mlsanityData to 7 Oct 2026

A Training-Free Regeneration Paradigm: Contrastive Reflection Memory Guided Self-Verification and Self-Improvement

arXiv 2603.20441 ↗First posted 20 Mar 2026cs.CLcited by 0

On the map: Synthetic data and self-improvement for LLMs → Self-rewarding alignment and self-training

Abstract (arXiv)

Verification-guided self-improvement has recently emerged as a promising approach to improving the accuracy of large language model (LLM) outputs. However, existing approaches face a trade-off between inference efficiency and accuracy: iterative verification-rectification is computationally expensive and prone to being trapped in faulty reasoning, while best-of-N selection requires extensive sampling without addressing internal model flaws. We propose a training-free regeneration paradigm that leverages an offline-curated contrastive Reflection Memory (RM) to provide corrective guidance, while regenerating from scratch helps break out of faulty reasoning. At inference time, the method performs RM-guided self-verification followed by a single RM-guided regeneration, avoiding both iterative correction and multi-sample selection. We evaluated our method on nine benchmarks that span algorithmic, reasoning, symbolic, and domain-specific tasks in both small- and large-scale LLMs. Experiment results show that our method outperforms prior methods while maintaining low computational cost.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Self-rewarding alignment and self-training, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 26 of its 34 references that are in the map. First posted on arXiv 20 Mar 2026.

Builds on

26 of 34 references.

References · 26

  1. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  2. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  3. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  4. 2021Unsupervised Dense Information Retrieval with Contrastive Learningcited by 1,203
  5. 2022Measuring and Narrowing the Compositionality Gap in Language Modelscited by 1,054
  6. 2020MPNet: Masked and Permuted Pre-training for Language Understandingcited by 973
  7. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 965
  8. 2023Teaching Large Language Models to Self-Debugcited by 963
  9. 2023CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquingcited by 751
  10. 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 736
  11. 2024Training Language Models to Self-Correct via Reinforcement Learningcited by 405
  12. 2024When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMscited by 230
14 more
  1. 2024V-STaR: Training Verifiers for Self-Taught Reasonerscited by 214
  2. 2024Chain-of-Thought Reasoning Without Promptingcited by 208
  3. 2024Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoningcited by 110
  4. 2020Impact of News on the Commodity Market: Dataset and Resultscited by 80
  5. 2024In-Context Principle Learning from Mistakescited by 42
  6. 2024Large Language Models Can Self-Correct with Key Condition Verificationcited by 39
  7. 2025Reverse-Engineered Reasoning for Open-Ended Generationcited by 32
  8. 2024In-Context Decision Transformer: Reinforcement Learning via Hierarchical Chain-of-Thoughtcited by 25
  9. 2024C-ICL: Contrastive In-context Learning for Information Extractioncited by 22
  10. 2024Customizing Language Model Responses with Contrastive In-Context Learningcited by 19
  11. 2025Can LLMs Correct Themselves? A Benchmark of Self-Correction in LLMscited by 12
  12. 2024Rethinking Chain-of-Thought from the Perspective of Self-Trainingcited by 10
  13. 2025Toward Structured Knowledge Reasoning: Contrastive Retrieval-Augmented Generation on Experiencecited by 7
  14. 2025Training-free LLM Verification via Recycling Few-shot Examplescited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes
  2. 2026Training-Free Inference-Time Self-Reflection and Cost-Bounded Early Stopping for Large Language Models
  3. 2026Can Language Models Learn to Reject Their Own Bad Reasoning Steps?
  4. 2025ReVISE: Learning to Refine at Test-Time via Intrinsic Self-Verification
  5. 2026ReTreVal: Reasoning Tree with Validation and Cross-Problem Memory for Large Language Models
  6. 2025Reflect, Retry, Reward: Self-Improving LLMs via Reinforcement Learning
  7. 2025Think Twice, Generate Once: Safeguarding by Progressive Self-Reflection
  8. 2025Instruct-of-Reflection: Enhancing Large Language Models Iterative Reflection Capabilities via Dynamic-Meta Instruction