In-Place Feedback: Reliable Refinement for Multi-Turn Expert-LLM Collaboration

2025/10/01 by Youn Seon Choi, M.-A. Lee, Choi, Youngbin +10
Computer Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Natural Language Processing Techniques #Semantic Web and Ontologies

paper · pdf · doi:10.48550/arxiv.2510.00777

openalex publication_date 2025/10/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

LLM-generated drafts often contain subtle factual or logical errors, yet prior work shows that models struggle to reliably integrate multi-turn feedback aimed at fixing them. We propose in-place feedback, an interaction paradigm in which the user directly edits the model's previous response and the model continues generation from the edited context. In-place feedback consistently outperforms standard multi-turn feedback across five reasoning-intensive benchmarks while requiring fewer tokens, and our fine-grained analysis shows that it applies corrections more reliably and propagates them to subsequent reasoning. A user study with domain experts refining LLM-generated summaries corroborates these findings: participants report higher final-output satisfaction and substantially lower fatigue with in-place feedback, and a mixed strategy combining in-place and multi-turn feedback scores highest on every measured dimension. These results suggest that editing errors directly is a more effective paradigm for expert-LLM collaboration.

Citations

• EDIT-Bench: Evaluating LLM Abilities to Perform Real-World Instructed Code Edits
• gpt-oss-120b & gpt-oss-20b Model Card
• Answer Matching Outperforms Multiple Choice for Language Model Evaluation
• LLMs Get Lost In Multi-Turn Conversation
• Training Turn-by-Turn Verifiers for Dialogue Tutoring Agents: The Curious Case of LLMs as Your Coding Tutors
• ZebraLogic: On the Scaling Limits of LLMs for Logical Reasoning
• CollabLLM: From Passive Responders to Active Collaborators
• MultiChallenge: A Realistic Multi-Turn Conversation Evaluation Benchmark Challenging to Frontier LLMs
• Modeling Future Conversation Turns to Teach LLMs to Ask Clarifying Questions
• Let the Code LLM Edit Itself When You Edit the Code
• MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
• Multi-turn Reinforcement Learning from Preference Human Feedback
• LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
• ArCHer: Training Language Model Agents via Hierarchical Multi-Turn RL
• GPQA: A Graduate-Level Google-Proof Q&A Benchmark
• MemGPT: Towards LLMs as Operating Systems
• Large Language Models Cannot Self-Correct Reasoning Yet
• MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback
• Efficient Memory Management for Large Language Model Serving with PagedAttention
• Self-Refine: Iterative Refinement with Self-Feedback
• Generating Sequences by Learning to Self-Correct
• In-context Learning and Induction Heads
• CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities
• Measuring Mathematical Problem Solving With the MATH Dataset

Related

• MAF: Multi-Aspect Feedback for Improving Reasoning in Large Language Models
• A NotSo Simple Way to Beat Simple Bench
• LLMRefine: Pinpointing and Refining Large Language Models via Fine-Grained Actionable Feedback
• Feedback Friction: LLMs Struggle to Fully Incorporate External Feedback
• Multi-turn Training with Basic Human Feedback Helps Little on LLM Reasoning