mlsanityData to 7 Oct 2026

JPU: Bridging Jailbreak Defense and Unlearning via On-Policy Path Rectification

arXiv 2601.03005 ↗First posted 6 Jan 2026cs.CR cs.AIcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks

also in 1 more
Abstract (arXiv)

Despite extensive safety alignment, Large Language Models (LLMs) often fail against jailbreak attacks. While machine unlearning has emerged as a promising defense by erasing specific harmful parameters, current methods remain vulnerable to diverse jailbreaks. We first conduct an empirical study and discover that this failure mechanism is caused by jailbreaks primarily activating non-erased parameters in the intermediate layers. Further, by probing the underlying mechanism through which these circumvented parameters reassemble into the prohibited output, we verify the persistent existence of dynamic jailbreak paths and show that the inability to rectify them constitutes the fundamental gap in existing unlearning defenses. To bridge this gap, we propose Jailbreak Path Unlearning (JPU), which is the first to rectify dynamic jailbreak paths towards safety anchors by dynamically mining on-policy adversarial samples to expose vulnerabilities and identify jailbreak paths. Extensive experiments demonstrate that JPU significantly enhances jailbreak resistance against dynamic attacks while preserving the model's utility.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 28 of its 45 references that are in the map. First posted on arXiv 6 Jan 2026.

Builds on

28 of 45 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Safety Alignment via Constrained Knowledge Unlearning24 May
  2. 2024Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications7 Feb

Other references · 26

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  3. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  4. 2018GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understandingcited by 4,854
  5. 2019HellaSwag: Can a Machine Really Finish Your Sentence?cited by 4,018
  6. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,114
  7. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,793
  8. 2023Jailbreaking Black Box Large Language Models in Twenty Queriescited by 1,326
  9. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 911
  10. 2023GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Promptscited by 520
  11. 2023Catastrophic Jailbreak of Open-source LLMs via Exploiting Generationcited by 439
  12. 2023SmoothLLM: Defending Large Language Models Against Jailbreaking Attackscited by 434
14 more
  1. 2024Improving Alignment and Robustness with Circuit Breakerscited by 302
  2. 2023A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easilycited by 242
  3. 2024SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decodingcited by 233
  4. 2024AmpleGCG: Learning a Universal and Transferable Generative Model of Adversarial Suffixes for Jailbreaking Both Open and Closed LLMscited by 160
  5. 2023Generative Language Models and Automated Influence Operations: Emerging Threats and Potential Mitigationscited by 157
  6. 2023RAIN: Your Language Models Can Align Themselves without Finetuningcited by 149
  7. 2024Reinforcement Learning for LLM Post-Training: A Surveycited by 91
  8. 2024GradSafe: Detecting Jailbreak Prompts for LLMs via Safety-Critical Gradient Analysiscited by 87
  9. 2024Defending Large Language Models Against Jailbreak Attacks via Layer-specific Editingcited by 77
  10. 2022Debiasing isn't enough! -- On the Effectiveness of Debiasing MLMs and their Social Biases in Downstream Taskscited by 48
  11. 2024Eraser: Jailbreaking Defense in Large Language Models via Unlearning Harmful Knowledgecited by 48
  12. 2024Improving the Robustness of Large Language Models via Consistency Alignmentcited by 36
  13. 2024RMCBench: Benchmarking Large Language Models' Resistance to Malicious Codecited by 27
  14. 2025DETAM: Defending LLMs Against Jailbreak Attacks via Targeted Attention Modificationcited by 10

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025JBShield: Defending Large Language Models from Jailbreak Attacks through Activated Concept Analysis and Manipulation
  2. 2025SafeLLM: Unlearning Harmful Outputs from Large Language Models against Jailbreak Attacks
  3. 2025Scalable Defense against In-the-wild Jailbreaking Attacks with Safety Context Retrieval
  4. 2026SAFEGuard: Detect Optimization-Based Jailbreak Attacks Through Harmful Semantic Analysis and Fluency Measurement
  5. 2026The Art of the Jailbreak: Formulating Jailbreak Attacks for LLM Security Beyond Binary Scoring
  6. 2025Defense-to-Attack: Bypassing Weak Defenses Enables Stronger Jailbreaks in Vision-Language Models
  7. 2026Jailbreaking Leaves a Trace: Understanding and Detecting Jailbreak Attacks from Internal Representations of Large Language Models
  8. 2026Circuit Discovery Helps Detect LLM Jailbreaking: A Mechanistic Interpretability Study