Article

Limits of Self-Correction in LLMs: An Information-Theoretic Analysis of Correlated Errors

2026-01-13

Abstract excerpt

Recent empirical work shows that large language models struggle to self-correct reasoning without external feedback. We propose a possible explanation: correlated error between generator and evaluator. When both components share failure modes, self-evaluation may provide weak evidence of correctness, and repeated self-critique may amplify confidence without adding information. We formalize this with two informatio...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: 96141299-b847-5c3c-935d-62b24d41f914
DOI: 10.20944/preprints202601.0892.v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Limits of Self-Correction in LLMs: An Information-Theoretic Analysis of Correlated ErrorsDOI 10.20944/preprints202601.0892.v1Read

Enhancing Reliability in Large Language Models: Self-Detection of Hallucinations With Spontaneous Self-ChecksSemantic proximity Read

Verifier Exploitation in NLI-Guided Iterative Refinement: A Controlled Empirical AnalysisSemantic proximity Read

RePCS: Diagnosing Data Memorization in LLM-Powered Retrieval-Augmented GenerationSemantic proximity Read

A Survey on Hallucination in Large Language and Foundation ModelsSemantic proximity Read

A Survey on Hallucination in Large Language and Foundation ModelsSemantic proximity Read

The Unreliable Judges: Assessing Reproducibility and Self-Preference Bias of LLMs as Free-Text EvaluatorsSemantic proximity Read

Mind the Gap: From Plausible to Valid Self-Explanations in Large Language ModelsSemantic proximity Read

From Illusion to Insight: A Taxonomic Survey of Hallucination Mitigation Techniques in LLMsSemantic proximity Read

Foresight Is Not Enough: Sentence-Level Future Signals, Self-Loop Hard Negatives, and the Calibration Gap in Small Language ModelsSemantic proximity Read

Graduated Dissent: Budgeted Disagreement Resolution for Multi-Model InferenceSemantic proximity Read

Addressee Framing Systematically Reduces LLM Structural Pathology: A Controlled Empirical Study of Directed Information FlowSemantic proximity Read

Explainable Automatic Evaluation: Interpretable Decision-Making in Large Language Model AutoratersSemantic proximity Read