective RL with Verifiable Rewards

arXiv에 게시된 'Blockwise Advantage Estimation for Multi-Objective RL with Verifiable Rewards' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Credit Assignment #Multi-Objective Optimization #Advantage Estimation #Calibration #Structured Generation #Group Relative Policy Optimization

2026년 2월 11일

[논문리뷰] Self-Improving World Modelling with Latent Actions

Anna Korhonen이 arXiv에 게시한 'Self-Improving World Modelling with Latent Actions' 논문에 대한 자세한 리뷰입니다.

#Review #World Modeling #Latent Actions #Self-Improvement #Reinforcement Learning #LLMs #VLMs #Inverse Dynamics Model #Forward World Modelling

2026년 2월 8일

[논문리뷰] Canzona: A Unified, Asynchronous, and Load-Balanced Framework for Distributed Matrix-based Optimizers

arXiv에 게시된 'Canzona: A Unified, Asynchronous, and Load-Balanced Framework for Distributed Matrix-based Optimizers' 논문에 대한 자세한 리뷰입니다.

#Review #Distributed Training #Matrix-based Optimizers #Load Balancing #Asynchronous Compute #Data Parallelism #Tensor Parallelism #ZeRO-1 #LLMs

2026년 2월 8일

[논문리뷰] Semantic Search over 9 Million Mathematical Theorems

arXiv에 게시된 'Semantic Search over 9 Million Mathematical Theorems' 논문에 대한 자세한 리뷰입니다.

#Review #Semantic Search #Theorem Retrieval #LLMs #Dense Retrieval #Mathematical Information Retrieval #Vector Embeddings #Mathematical Dataset #RAG

2026년 2월 5일

[논문리뷰] Length-Unbiased Sequence Policy Optimization: Revealing and Controlling Response Length Variation in RLVR

Zhixiong Zeng이 arXiv에 게시한 'Length-Unbiased Sequence Policy Optimization: Revealing and Controlling Response Length Variation in RLVR' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning with Verifiable Rewards #LLMs #Policy Optimization #Response Length Bias #Sequence-level Clipping #Length-Unbiased Optimization #Multimodal Reasoning

2026년 2월 5일

[논문리뷰] Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations

arXiv에 게시된 'Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #Kernel Generation #Triton #GPU Optimization #LLMs #Reward Hacking #Multi-turn Interaction #Code Generation

2026년 2월 5일

[논문리뷰] FASA: Frequency-aware Sparse Attention

arXiv에 게시된 'FASA: Frequency-aware Sparse Attention' 논문에 대한 자세한 리뷰입니다.

#Review #Sparse Attention #KV Cache Optimization #Rotary Positional Embedding (RoPE)#Frequency Chunks (FCs)#LLMs #Long-Context #Training-Free

2026년 2월 4일

[논문리뷰] Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection

Jae-Joon Kim이 arXiv에 게시한 'Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection' 논문에 대한 자세한 리뷰입니다.

#Review #Sparse Attention #Long-Context Inference #LLMs #Token Selection #Efficiency #Transformer #Dynamic Sparsity

2026년 2월 3일

[논문리뷰] CoBA-RL: Capability-Oriented Budget Allocation for Reinforcement Learning in LLMs

arXiv에 게시된 'CoBA-RL: Capability-Oriented Budget Allocation for Reinforcement Learning in LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Budget Allocation #Adaptive Learning #Capability-Oriented Value Function #Exploration-Exploitation #Resource Efficiency

2026년 2월 3일

[논문리뷰] Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors

Bin Liang이 arXiv에 게시한 'Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors' 논문에 대한 자세한 리뷰입니다.

#Review #Tool Use #Execution Errors #Error Recovery #Reinforcement Learning #LLMs #Agentic AI #GRPO #FISSION

2026년 2월 1일

[논문리뷰] ReGuLaR: Variational Latent Reasoning Guided by Rendered Chain-of-Thought

Zhifeng Gao이 arXiv에 게시한 'ReGuLaR: Variational Latent Reasoning Guided by Rendered Chain-of-Thought' 논문에 대한 자세한 리뷰입니다.

#Review #Latent Reasoning #Chain-of-Thought #Variational Autoencoder #Visual-Text Compression #LLMs #Multi-modal Reasoning #Computational Efficiency

2026년 2월 1일

[논문리뷰] ConceptMoE: Adaptive Token-to-Concept Compression for Implicit Compute Allocation

arXiv에 게시된 'ConceptMoE: Adaptive Token-to-Concept Compression for Implicit Compute Allocation' 논문에 대한 자세한 리뷰입니다.

#Review #MoE #LLMs #Adaptive Compression #Token Merging #Compute Allocation #Efficiency #Vision-Language Models #Continual Training

2026년 1월 29일

[논문리뷰] Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation

arXiv에 게시된 'Harder Is Better: Boosting Mathematical Reasoning via Difficulty-Aware GRPO and Multi-Aspect Question Reformulation' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #Mathematical Reasoning #Difficulty-Aware Optimization #Data Augmentation #Policy Optimization #LLMs #GRPO #MQR

2026년 1월 28일

[논문리뷰] SAGE: Steerable Agentic Data Generation for Deep Search with Execution Feedback

arXiv에 게시된 'SAGE: Steerable Agentic Data Generation for Deep Search with Execution Feedback' 논문에 대한 자세한 리뷰입니다.

#Review #Deep Search #Agentic Data Generation #LLMs #Execution Feedback #Reinforcement Learning #Question Answering #Synthetic Data

2026년 1월 26일

[논문리뷰] Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs

arXiv에 게시된 'Can LLMs Clean Up Your Mess? A Survey of Application-Ready Data Preparation with LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Data Preparation #LLMs #Data Cleaning #Data Integration #Data Enrichment #AI Agents #Semantic Reasoning #Workflow Automation

2026년 1월 26일

[논문리뷰] Knowledge is Not Enough: Injecting RL Skills for Continual Adaptation

arXiv에 게시된 'Knowledge is Not Enough: Injecting RL Skills for Continual Adaptation' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Continual Adaptation #Reinforcement Learning #Supervised Fine-Tuning #Skill Transfer #Task Arithmetic #Tool Use

2026년 1월 25일

[논문리뷰] Aligning Agentic World Models via Knowledgeable Experience Learning

arXiv에 게시된 'Aligning Agentic World Models via Knowledgeable Experience Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Agentic AI #World Models #Experience Learning #LLMs #Physical Hallucinations #Embodied AI #Predictive Coding #Knowledge Repository

2026년 1월 20일

[논문리뷰] Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs

Lecheng Yan이 arXiv에 게시한 'Spurious Rewards Paradox: Mechanistically Understanding How RLVR Activates Memorization Shortcuts in LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #RLVR #LLMs #Mechanistic Interpretability #Memorization Shortcuts #Data Contamination #Anchor-Adapter Circuit #Path Patching #Logit Lens

2026년 1월 19일

[논문리뷰] MatchTIR: Fine-Grained Supervision for Tool-Integrated Reasoning via Bipartite Matching

arXiv에 게시된 'MatchTIR: Fine-Grained Supervision for Tool-Integrated Reasoning via Bipartite Matching' 논문에 대한 자세한 리뷰입니다.

#Review #Tool-Integrated Reasoning #LLMs #Reinforcement Learning #Fine-Grained Supervision #Bipartite Matching #Credit Assignment #Advantage Estimation

2026년 1월 15일

[논문리뷰] LSRIF: Logic-Structured Reinforcement Learning for Instruction Following

arXiv에 게시된 'LSRIF: Logic-Structured Reinforcement Learning for Instruction Following' 논문에 대한 자세한 리뷰입니다.

#Review #Instruction Following #Reinforcement Learning #Logical Structures #LLMs #Reward Modeling #Dataset Construction #Attention Mechanism

2026년 1월 15일

[논문리뷰] JudgeRLVR: Judge First, Generate Second for Efficient Reasoning

Sujian Li이 arXiv에 게시한 'JudgeRLVR: Judge First, Generate Second for Efficient Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #RLVR #LLMs #Reasoning #Judge-then-Generate #Quality-Efficiency #Discriminative Supervision #Mathematical Reasoning #Backtracking Reduction

2026년 1월 13일

[논문리뷰] Aligning Text, Code, and Vision: A Multi-Objective Reinforcement Learning Framework for Text-to-Visualization

arXiv에 게시된 'Aligning Text, Code, and Vision: A Multi-Objective Reinforcement Learning Framework for Text-to-Visualization' 논문에 대한 자세한 리뷰입니다.

#Review #Text-to-Visualization #Reinforcement Learning #Multi-Objective Optimization #GRPO #Multimodal Feedback #LLMs #Code Generation

2026년 1월 13일

[논문리뷰] Structured Episodic Event Memory

arXiv에 게시된 'Structured Episodic Event Memory' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #RAG #Episodic Memory #Graph Memory #Memory Architecture #Narrative Coherence #Long-term Reasoning #Event Frames

2026년 1월 12일

[논문리뷰] PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning

arXiv에 게시된 'PaCoRe: Learning to Scale Test-Time Compute with Parallel Coordinated Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #PaCoRe #Test-Time Compute Scaling #LLMs #Parallel Reasoning #Reinforcement Learning #Reasoning Synthesis #Message Passing #Mathematical Reasoning

2026년 1월 12일

[논문리뷰] GenCtrl -- A Formal Controllability Toolkit for Generative Models

arXiv에 게시된 'GenCtrl -- A Formal Controllability Toolkit for Generative Models' 논문에 대한 자세한 리뷰입니다.

#Review #Generative Models #Controllability #Reachability #Control Theory #Dialogue Systems #LLMs #T2IMs #PAC Bounds #Formal Verification

2026년 1월 11일

[논문리뷰] GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization

arXiv에 게시된 'GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimization' 논문에 대한 자세한 리뷰입니다.

#Review #Multi-reward RL #Policy Optimization #Reward Normalization #GRPO #GDPO #LLMs #Training Stability

2026년 1월 8일

[논문리뷰] Mindscape-Aware Retrieval Augmented Generation for Improved Long Context Understanding

arXiv에 게시된 'Mindscape-Aware Retrieval Augmented Generation for Improved Long Context Understanding' 논문에 대한 자세한 리뷰입니다.

#Review #Retrieval Augmented Generation #Long Context Understanding #Mindscape-Aware #Hierarchical Summarization #Context-Aware Embeddings #Integrative Reasoning #LLMs

2025년 12월 28일

[논문리뷰] Step-DeepResearch Technical Report

arXiv에 게시된 'Step-DeepResearch Technical Report' 논문에 대한 자세한 리뷰입니다.

#Review #Deep Research Agents #LLMs #Reinforcement Learning #Supervised Fine-tuning #Agentic AI #Multi-hop Reasoning #Benchmarking #Cost-effectiveness

2025년 12월 23일

[논문리뷰] Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows

Yuhao Zhou이 arXiv에 게시한 'Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows' 논문에 대한 자세한 리뷰입니다.

#Review #Scientific General Intelligence (SGI)#LLMs #Benchmarking #Scientist-Aligned Workflows #Practical Inquiry Model #Multi-modal Reasoning #Code Generation #Test-Time Reinforcement Learning (TTRL)

2025년 12월 21일

[논문리뷰] NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents

chongyang09이 arXiv에 게시한 'NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents' 논문에 대한 자세한 리뷰입니다.

#Review #Coding Agents #LLMs #Software Engineering #Repository Generation #Long-Horizon Reasoning #Benchmark #Python Development #Autonomous Systems

2025년 12월 15일

[논문리뷰] Memory in the Age of AI Agents

Yanwei Yue이 arXiv에 게시한 'Memory in the Age of AI Agents' 논문에 대한 자세한 리뷰입니다.

#Review #AI Agents #Memory Systems #LLMs #Taxonomy #Continual Learning #Self-Evolution #Multimodality #Reinforcement Learning

2025년 12월 15일

[논문리뷰] CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning

arXiv에 게시된 'CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning' 논문에 대한 자세한 리뷰입니다.

#Review #CUDA #Matrix Multiplication #Reinforcement Learning #LLMs #Kernel Optimization #HGEMM #GPU Performance #cuBLAS

2025년 12월 2일

[논문리뷰] The Art of Scaling Test-Time Compute for Large Language Models

Tanmoy Chakraborty이 arXiv에 게시한 'The Art of Scaling Test-Time Compute for Large Language Models' 논문에 대한 자세한 리뷰입니다.

#Review #Test-Time Scaling #LLMs #Reasoning #Compute Efficiency #Inference Optimization #Decoding Strategies #Model Behavior

2025년 12월 1일

[논문리뷰] SpeContext: Enabling Efficient Long-context Reasoning with Speculative Context Sparsity in LLMs

arXiv에 게시된 'SpeContext: Enabling Efficient Long-context Reasoning with Speculative Context Sparsity in LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Long-context Reasoning #KV Cache Optimization #Speculative Sparsity #Knowledge Distillation #Adaptive Memory Management #Throughput

2025년 12월 1일

[논문리뷰] Unveiling Intrinsic Dimension of Texts: from Academic Abstract to Creative Story

Kristian Kuznetsov이 arXiv에 게시한 'Unveiling Intrinsic Dimension of Texts: from Academic Abstract to Creative Story' 논문에 대한 자세한 리뷰입니다.

#Review #Intrinsic Dimension #LLMs #Text Complexity #Sparse Autoencoders #Text Semantics #Genre Analysis #Embedding Space #Text Generation

2025년 11월 23일

[논문리뷰] Agent READMEs: An Empirical Study of Context Files for Agentic Coding

Kundjanasith Thonglek이 arXiv에 게시한 'Agent READMEs: An Empirical Study of Context Files for Agentic Coding' 논문에 대한 자세한 리뷰입니다.

#Review #Agentic Coding #Context Files #READMEs for Agents #Empirical Study #Software Engineering #Documentation Maintenance #Non-functional Requirements #LLMs

2025년 11월 18일

[논문리뷰] ATLAS: A High-Difficulty, Multidisciplinary Benchmark for Frontier Scientific Reasoning

Yuqiang Li이 arXiv에 게시한 'ATLAS: A High-Difficulty, Multidisciplinary Benchmark for Frontier Scientific Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Benchmark #LLMs #Scientific Reasoning #Multidisciplinary #AI4S #Data Contamination #Evaluation #LRM-as-Judge

2025년 11월 18일

[논문리뷰] Assessing LLMs for Serendipity Discovery in Knowledge Graphs: A Case for Drug Repurposing

arXiv에 게시된 'Assessing LLMs for Serendipity Discovery in Knowledge Graphs: A Case for Drug Repurposing' 논문에 대한 자세한 리뷰입니다.

#Review #Serendipity Discovery #Knowledge Graphs #Drug Repurposing #LLMs #KGQA #RNS Metric #Biomedical AI

2025년 11월 17일

[논문리뷰] MarsRL: Advancing Multi-Agent Reasoning System via Reinforcement Learning with Agentic Pipeline Parallelism

arXiv에 게시된 'MarsRL: Advancing Multi-Agent Reasoning System via Reinforcement Learning with Agentic Pipeline Parallelism' 논문에 대한 자세한 리뷰입니다.

#Review #Multi-Agent Systems #Reinforcement Learning #LLMs #Pipeline Parallelism #Reasoning #Reward Shaping #Agentic AI

2025년 11월 16일

[논문리뷰] RLoop: An Self-Improving Framework for Reinforcement Learning with Iterative Policy Initialization

Wenhao Huang이 arXiv에 게시한 'RLoop: An Self-Improving Framework for Reinforcement Learning with Iterative Policy Initialization' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Generalization #Overfitting #Catastrophic Forgetting #Iterative Policy Optimization #Policy Diversity

2025년 11월 10일

[논문리뷰] Shorter but not Worse: Frugal Reasoning via Easy Samples as Length Regularizers in Math RLVR

arXiv에 게시된 'Shorter but not Worse: Frugal Reasoning via Easy Samples as Length Regularizers in Math RLVR' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #RLVR #Length Regularization #Mathematical Reasoning #Data Curation #Model Efficiency #Emergent Brevity

2025년 11월 9일

[논문리뷰] Multi-Step Knowledge Interaction Analysis via Rank-2 Subspace Disentanglement

Isabelle Augenstein이 arXiv에 게시한 'Multi-Step Knowledge Interaction Analysis via Rank-2 Subspace Disentanglement' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Knowledge Interaction #Parametric Knowledge #Contextual Knowledge #Subspace Disentanglement #NLE Generation #Hallucination Detection #Chain-of-Thought

2025년 11월 9일

[논문리뷰] Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph

arXiv에 게시된 'Generalizing Test-time Compute-optimal Scaling as an Optimizable Graph' 논문에 대한 자세한 리뷰입니다.

#Review #Test-Time Scaling #LLMs #Graph Optimization #REINFORCE #Multi-agent Systems #Adaptive Architectures #Compute-optimal Scaling #Probabilistic Graphs

2025년 11월 9일

[논문리뷰] SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens

arXiv에 게시된 'SemCoT: Accelerating Chain-of-Thought Reasoning through Semantically-Aligned Implicit Tokens' 논문에 대한 자세한 리뷰입니다.

#Review #Chain-of-Thought (CoT)#Implicit Reasoning #LLMs #Semantic Alignment #Efficiency Optimization #Knowledge Distillation

2025년 11월 9일

[논문리뷰] Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning

arXiv에 게시된 'Supervised Reinforcement Learning: From Expert Trajectories to Step-wise Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Supervised Reinforcement Learning #LLMs #Multi-step Reasoning #Reward Shaping #Expert Trajectories #Math Reasoning #Agentic AI

2025년 10월 31일

[논문리뷰] A Survey of Data Agents: Emerging Paradigm or Overstated Hype?

Boyan Li이 arXiv에 게시한 'A Survey of Data Agents: Emerging Paradigm or Overstated Hype?' 논문에 대한 자세한 리뷰입니다.

#Review #Data Agents #LLMs #Autonomy Levels #Hierarchical Taxonomy #SAE J3016 #Data Management #Data Preparation #Data Analysis #Autonomous Orchestration

2025년 10월 28일

[논문리뷰] Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers

arXiv에 게시된 'Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers' 논문에 대한 자세한 리뷰입니다.

#Review #MoE #Reinforcement Learning #Training Stability #Routing #Policy Alignment #Rollout Routing Replay #LLMs

2025년 10월 27일

[논문리뷰] Reasoning with Sampling: Your Base Model is Smarter Than You Think

arXiv에 게시된 'Reasoning with Sampling: Your Base Model is Smarter Than You Think' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #MCMC #Sampling #Reasoning #Distribution Sharpening #Reinforcement Learning (RL)#Inference-time Optimization #Training-free

2025년 10월 27일

[논문리뷰] RECALL: REpresentation-aligned Catastrophic-forgetting ALLeviation via Hierarchical Model Merging

arXiv에 게시된 'RECALL: REpresentation-aligned Catastrophic-forgetting ALLeviation via Hierarchical Model Merging' 논문에 대한 자세한 리뷰입니다.

#Review #Catastrophic Forgetting #Continual Learning #Model Merging #LLMs #Representation Learning #Data-free Learning #Hierarchical Parameter Fusion

2025년 10월 27일

[논문리뷰] Diff-XYZ: A Benchmark for Evaluating Diff Understanding

arXiv에 게시된 'Diff-XYZ: A Benchmark for Evaluating Diff Understanding' 논문에 대한 자세한 리뷰입니다.

#Review #Diff Understanding #Code Diff #Benchmark #LLMs #Code Editing #Software Engineering #Unified Diff Format #Search-Replace

2025년 10월 24일

[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

[논문리뷰] InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training

Congkai Xie이 arXiv에 게시한 'InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Reinforcement Learning #Rubric-Based Training #Medical Dialogue #Open-Ended Tasks #HealthBench #RAG

2025년 10월 20일

[논문리뷰] Large Language Models Do NOT Really Know What They Don't Know

arXiv에 게시된 'Large Language Models Do NOT Really Know What They Don't Know' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Hallucination Detection #Mechanistic Interpretability #Internal States #Knowledge Recall #Refusal Tuning #Factual Associations #Associated Hallucinations

2025년 10월 17일

[논문리뷰] The Art of Scaling Reinforcement Learning Compute for LLMs

arXiv에 게시된 'The Art of Scaling Reinforcement Learning Compute for LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Scaling Laws #Compute Efficiency #Predictability #Sigmoidal Curves #ScaleRL #Off-Policy RL

2025년 10월 16일

[논문리뷰] NOSA: Native and Offloadable Sparse Attention

Zhiyuan Liu이 arXiv에 게시한 'NOSA: Native and Offloadable Sparse Attention' 논문에 대한 자세한 리뷰입니다.

#Review #Sparse Attention #KV Cache Offloading #LLMs #Decoding Throughput #Locality Constraint #Memory Optimization #Trainable Sparse Attention

2025년 10월 16일

[논문리뷰] Dr.LLM: Dynamic Layer Routing in LLMs

arXiv에 게시된 'Dr.LLM: Dynamic Layer Routing in LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Dynamic Routing #LLMs #Adaptive Depth #Computational Efficiency #Monte Carlo Tree Search (MCTS)#Retrofittable Framework #Supervised Learning #Accuracy Improvement

2025년 10월 15일

[논문리뷰] ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall

Jiaqi Tang이 arXiv에 게시한 'ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall' 논문에 대한 자세한 리뷰입니다.

#Review #Knowledge Editing #LLMs #Multi-hop Reasoning #Mechanistic Interpretability #Neuron-level Attribution #Factual Recall #Transformer Networks

2025년 10월 13일

[논문리뷰] Training Dynamics Impact Post-Training Quantization Robustness

Jonas Geiping이 arXiv에 게시한 'Training Dynamics Impact Post-Training Quantization Robustness' 논문에 대한 자세한 리뷰입니다.

#Review #Post-Training Quantization #Quantization Robustness #Training Dynamics #Learning Rate Schedules #Weight Averaging #Large Language Models #LLMs #Hyperparameter Tuning

2025년 10월 8일

[논문리뷰] Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning

Zhuoshi Pan이 arXiv에 게시한 'Scaling Code-Assisted Chain-of-Thoughts and Instructions for Model Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Code-Assisted Reasoning #Chain-of-Thought (CoT)#Instruction Tuning #Data Augmentation #LLMs #Mathematical Reasoning #Self-Verification #Code Generation

2025년 10월 8일

[논문리뷰] DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning

Zheli Liu이 arXiv에 게시한 'DRIFT: Learning from Abundant User Dissatisfaction in Real-World Preference Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Preference Learning #LLMs #User Feedback #Dissatisfaction Signals #DPO #Iterative Training #RLHF #Exploration

2025년 10월 8일

[논문리뷰] REPAIR: Robust Editing via Progressive Adaptive Intervention and Reintegration

arXiv에 게시된 'REPAIR: Robust Editing via Progressive Adaptive Intervention and Reintegration' 논문에 대한 자세한 리뷰입니다.

#Review #Model Editing #Lifelong Learning #LLMs #Continual Learning #Knowledge Distillation #Error Feedback #Memory Management #Parameter Merging

2025년 10월 6일

[논문리뷰] Free Lunch Alignment of Text-to-Image Diffusion Models without Preference Image Pairs

arXiv에 게시된 'Free Lunch Alignment of Text-to-Image Diffusion Models without Preference Image Pairs' 논문에 대한 자세한 리뷰입니다.

#Review #Text-to-Image Models #Diffusion Models #Preference Optimization #LLMs #RLHF #Prompt Editing #Free Lunch Alignment #TDPO #TKTO

2025년 10월 6일

[논문리뷰] PIPer: On-Device Environment Setup via Online Reinforcement Learning

arXiv에 게시된 'PIPer: On-Device Environment Setup via Online Reinforcement Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Environment Setup #LLMs #Reinforcement Learning #Supervised Fine-tuning #On-device AI #Software Engineering #Verifiable Rewards

2025년 10월 2일

[논문리뷰] CurES: From Gradient Analysis to Efficient Curriculum Learning for Reasoning LLMs

Hengyi Cai이 arXiv에 게시한 'CurES: From Gradient Analysis to Efficient Curriculum Learning for Reasoning LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Curriculum Learning #LLMs #Reasoning #Gradient Optimization #Reinforcement Learning #Bayesian Inference #Sample Efficiency

2025년 10월 2일

[논문리뷰] BroRL: Scaling Reinforcement Learning via Broadened Exploration

arXiv에 게시된 'BroRL: Scaling Reinforcement Learning via Broadened Exploration' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Scaling Laws #Exploration #Rollout Size #Verifiable Rewards #PPO #Mass Balance Equation

2025년 10월 2일

[논문리뷰] SPARK: Synergistic Policy And Reward Co-Evolving Framework

arXiv에 게시된 'SPARK: Synergistic Policy And Reward Co-Evolving Framework' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #LVLMs #Reward Modeling #Policy Optimization #Self-Reflection #Verifiable Rewards #Co-evolution

2025년 9월 29일

[논문리뷰] Instruction-Following Evaluation in Function Calling for Large Language Models

NikolaiSkripko이 arXiv에 게시한 'Instruction-Following Evaluation in Function Calling for Large Language Models' 논문에 대한 자세한 리뷰입니다.

#Review #Function Calling #LLMs #Instruction Following #Benchmarking #JSON Schema #AI Agents #Evaluation Metrics

2025년 9월 29일

[논문리뷰] SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent

Siyuan Huang이 arXiv에 게시한 'SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent' 논문에 대한 자세한 리뷰입니다.

#Review #3D Scene Synthesis #Agentic Framework #LLMs #Self-Reflection #Tool-Use #Physical Plausibility #Iterative Refinement #Embodied AI

2025년 9월 26일

[논문리뷰] SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?

Yannis Yiming He이 arXiv에 게시한 'SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?' 논문에 대한 자세한 리뷰입니다.

#Review #AI Agents #Software Engineering #LLMs #Code Generation #Benchmark #Contamination Resistance #Long-Horizon Tasks #Enterprise Software

2025년 9월 23일

[논문리뷰] From Uniform to Heterogeneous: Tailoring Policy Optimization to Every Token's Nature

Bin Cui이 arXiv에 게시한 'From Uniform to Heterogeneous: Tailoring Policy Optimization to Every Token's Nature' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Policy Optimization #Token Heterogeneity #Adaptive Sampling #Advantage Redistribution #Asymmetric Clipping #Entropy-based RL

2025년 9월 23일

[논문리뷰] EpiCache: Episodic KV Cache Management for Long Conversational Question Answering

Minsik Cho이 arXiv에 게시한 'EpiCache: Episodic KV Cache Management for Long Conversational Question Answering' 논문에 대한 자세한 리뷰입니다.

#Review #KV Cache Management #Long Conversational QA #LLMs #Memory Efficiency #Episodic Clustering #Block Prefill Eviction #Sensitivity-aware Allocation

2025년 9월 23일

[논문리뷰] CodeFuse-CR-Bench: A Comprehensiveness-aware Benchmark for End-to-End Code Review Evaluation in Python Projects

Hang Yu이 arXiv에 게시한 'CodeFuse-CR-Bench: A Comprehensiveness-aware Benchmark for End-to-End Code Review Evaluation in Python Projects' 논문에 대한 자세한 리뷰입니다.

#Review #Code Review #LLMs #Benchmark #Python Projects #End-to-End Evaluation #Context-Awareness #Software Engineering #LLM-as-a-Judge

2025년 9월 23일

[논문리뷰] RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation

Steven Liu이 arXiv에 게시한 'RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation' 논문에 대한 자세한 리뷰입니다.

#Review #Code Generation #LLMs #Repository Planning #Graph-based Representation #Software Engineering #Agent Frameworks #Scalable Codebase

2025년 9월 22일

[논문리뷰] Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Delibration

Zhilin Wang이 arXiv에 게시한 'Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Delibration' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Specification Alignment #Test-Time Deliberation #Safety-Behavior Trade-off #ALIGN3 #SPECBENCH #Prompt Engineering

2025년 9월 19일

[논문리뷰] Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation

Kishan Panaganti이 arXiv에 게시한 'Evolving Language Models without Labels: Majority Drives Selection, Novelty Promotes Variation' 논문에 대한 자세한 리뷰입니다.

#Review #Label-free Reinforcement Learning #LLMs #Self-improvement #Entropy Collapse #Novelty Reward #Test-Time RL #GRPO #Evolutionary Computing Principles

2025년 9월 19일

[논문리뷰] WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning

Huifeng Yin이 arXiv에 게시한 'WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Web Agents #Reinforcement Learning #Synthetic Data #Knowledge Graphs #LLMs #Supervised Fine-Tuning #Sim-to-Real Transfer #Agentic AI

2025년 9월 17일

[논문리뷰] GAPrune: Gradient-Alignment Pruning for Domain-Aware Embeddings

Yixuan Tang이 arXiv에 게시한 'GAPrune: Gradient-Alignment Pruning for Domain-Aware Embeddings' 논문에 대한 자세한 리뷰입니다.

#Review #Model Pruning #Domain Adaptation #Embedding Models #Gradient Alignment #Fisher Information #Model Compression #LLMs

2025년 9월 16일

[논문리뷰] <think> So let's replace this phrase with insult... </think> Lessons learned from generation of toxic texts with LLMs

Alexander Panchenko이 arXiv에 게시한 '<think> So let's replace this phrase with insult... </think> Lessons learned from generation of toxic texts with LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #Toxic Text Generation #LLMs #Text Detoxification #Lexical Diversity #Synthetic Data #Human Annotation #Style Transfer

2025년 9월 11일

[논문리뷰] A Survey of Reinforcement Learning for Large Reasoning Models

Runze Liu이 arXiv에 게시한 'A Survey of Reinforcement Learning for Large Reasoning Models' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #Large Reasoning Models #LLMs #Reward Design #Policy Optimization #Verifiable Rewards #Agentic AI #Multimodal AI

2025년 9월 11일

[논문리뷰] ΔL Normalization: Rethink Loss Aggregation in RLVR

Lili Qiu이 arXiv에 게시한 'ΔL Normalization: Rethink Loss Aggregation in RLVR' 논문에 대한 자세한 리뷰입니다.

#Review #Reinforcement Learning #LLMs #Gradient Variance #Loss Aggregation #Unbiased Estimator #RLVR #Policy Gradient #Normalization

2025년 9월 10일

[논문리뷰] Reverse-Engineered Reasoning for Open-Ended Generation

Wangchunshu Zhou이 arXiv에 게시한 'Reverse-Engineered Reasoning for Open-Ended Generation' 논문에 대한 자세한 리뷰입니다.

#Review #Deep Reasoning #Open-Ended Generation #Reverse-Engineered Reasoning (REER)#LLMs #Synthetic Data #Iterative Refinement #Perplexity Minimization #DeepWriting-20K

2025년 9월 9일

[논문리뷰] Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions?

Yu Fu이 arXiv에 게시한 'Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions?' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Instruction Following #Benchmark #Cognitive Inertia #Out-of-Distribution #Supervised Fine-Tuning #Evaluation #Robustness

2025년 9월 5일

[논문리뷰] The Gold Medals in an Empty Room: Diagnosing Metalinguistic Reasoning in LLMs with Camlang

Solomon Tsai이 arXiv에 게시한 'The Gold Medals in an Empty Room: Diagnosing Metalinguistic Reasoning in LLMs with Camlang' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Metalinguistic Reasoning #Constructed Language #Camlang #Second Language Acquisition #Zero-shot Learning #Natural Language Understanding #Commonsense Reasoning

2025년 9월 3일

[논문리뷰] Reasoning Vectors: Transferring Chain-of-Thought Capabilities via Task Arithmetic

Bernard Ghanem이 arXiv에 게시한 'Reasoning Vectors: Transferring Chain-of-Thought Capabilities via Task Arithmetic' 논문에 대한 자세한 리뷰입니다.

#Review #Reasoning Vectors #Task Arithmetic #Chain-of-Thought #LLMs #Reinforcement Learning #Model Merging #Parameter Transfer

2025년 9월 3일

[논문리뷰] Metis: Training Large Language Models with Advanced Low-Bit Quantization

Hengjie Cao이 arXiv에 게시한 'Metis: Training Large Language Models with Advanced Low-Bit Quantization' 논문에 대한 자세한 리뷰입니다.

#Review #Low-Bit Quantization #LLMs #Spectral Decomposition #Anisotropy #Adaptive Learning Rate #Regularization #FP8 Training #FP4 Training

2025년 9월 3일

[논문리뷰] FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games

Dongmin Park이 arXiv에 게시한 'FlashAdventure: A Benchmark for GUI Agents Solving Full Story Arcs in Diverse Adventure Games' 논문에 대한 자세한 리뷰입니다.

#Review #GUI Agents #Adventure Games #Benchmark #Full Story Arc #Observation-Behavior Gap #LLMs #Automated Evaluation

2025년 9월 3일

[논문리뷰] UltraMemV2: Memory Networks Scaling to 120B Parameters with Superior Long-Context Learning

Ran Guo이 arXiv에 게시한 'UltraMemV2: Memory Networks Scaling to 120B Parameters with Superior Long-Context Learning' 논문에 대한 자세한 리뷰입니다.

#Review #Memory Networks #Mixture of Experts (MoE)#Long-Context Learning #Sparse Models #Transformer Architecture #LLMs #Efficient Inference

2025년 8월 27일

[논문리뷰] ThinkDial: An Open Recipe for Controlling Reasoning Effort in Large Language Models

Jiangjie Chen이 arXiv에 게시한 'ThinkDial: An Open Recipe for Controlling Reasoning Effort in Large Language Models' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Controllable Reasoning #Computational Efficiency #Reinforcement Learning #Supervised Fine-tuning #Reasoning Compression #Budget-Aware Training

2025년 8월 27일

[논문리뷰] ST-Raptor: LLM-Powered Semi-Structured Table Question Answering

Wei Zhou이 arXiv에 게시한 'ST-Raptor: LLM-Powered Semi-Structured Table Question Answering' 논문에 대한 자세한 리뷰입니다.

#Review #Semi-structured Tables #Question Answering #LLMs #Hierarchical Orthogonal Tree #Table Layout Understanding #Pipeline Generation #Verification Mechanism

2025년 8월 26일

[논문리뷰] CRISP: Persistent Concept Unlearning via Sparse Autoencoders

Yonatan Belinkov이 arXiv에 게시한 'CRISP: Persistent Concept Unlearning via Sparse Autoencoders' 논문에 대한 자세한 리뷰입니다.

#Review #Concept Unlearning #Sparse Autoencoders (SAEs)#LLMs #Parameter-Efficient Fine-Tuning #Model Interpretability #Safety-Critical AI #Feature Suppression #WMDP Benchmark

2025년 8월 25일

[논문리뷰] aiXiv: A Next-Generation Open Access Ecosystem for Scientific Discovery Generated by AI Scientists

Heng Zhang이 arXiv에 게시한 'aiXiv: A Next-Generation Open Access Ecosystem for Scientific Discovery Generated by AI Scientists' 논문에 대한 자세한 리뷰입니다.

#Review #AI Agents #Open Access #Scientific Discovery #Peer Review #LLMs #Multi-agent Systems #Prompt Injection #Iterative Refinement

2025년 8월 22일

[논문리뷰] Mind the Generation Process: Fine-Grained Confidence Estimation During LLM Generation

Xinyi Wang이 arXiv에 게시한 'Mind the Generation Process: Fine-Grained Confidence Estimation During LLM Generation' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Confidence Estimation #Fine-Grained #Generation Process #Calibration #Monte Carlo Sampling #Backward Confidence Integration

2025년 8월 20일

[논문리뷰] Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL

Liam-Liu이 arXiv에 게시한 'Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL' 논문에 대한 자세한 리뷰입니다.

#Review #Chain-of-Agents #Agent Foundation Models #Multi-Agent Systems #Tool-Integrated Reasoning #Multi-agent Distillation #Agentic Reinforcement Learning #LLMs #End-to-End Learning

2025년 8월 20일

[논문리뷰] BiasGym: Fantastic Biases and How to Find (and Remove) Them

Arnav Arora이 arXiv에 게시한 'BiasGym: Fantastic Biases and How to Find (and Remove) Them' 논문에 대한 자세한 리뷰입니다.

#Review #Bias Mitigation #LLMs #Mechanistic Interpretability #Fine-tuning #Attention Steering #Stereotype Analysis #Safety Alignment

2025년 8월 13일

[논문리뷰] Less Is More: Training-Free Sparse Attention with Global Locality for Efficient Reasoning

Baihong Yuan이 arXiv에 게시한 'Less Is More: Training-Free Sparse Attention with Global Locality for Efficient Reasoning' 논문에 대한 자세한 리뷰입니다.

#Review #Sparse Attention #LLMs #Reasoning Tasks #Efficiency #Training-Free #Global Locality #KV Cache Optimization

2025년 8월 12일

[논문리뷰] Grove MoE: Towards Efficient and Superior MoE LLMs with Adjugate Experts

Tieyuan Chen이 arXiv에 게시한 'Grove MoE: Towards Efficient and Superior MoE LLMs with Adjugate Experts' 논문에 대한 자세한 리뷰입니다.

#Review #Mixture of Experts #LLMs #MoE Architecture #Dynamic Activation #Adjugate Experts #Upcycling Strategy #Load Balancing

2025년 8월 12일

[논문리뷰] Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs

Robert Kirk이 arXiv에 게시한 'Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #데이터 필터링 #사전 학습 #변조 저항성 #바이오위협 #AI 안전 #서킷 브레이킹 #머신 언러닝

2025년 8월 12일

[논문리뷰] BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent

Kai Zou이 arXiv에 게시한 'BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent' 논문에 대한 자세한 리뷰입니다.

#Review #Benchmarking #Deep-Research Agents #LLMs #Retrieval #Curated Corpus #Evaluation #Fairness #Transparency #Reproducibility

2025년 8월 12일

[논문리뷰] Pruning the Unsurprising: Efficient Code Reasoning via First-Token Surprisal

Chengcheng Wan이 arXiv에 게시한 'Pruning the Unsurprising: Efficient Code Reasoning via First-Token Surprisal' 논문에 대한 자세한 리뷰입니다.

#Review #Code Reasoning #CoT Compression #LLMs #Efficiency #Surprisal #Pruning #Fine-tuning #Large Reasoning Models

2025년 8월 11일

[논문리뷰] MeshLLM: Empowering Large Language Models to Progressively Understand and Generate 3D Mesh

Yi Yang이 arXiv에 게시한 'MeshLLM: Empowering Large Language Models to Progressively Understand and Generate 3D Mesh' 논문에 대한 자세한 리뷰입니다.

#Review #3D Mesh Generation #LLMs #Mesh Understanding #Text-to-3D #Primitive-Mesh Decomposition #Progressive Training #Multimodal AI

2025년 8월 11일

[논문리뷰] Light-IF: Endowing LLMs with Generalizable Reasoning via Preview and Self-Checking for Complex Instruction Following

Liang Xu이 arXiv에 게시한 'Light-IF: Endowing LLMs with Generalizable Reasoning via Preview and Self-Checking for Complex Instruction Following' 논문에 대한 자세한 리뷰입니다.

#Review #LLMs #Instruction Following #Reasoning #Reinforcement Learning #Supervised Fine-tuning #Entropy Regularization #Self-Checking #Previewing

2025년 8월 7일

[논문리뷰] Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens

Zhen Tan이 arXiv에 게시한 'Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens' 논문에 대한 자세한 리뷰입니다.

#Review #Chain-of-Thought #LLMs #OOD Generalization #Data Distribution Shift #Reasoning #Pattern Matching #DataAlchemy

2025년 8월 7일

[논문리뷰] IFDECORATOR: Wrapping Instruction Following Reinforcement Learning with Verifiable Rewards

Ling-I Wu이 arXiv에 게시한 'IFDECORATOR: Wrapping Instruction Following Reinforcement Learning with Verifiable Rewards' 논문에 대한 자세한 리뷰입니다.

#Review #Instruction Following #Reinforcement Learning #Reward Hacking #LLMs #Curriculum Learning #Data Flywheel #Verifiable Rewards

2025년 8월 7일