# Decomposing LLM Self-Correction: The Accuracy-Correction Paradox and Error Depth Hypothesis

arXiv: https://arxiv.org/abs/2601.00828 · HTML: https://mlsanity.com/paper/2601.00828/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 8 references that are in the map.
- First posted on arXiv 24 Dec 2025.

## Builds on (most cited first, 5 of 5)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Self-Refine: Iterative Refinement with Self-Feedback](https://mlsanity.com/paper/2303.17651/) (2023)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)

## Built on by (most cited first, 8 of 8)

- [The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement](https://mlsanity.com/paper/2609.11873/) (2026)
- [Precise but Uncoupled: Reviewer Precision Does Not Guarantee Critique Uptake in Multi-Agent Math Reasoning](https://mlsanity.com/paper/2607.15388/) (2026)
- [From Hallucination to Structure Snowballing: The Alignment Tax of Constrained Decoding in LLM Reflection](https://mlsanity.com/paper/2604.06066/) (2026)
- [It's Not the Capability: Harness Sensitivity Is Non-Monotone Across LLM Agent Tiers](https://mlsanity.com/paper/2605.26731/) (2026)
- [The Horizon Gap: Planning, Memory, Execution, Training, and Evaluation for Long-Horizon LLM Agents](https://mlsanity.com/paper/2608.06663/) (2026)
- [ProCrit: Self-Elicited Multi-Perspective Reasoning with Critic-Guided Revision for Multimodal Sarcasm Detection](https://mlsanity.com/paper/2605.20867/) (2026)
- [Spectral Origins of the Self-Correction Blind Spot in Autoregressive Generation](https://mlsanity.com/paper/2607.09803/) (2026)
- [Why Didn't It Check? Unsupported Final Claims and Their Repair in Two Tool-Equipped Language Models](https://mlsanity.com/paper/2608.27768/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) are widely believed to possess self-correction capabilities, yet recent studies suggest that intrinsic self-correction--where models correct their own outputs without external feedback--remains largely ineffective. In this work, we systematically decompose self-correction into three distinct sub-capabilities: error detection, error localization, and error correction. Through cross-model experiments on GSM8K-Complex (n=500 per model, 346 total errors) with three major LLMs, we uncover a striking Accuracy-Correction Paradox: weaker models (GPT-3.5, 66% accuracy) achieve 1.6x higher intrinsic correction rates than stronger models (DeepSeek, 94% accuracy)--26.8% vs 16.7%. We propose the Error Depth Hypothesis: stronger models make fewer but deeper errors that resist self-correction. Error detection rates vary dramatically across architectures (10% to 82%), yet detection capability does not predict correction success--Claude detects only 10% of errors but corrects 29% intrinsically. Surprisingly, providing error location hints hurts all models. Our findings challenge linear assumptions about model capability and self-improvement, with important implications for the design of self-refinement pipelines.