Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs

• URL: http://arxiv.org/abs/2608.11573v1
• Date: Wed, 12 Aug 2026 02:31:02 GMT
• Title: Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs
• Abstract summary: Self-Fix Step-DPO is a reinforcement learning based framework for step-level self-verification and self-correction.<n>We show that SFS-DPO and SFS-DPO-R consistently outperform prior step-level training baselines.
• Score: 9.698213596197528
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Achieving effective self-correction, where models verify and correct their own mistakes, remains a fundamental challenge for large language models (LLMs). In this work, we propose Self-Fix Step-DPO (SFS-DPO), a reinforcement learning based, two-stage framework for step-level self-verification and self-correction. The first stage strengthens step-level reasoning via step-level preference optimization, while the second stage explicitly trains models to self-verify and self-correct. We further introduce a teacher-assisted variant, SFS-DPO-R, which incorporates explanatory rationales for error verification to provide stronger corrective signals. Comprehensive in-domain and out-of-domain evaluations across multiple LLMs demonstrate that SFS-DPO and SFS-DPO-R consistently outperform prior step-level training baselines. Our analysis further reveals improvements in self-correction frequency and effectiveness, highlighting the importance of strengthening step-level reasoning for robust performance.

Related papers

• Diagnosing Multi-step Reasoning Failures in Black-box LLMs via Stepwise Confidence Attribution [12.955460962298588]
  We introduce Stepwise Confidence Attribution (SCA), a framework for closed-source LLMs that assigns step-level confidence based only on generated reasoning traces.<n>SCA reliably identifies low-confidence steps strongly correlated with reasoning errors.<n>Using step-level confidence to guide self-correction improves the correction success rate by up to 13.5% over answer-level feedback.
  arXiv  Detail & Related papers  (2026-05-19T00:57:51Z)
• On Predicting the Post-training Potential of Pre-trained LLMs [60.07459271263409]
  We introduce a new task of predicting post-training potential - forecasting a base model's performance before post-training.<n>We propose RuDE, a unified framework that bypasses the generation gap of base models by leveraging response discrimination.<n>Experiments demonstrate a correlation greater than 90% with post-training performance.
  arXiv  Detail & Related papers  (2026-05-12T11:33:49Z)
• Team-Based Self-Play With Dual Adaptive Weighting for Fine-Tuning LLMs [24.816746511570244]
  Team-based self-Play with dual Adaptive Weighting (TPAW) is a novel self-play algorithm designed to improve alignment in a fully self-supervised setting.<n>TPAW iteratively refines alignment without requiring additional human supervision.
  arXiv  Detail & Related papers  (2026-05-11T03:17:31Z)
• Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards [71.19033708090389]
  Reinforcement Learning from Verifiable Rewards (RLVR) significantly enhances large language models (LLMs) reasoning but severely suffers from calibration degeneration.<n>We propose DCPO, a framework that systematically decouples reasoning and calibration objectives.
  arXiv  Detail & Related papers  (2026-03-10T02:47:59Z)
• Data Distribution as a Lever for Guiding Optimizers Toward Superior Generalization in LLMs [60.68927774057402]
  We show, for the first time, that a lower simplicity bias induces a better generalization.<n>Motivated by this insight, we demonstrate that the training data distribution by upsampling or augmenting examples learned later in training similarly reduces SB and leads to improved generalization.<n>Our strategy improves the performance of multiple language models including Phi2-2.7B, Llama3.2-1B, Gemma3-1B-PT, Qwen3-0.6B-Base-achieving relative accuracy gains up to 18% when fine-tuned with AdamW and Muon.
  arXiv  Detail & Related papers  (2026-01-31T07:40:36Z)
• PAG: Multi-Turn Reinforced LLM Self-Correction with Policy as Generative Verifier [18.771754895027616]
  Policy as Generative Verifier (PAG) is a framework that empowers Large Language Models to self-correct by alternating between policy and verifier roles.<n>It alleviates model collapse and jointly enhances both reasoning and verification abilities.
  arXiv  Detail & Related papers  (2025-06-12T06:59:35Z)
• S$^2$R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning [51.84977135926156]
  We introduce S$2$R, an efficient framework that enhances LLM reasoning by teaching models to self-verify and self-correct during inference.<n>Our results demonstrate that Qwen2.5-math-7B achieves an accuracy improvement from 51.0% to 81.6%, outperforming models trained on an equivalent amount of long-CoT distilled data.
  arXiv  Detail & Related papers  (2025-02-18T13:40:22Z)
• LEAF: Learning and Evaluation Augmented by Fact-Checking to Improve Factualness in Large Language Models [11.453585039783901]
  LEAF: Learning and Evaluation Augmented by Fact-Checking, is a novel approach designed to enhance the factual reliability of large language models (LLMs) The first strategy, Fact-Check-Then-RAG, improves Retrieval-Augmented Generation (RAG) by incorporating fact-checking results to guide the retrieval process without updating model parameters. The second strategy, Learning from Fact-Checks via Self-Training, involves supervised fine-tuning (SFT) on fact-checked responses or applying Simple Preference Optimization (SimPO) with fact-checking as a ranking mechanism.
  arXiv  Detail & Related papers  (2024-10-31T00:18:05Z)
• Training Language Models to Self-Correct via Reinforcement Learning [98.35197671595343]
  Self-correction has been found to be largely ineffective in modern large language models (LLMs) We develop a multi-turn online reinforcement learning approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. We find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on MATH and HumanEval.
  arXiv  Detail & Related papers  (2024-09-19T17:16:21Z)
• Model Extrapolation Expedites Alignment [135.12769233630362]
  We propose a method called ExPO to expedite alignment training with human preferences.<n>We demonstrate that ExPO boosts a DPO model trained with only 20% steps to outperform the fully-trained one.<n>We show that ExPO notably improves existing open-source LLMs on the leading AlpacaEval 2.0 and MT-Bench benchmarks.
  arXiv  Detail & Related papers  (2024-04-25T17:39:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.