# Self-Correction Bench: Uncovering and Addressing the Self-Correction Blind Spot in Large Language Models

arXiv: https://arxiv.org/abs/2507.02778 · HTML: https://mlsanity.com/paper/2507.02778/

## Key facts

- Cited by 20 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 36 of its 54 references that are in the map.
- First posted on arXiv 3 Jul 2025.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 30 of 36)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (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)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions](https://mlsanity.com/paper/2311.05232/) (2023)
- [s1: Simple test-time scaling](https://mlsanity.com/paper/2501.19393/) (2025)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Large Language Models Cannot Self-Correct Reasoning Yet](https://mlsanity.com/paper/2310.01798/) (2023)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [On Faithfulness and Factuality in Abstractive Summarization](https://mlsanity.com/paper/2005.00661/) (2020)
- [A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity](https://mlsanity.com/paper/2302.04023/) (2023)
- [Large Language Models Can Be Easily Distracted by Irrelevant Context](https://mlsanity.com/paper/2302.00093/) (2023)
- [OpenAssistant Conversations -- Democratizing Large Language Model Alignment](https://mlsanity.com/paper/2304.07327/) (2023)
- [Training Language Models to Self-Correct via Reinforcement Learning](https://mlsanity.com/paper/2409.12917/) (2024)
- [Language Models can Solve Computer Tasks](https://mlsanity.com/paper/2303.17491/) (2023)
- [Formalizing and Benchmarking Prompt Injection Attacks and Defenses](https://mlsanity.com/paper/2310.12815/) (2023)
- [How Language Model Hallucinations Can Snowball](https://mlsanity.com/paper/2305.13534/) (2023)
- [When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMs](https://mlsanity.com/paper/2406.01297/) (2024)
- [The Effect of Sampling Temperature on Problem Solving in Large Language Models](https://mlsanity.com/paper/2402.05201/) (2024)
- [Benchmarking Cognitive Biases in Large Language Models as Evaluators](https://mlsanity.com/paper/2309.17012/) (2023)
- [Cognitive Bias in Decision-Making with LLMs](https://mlsanity.com/paper/2403.00811/) (2024)
- [PRMBench: A Fine-grained and Challenging Benchmark for Process-Level Reward Models](https://mlsanity.com/paper/2501.03124/) (2025)
- [Capturing Failures of Large Language Models via Human Cognitive Biases](https://mlsanity.com/paper/2202.12299/) (2022)

## Built on by (most cited first, 20 of 20)

- [Why Reasoning Fails to Plan: A Planning-Centric Analysis of Long-Horizon Decision Making in LLM Agents](https://mlsanity.com/paper/2601.22311/) (2026)
- [The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement](https://mlsanity.com/paper/2609.11873/) (2026)
- [The Illusion of Insight in Reasoning Models](https://mlsanity.com/paper/2601.00514/) (2026)
- [OptiRepair: Closed-Loop Diagnosis and Repair of Supply Chain Optimization Models with LLM Agents](https://mlsanity.com/paper/2602.19439/) (2026)
- [Self-Improvement of Large Language Models: A Technical Overview and Future Outlook](https://mlsanity.com/paper/2603.25681/) (2026)
- [The Narrative Continuity Test: A Conceptual Framework for Evaluating Identity Persistence in AI Systems](https://mlsanity.com/paper/2510.24831/) (2025)
- [Structure Enables Effective Self-Localization of Errors in LLMs](https://mlsanity.com/paper/2602.02416/) (2026)
- [Self-Attribution Bias: When AI Monitors Go Easy on Themselves](https://mlsanity.com/paper/2603.04582/) (2026)
- [Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Models](https://mlsanity.com/paper/2604.16565/) (2026)
- [Partially Correlated Verifier Cascades in LLM Harnesses: Concave Log-Odds, Polynomial Reliability, and Blind-Spot Ceilings](https://mlsanity.com/paper/2607.13918/) (2026)
- [Governing Reflective Human-AI Collaboration: A Framework for Epistemic Scaffolding and Traceable Reasoning](https://mlsanity.com/paper/2604.14898/) (2026)
- [ReFlect: An Effective Harness System for Complex Long-Horizon LLM Reasoning](https://mlsanity.com/paper/2605.05737/) (2026)
- [ESC: Emotional Self-Correction for Reliable Vision-Language Models](https://mlsanity.com/paper/2607.02089/) (2026)
- [CUSUM-Shaped Inference-Time Monitoring and Targeted Re-Decoding for Quantized Small Language Model Reasoning](https://mlsanity.com/paper/2607.20129/) (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)
- [Calibrated e-CUSUM Decoding for Quantized Reasoning Models: Why Token Log-Probability Is the Wrong Observable for Decoding Monitors](https://mlsanity.com/paper/2607.11317/) (2026)
- [Think Through a Bottleneck: Hourglass Reasoning for Rigorous Induction](https://mlsanity.com/paper/2607.11696/) (2026)
- [When Do Models Admit They Are Wrong? Failure Disclosure Is Unstable Under Reinforcement Learning](https://mlsanity.com/paper/2609.33220/) (2026)
- [Anatomy of LLM Sycophancy: What a Flip Rate Hides](https://mlsanity.com/paper/2610.06522/) (2026)

## Abstract (arXiv)

Although large language models (LLMs) have transformed AI, they still make errors and follow unproductive reasoning paths. Self-correction is vital for safety-critical applications, but studying it requires disentangling activation failure from knowledge deficiency: when a model fails to correct an error, is it because it cannot, or because it does not? We introduce Self-Correction Bench, a controlled evaluation framework that isolates this distinction by injecting the same error as either an external (user-attributed) or internal (model-attributed) error, keeping all other context identical. Testing 14 open-source non-reasoning models reveals a 64.5% Self-Correction Blind Spot: models correct external errors but fail on identical internal ones, proving the capability exists but is not activated. On models' own naturally generated errors, a measurable share of what a model fails to catch in its own output is caught when the identical error is presented externally. We trace the cause to post-training data composition: supervised fine-tuning datasets lack error-correction sequences, and fine-tuning with as few as 5,306 such traces already reduces the blind spot by 76.0%. Mechanistically, we identify a transferable conversational-role direction in representation space that causally gates self-correction. Appending "Wait" requires no training yet reduces the blind spot by 89.3%, and operates through a nearly independent pathway, indicating that correction activation is not reducible to this single mechanism.