When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMs
Authors:
Top Cited Papers
Open Access
- 4 November 2024
- journal article
- research article
- Published by MIT Press in Transactions of the Association for Computational Linguistics
- Vol. 12, 1417-1440
Abstract
Self-correction is an approach to improving responses from large language models (LLMs) by refining the responses using LLMs during inference. Prior work has proposed various self-correction frameworks using different sources of feedback, including self-evaluation and external feedback. However, there is still no consensus on the question of when LLMs can correct their own mistakes, as recent studies also report negative results. In this work, we critically survey broad papers and discuss the conditions required for successful self-correction. We first find that prior studies often do not define their research questions in detail and involve impractical frameworks or unfair evaluations that over-evaluate self-correction. To tackle these issues, we categorize research questions in self-correction research and provide a checklist for designing appropriate experiments. Our critical survey based on the newly categorized research questions shows that (1) no prior work demonstrates successful self-correction with feedback from prompted LLMs, except for studies in tasks that are exceptionally suited for self-correction, (2) self-correction works well in tasks that can use reliable external feedback, and (3) large-scale fine-tuning enables self-correction.Keywords
All Related Versions
This publication has 54 references indexed in Scilit:
- Automatically Correcting Large Language Models: Surveying the Landscape of Diverse Automated Correction StrategiesTransactions of the Association for Computational Linguistics, 2024
- Small Language Model Can Self-CorrectProceedings of the AAAI Conference on Artificial Intelligence, 2024
- Baldur: Whole-Proof Generation and Repair with Large Language Models2023
- Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPTransactions of the Association for Computational Linguistics, 2021
- Automatic Fact-Guided Sentence ModificationProceedings of the AAAI Conference on Artificial Intelligence, 2020
- Asking and Answering Questions to Evaluate the Factual Consistency of Summaries2020
- Getafix: learning to fix bugs automaticallyProceedings of the ACM on Programming Languages, 2019
- SEQUENCER: Sequence-to-Sequence Learning for End-to-End Program RepairIEEE Transactions on Software Engineering, 2019
- DeepDelta: learning to repair compilation errors2019
- DeepFix: Fixing Common C Language Errors by Deep LearningProceedings of the AAAI Conference on Artificial Intelligence, 2017