Reinforcement Learning with Backtracking Feedback

• URL: http://arxiv.org/abs/2602.08377v1
• Date: Mon, 09 Feb 2026 08:23:19 GMT
• Title: Reinforcement Learning with Backtracking Feedback
• Authors: Bilgehan Sel, Vaishakh Keshava, Phillip Wallis, Lukas Rutishauser, Ming Jin, Dingcheng Li,
• Abstract summary: We introduce Reinforcement Learning with Backtracking Feedback (RLBF)<n>This framework advances upon prior methods, such as BSAFE.<n>We show that RLBF significantly reduces attack success rates across diverse benchmarks and model scales.
• Score: 12.680874918250069
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Addressing the critical need for robust safety in Large Language Models (LLMs), particularly against adversarial attacks and in-distribution errors, we introduce Reinforcement Learning with Backtracking Feedback (RLBF). This framework advances upon prior methods, such as BSAFE, by primarily leveraging a Reinforcement Learning (RL) stage where models learn to dynamically correct their own generation errors. Through RL with critic feedback on the model's live outputs, LLMs are trained to identify and recover from their actual, emergent safety violations by emitting an efficient "backtrack by x tokens" signal, then continuing generation autoregressively. This RL process is crucial for instilling resilience against sophisticated adversarial strategies, including middle filling, Greedy Coordinate Gradient (GCG) attacks, and decoding parameter manipulations. To further support the acquisition of this backtracking capability, we also propose an enhanced Supervised Fine-Tuning (SFT) data generation strategy (BSAFE+). This method improves upon previous data creation techniques by injecting violations into coherent, originally safe text, providing more effective initial training for the backtracking mechanism. Comprehensive empirical evaluations demonstrate that RLBF significantly reduces attack success rates across diverse benchmarks and model scales, achieving superior safety outcomes while critically preserving foundational model utility.

Related papers

• Native Reasoning Models: Training Language Models to Reason on Unverifiable Data [16.065264121785294]
  We introduce NRT (Native Reasoning Training), a novel framework that cultivates complex reasoning.<n>NRT reframes the training problem by treating the reasoning process as a latent variable.<n>NRT achieves state-of-the-art performance among verifier-free methods.
  arXiv  Detail & Related papers  (2026-02-12T04:15:46Z)
• Towards Sample-Efficient and Stable Reinforcement Learning for LLM-based Recommendation [56.92367609590823]
  Long Chain-of-Thought (Long CoT) reasoning has shown promise in Large Language Models (LLMs)<n>We argue that Long CoT is inherently ill-suited for the sequential recommendation domain.<n>We propose RISER, a novel Reinforced Item Space Exploration framework for Recommendation.
  arXiv  Detail & Related papers  (2026-01-31T10:02:43Z)
• Deep Leakage with Generative Flow Matching Denoiser [54.05993847488204]
  We introduce a new deep leakage (DL) attack that integrates a generative Flow Matching (FM) prior into the reconstruction process.<n>Our approach consistently outperforms state-of-the-art attacks across pixel-level, perceptual, and feature-based similarity metrics.
  arXiv  Detail & Related papers  (2026-01-21T14:51:01Z)
• Principled RL for Diffusion LLMs Emerges from a Sequence-Level Perspective [85.06838178922791]
  Reinforcement Learning (RL) has proven highly effective for autoregressive language models.<n>But adapting these methods to diffusion large language models (dLLMs) presents fundamental challenges.<n>We propose a principled RL framework that treats entire sequence generation as a single action and uses the ELBO as a tractable sequence-level likelihood proxy.
  arXiv  Detail & Related papers  (2025-12-03T13:05:32Z)
• Rethinking Deep Alignment Through The Lens Of Incomplete Learning [14.306119791052575]
  We show that position-dependent gradient weakening during autoregressive training creates signal decay, leading to incomplete safety learning.<n>We introduce base-favored tokens as computational indicators of incomplete safety learning.<n> Experimental evaluation across Llama and Qwen model families demonstrates dramatic improvements in adversarial robustness.
  arXiv  Detail & Related papers  (2025-11-15T10:53:03Z)
• Beyond Reasoning Gains: Mitigating General Capabilities Forgetting in Large Reasoning Models [33.214586668992965]
  Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning.<n>We propose RECAP-a replay strategy with dynamic objective reweighting for general knowledge.<n>Our method is end-to-end and readily applicable to existing RLVR pipelines without training additional models or heavy tuning.
  arXiv  Detail & Related papers  (2025-10-24T19:08:48Z)
• Large Reasoning Models Learn Better Alignment from Flawed Thinking [56.08883934423522]
  Large reasoning models (LRMs) "think" by generating structured chain-of-thought (CoT) before producing a final answer.<n>We propose RECAP, a principled reinforcement learning (RL) method for post-training that explicitly teaches models to override flawed reasoning trajectories.
  arXiv  Detail & Related papers  (2025-10-01T14:15:43Z)
• Reinforcement Learning on Pre-Training Data [55.570379963147424]
  We introduce Reinforcement Learning on Pre-Training data (R), a new training-time scaling paradigm for optimizing large language models (LLMs)<n>R enables the policy to autonomously explore meaningful trajectories to learn from pre-training data and improve its capability through reinforcement learning (RL)<n>Extensive experiments on both general-domain and mathematical reasoning benchmarks across multiple models validate the effectiveness of R.
  arXiv  Detail & Related papers  (2025-09-23T17:10:40Z)
• RL as Regressor: A Reinforcement Learning Approach for Function Approximation [0.0]
  We propose framing regression as a Reinforcement Learning (RL) problem.<n>We demonstrate this by treating a model's prediction as an action and defining a custom reward signal based on the prediction error.<n>We show that the RL framework not only successfully solves the regression problem but also offers enhanced flexibility in defining objectives and guiding the learning process.
  arXiv  Detail & Related papers  (2025-07-31T21:39:24Z)
• One Token to Fool LLM-as-a-Judge [52.45386385722788]
  Large language models (LLMs) are increasingly trusted as automated judges, assisting evaluation and providing reward signals for training other models.<n>We uncover a critical vulnerability even in this reference-based paradigm: generative reward models are systematically susceptible to reward hacking.
  arXiv  Detail & Related papers  (2025-07-11T17:55:22Z)
• Learn from the Past: A Proxy Guided Adversarial Defense Framework with Self Distillation Regularization [53.04697800214848]
  Adversarial Training (AT) is pivotal in fortifying the robustness of deep learning models. AT methods, relying on direct iterative updates for target model's defense, frequently encounter obstacles such as unstable training and catastrophic overfitting. We present a general proxy guided defense framework, LAST' (bf Learn from the Pbf ast)
  arXiv  Detail & Related papers  (2023-10-19T13:13:41Z)
• Robust Reinforcement Learning Objectives for Sequential Recommender Systems [7.44049827436013]
  We develop recommender systems that incorporate direct user feedback in the form of rewards, enhancing personalization for users. employing RL algorithms presents challenges, including off-policy training, expansive action spaces, and the scarcity of datasets with sufficient reward signals. We introduce an enhanced methodology aimed at providing a more effective solution to these challenges.
  arXiv  Detail & Related papers  (2023-05-30T08:09:08Z)

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.