LLM Self-Refinement

LLM: Large Language Model

Follow this topic Ask about this topic

Momentum

6 papers in the last four weeks, up 50% on the four weeks before. 0.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 62

All topics

The Winner's Curse in LLM Self-Improvement Loops: Selection Noise, Lock-in, and Acceptance Rules
Oct 7, 2026Litao Hu, Yutong TangLLM Self-Refinement Language Model Self-Improvement
Meta · Microsoft
RefCon: Iterative Refinement and Contrastive Memory Extraction for Context-Evolving Agent
Sep 30, 2026Ubaidillah Ariq Prathama, Bo Liu, Yeo Boon Hong +3Contrastive Learning LLM Self-Refinement
Huawei Technologies, Co., Ltd.
AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks
Sep 29, 2026Hongjin Qian, Chaofan Li, Kun Luo +11LLM Self-Refinement LLM Agent Training
Beijing Academy of Artificial Intelligence (BAAI)
WeaveData: A Multimodal Data Analysis System with Self-Critiquing and Self-Evolving LLM Plans
Sep 28, 2026Min Jia, Shihao Zhou, Jun-Peng Zhu +9LLM Self-Refinement LLM Planning
Northwest A&F University · PingCAP · East China Normal University +1
DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement
Sep 18, 2026Siyuan Liu, Fan Yu, Dongyu Ru +5LLM Self-Refinement AI Agent Evaluation
Fudan University · Meituan Longcat Team
What Drives Recovery in Agentic Text-to-Cypher? LAST-CQ: An LLM Agent Self-Refinement Framework
Sep 14, 2026Ioannis Prokopiou, Athanasios Aidinis, Panagiotis-Christos Kyrmpatsos +1LLM Self-Refinement LLM Agent Evaluation
Athens University of Economics and Business · Orfium
Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents
Sep 11, 2026Toshiaki Koike-Akino, Vladislav Blaykhman, Ye Wang +2LLM Self-Refinement LLM-as-a-Judge
Mitsubishi Electric Research Laboratories (MERL) 202 Broadway, Cambrdige, MA 02139
ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide Generation
Aug 31, 2026Muzhao Tian, Zezi Zeng, Yifan Yang +14LLM Self-Refinement Generative AI Evaluation
Fudan University · Work done during an internship at Microsoft. · Xi’an Jiaotong University +2
Co-Evolving Actor-Conditioned Critics for Non-Verifiable Generation
Aug 31, 2026Jinyoung Kim, Muhammad Khalifa, Lajanugen Logeswaran +4RL for Language Models LLM Evaluation
University of Michigan · LG AI Research · University of Illinois at Chicago
Optimal Stopping of Self-Refining Foundation Models
Aug 11, 2026Kim Hammar, Tansu Alpcan, Emil C. LupuLLM Self-Refinement Optimal Stopping
Imperial College London, United Kingdom · University of Melbourne, Australia
SafeQL: Search-based Refinement for Safe and Efficient LLM-based Text-to-SQL
Aug 10, 2026Geonho Lee, Min-Soo KimLLM Self-Refinement Text-to-SQL
Korea Advanced Institute of Science and Technology Republic of Korea
Progressive Content Refinement with Decaying Reward Joint LinUCB
Aug 7, 2026Shion Ishikawa, Pablo Loyola, Young-joo Chung +1LLM Self-Refinement Large Language Model-Guided Optimization
Rakuten Group, Inc. Tokyo, Japan · Rakuten Group, Inc. San Mateo, USA
Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning
Aug 6, 2026Ahsan Bilal, Muhammad Ahmed Mohsin, Muhammad Umer +4Numerical Reasoning in Language Models LLM Self-Correction
University of Oklahoma · Stanford University · Universitat Pompeu Fabra +1
RefactorAssist: Agentic Refinement for Reliable Code Refactoring
Aug 2, 2026Jonathan Cordeiro, Shayan Noei, Ying ZouLLM Self-Refinement Automated Program Repair
Queen’s University, Kingston, Canada
When Does LLM Orchestration Pay Off? A Controlled Evaluation of Accuracy, Cost, and Task Difficulty
Aug 1, 2026Nicolas Leins, Nico Pelleriti, Jana Gonnermann-Müller +1LLM Evaluation LLM Self-Refinement
Zuse Institute Berlin & TU Berlin, Berlin, Germany · Zuse Institute Berlin & Weizenbaum Institute Berlin, Berlin, Germany
Reflection or Re-Generation? Why LLM Revision Fails Where Human Revision Succeeds
Jul 31, 2026Yefan Tao, Gerald Friedland, Madhusudhanan Chandrasekaran +1LLM Self-Correction LLM Self-Refinement
Amazon
Sample More, Reflect Less: Self-Refine and Reflexion Lose to Repeated Sampling at Equal Token Cost, from 1.5B to 7B
Jul 30, 2026Iliya MirzaeiLLM Self-Refinement Language Model Generation Evaluation
Department of Computer Science Stony Brook University
Two Calls Beat Five Agents: Evaluating Multi-Agent Pipelines Against Self-Refinement for Local Language Models
Jul 29, 2026Ashish Prajapati, Om MohiteMulti-Agent LLM Systems LLM Self-Refinement
Independent Researcher Mumbai, India
When Do Agent Loops Mistake Stagnation for Progress? Self-Evaluation Bias and Externally Grounded Verification in Long-Running Autonomous LLM Agent Loops
Jul 27, 2026Hyundoo Park, Byungho ChoiAI Agent Reliability LLM Self-Refinement
AsyncSite
Failures Reveal What Metrics Miss: An Evidence-Driven Agent for Recursive Refinement of ECG Classifiers
Jul 27, 2026Jinliang Deng, Yiming Niu, Yibo Pan +3LLM Self-Refinement Electrocardiogram Classification
School of Computer Science and Engineering, Beihang University, Beijing, China · School of Artificial Intelligence, Beihang University, Beijing, China · School of Economics and Business Administration, Chongqing University, Chongqing, China +1
Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement Learning
Jul 24, 2026Shujin Wu, Cheng Qian, Xiusi Chen +1RL for Code Generation LLM Self-Refinement
University of Illinois Urbana-Champaign
Think Through a Bottleneck: Hourglass Reasoning for Rigorous Induction
Jul 13, 2026Huan ZhuLLM Self-Refinement LLM-Based Program Synthesis
Peking University
Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops
Jul 8, 2026Mingguang Chen, Licheng Wang, Bo QuLLM Self-Refinement Human-in-the-Loop Evaluation
University of California, Riverside (UCR) · AlphaAvatar · Illinois Institute of Technology (IIT)
Contrastive Reflection for Iterative Prompt Optimization
Jun 29, 2026Derek Koh, Jinghui Mo, Benjamin H. Le +7LLM Self-Refinement LLM Agent Workflow Optimization
LinkedIn Sunnyvale, California, USA
What Drives Interactive Improvement from Feedback?
Jun 29, 2026Bartłomiej Cupiał, Jan Łojek, Mikołaj Garstecki +3LLM Self-Refinement LLM Agent Evaluation
University of Warsaw · AKCES NCBR · Princeton University +2
Towards Reliable and Robust LLM Planning: Symbolic Feedback-Driven Iterative Self-Refinement Framework
Jun 26, 2026Jiajing Zhang, Jiamei Jiang, Chenyang Zhang +3LLM Self-Refinement LLM Planning
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing, China · School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China · School of Industry-education Integration, University of Chinese Academy of Sciences, Beijing, China
Perception, Verdict, and Evolution: Hindsight-Driven Self-Refining Forensics Agent for AI-Generated Image Detection
Jun 25, 2026Yangjun Wu, Keyu Yan, Yu Liu +5LLM Self-Refinement AI-Generated Image Detection
Zhejiang University, Zhejiang, China · Alibaba Group
When Does Intrinsic Self-Correction Help? A Task-Sensitive Analysis
Jun 22, 2026Elroy Stav, Dvir Berlowitz, Maayan Orner +1LLM Self-Correction LLM Evaluation
Department of Computer Science and Artificial Intelligence, Bar-Ilan University · Independent Researcher
Denoising Iterative Self-Correction: Structured Verification Loops for Reliable LLM Reasoning
Jun 19, 2026Shen Yin, David Ken, Joel StremmelLLM Self-Correction LLM Self-Refinement
Thomson Reuters Labs
ScholarSum: Student-Teacher Abstractive Summarization via Knowledge Graph Reasoning and Reflective Refinement
Jun 17, 2026Bohou Zhang, Xiaoyu Tao, Mingyue Cheng +2LLM Self-Refinement Graph Retrieval-Augmented Generation
State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China

Oct 7, 2026, cs.AIJ/K move · Enter open · S save

The Winner's Curse in LLM Self-Improvement Loops: Selection Noise, Lock-in, and Acceptance Rules

Litao Hu, Yutong Tang

Meta · Microsoft

LLM Self-Refinement Language Model Self-Improvement LLM Evaluation

Open paper AskPDF arXiv

Self-improving LLM systems propose changes to themselves and keep those that score better on a small evaluation set. We treat this keep-if-better step as selection under measurement noise, model the correlated errors of the candidates in a single decision, and study empirically what happens when the evaluation set is reused. In runs where Qwen models rewrite their own instructions and every candidate is also scored on 600 held-out items, most proposals after the first are harmful, and the model gives the size of the winner's curse of a generation's best candidate. With a prior from a separate pilot, it matches the average overstatement of first-generation commits in native loops, though not setting by setting. In a pre-registered study, the final selection-set score of greedy loops exceeded held-out accuracy by 13 to 20 points with 16 selection items and by 1 to 5 points with 256. Held-out gains grew with the selection set on TREC but not on GSM8K, and the tested acceptance rules did not beat greedy acceptance over whole runs. Gains measured on the selection set also exceeded held-out gains when a current model refined a competent instruction, and in the validation scores of GEPA and MIPROv2. Scoring the starting and the current instruction on 64 items never used for selection removes the average bias of a loop's reported gain, but single estimates remain off by about 6 points. Self-improvement studies should report held-out gains with their uncertainty.

Momentum

Related topics

Latest papers 62