# Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training

arXiv: https://arxiv.org/abs/2407.09121 · HTML: https://mlsanity.com/paper/2407.09121/

## Key facts

- Cited by 69 papers in the map (data through 14 Oct 2026).
- Builds on 54 of its 82 references that are in the map.
- First posted on arXiv 12 Jul 2024.

## Builds on (most cited first, 30 of 54)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Jailbreaking Black Box Large Language Models in Twenty Queries](https://mlsanity.com/paper/2310.08419/) (2023)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations](https://mlsanity.com/paper/2312.06674/) (2023)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [WizardLM: Empowering large pre-trained language models to follow complex instructions](https://mlsanity.com/paper/2304.12244/) (2023)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Self-Rewarding Language Models](https://mlsanity.com/paper/2401.10020/) (2024)
- [Baseline Defenses for Adversarial Attacks Against Aligned Language Models](https://mlsanity.com/paper/2309.00614/) (2023)
- [How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs](https://mlsanity.com/paper/2401.06373/) (2024)
- [GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts](https://mlsanity.com/paper/2309.10253/) (2023)
- [The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions](https://mlsanity.com/paper/2404.13208/) (2024)
- [Safety Alignment Should Be Made More Than Just a Few Tokens Deep](https://mlsanity.com/paper/2406.05946/) (2024)
- [SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks](https://mlsanity.com/paper/2310.03684/) (2023)
- [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](https://mlsanity.com/paper/2312.09390/) (2023)
- [Detecting Language Model Attacks with Perplexity](https://mlsanity.com/paper/2308.14132/) (2023)
- [Jailbreak and Guard Aligned Language Models with Only Few In-Context Demonstrations](https://mlsanity.com/paper/2310.06387/) (2023)

## Built on by (most cited first, 30 of 69)

- [LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yet](https://mlsanity.com/paper/2408.15221/) (2024)
- [LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution Shifts](https://mlsanity.com/paper/2410.10700/) (2024)
- [Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities](https://mlsanity.com/paper/2502.05209/) (2025)
- [Refusal Direction is Universal Across Safety-Aligned Languages](https://mlsanity.com/paper/2505.17306/) (2025)
- [Nuclear Deployed: Analyzing Catastrophic Risks in Decision-making of Autonomous LLM Agents](https://mlsanity.com/paper/2502.11355/) (2025)
- [There Is More to Refusal in Large Language Models than a Single Direction](https://mlsanity.com/paper/2602.02132/) (2026)
- [FORTRESS: Frontier Risk Evaluation for National Security and Public Safety](https://mlsanity.com/paper/2506.14922/) (2025)
- [Locking Down the Finetuned LLMs Safety](https://mlsanity.com/paper/2410.10343/) (2024)
- [From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring](https://mlsanity.com/paper/2506.09996/) (2025)
- [SoK: Evaluating Jailbreak Guardrails for Large Language Models](https://mlsanity.com/paper/2506.10597/) (2025)
- [X-Boundary: Establishing Exact Safety Boundary to Shield LLMs from Multi-Turn Jailbreaks without Compromising Usability](https://mlsanity.com/paper/2502.09990/) (2025)
- [Steering Dialogue Dynamics for Robustness against Multi-turn Jailbreaking Attacks](https://mlsanity.com/paper/2503.00187/) (2025)
- [Inverse Reinforcement Learning with Dynamic Reward Scaling for LLM Alignment](https://mlsanity.com/paper/2503.18991/) (2025)
- [SafeKey: Amplifying Aha-Moment Insights for Safety Reasoning](https://mlsanity.com/paper/2505.16186/) (2025)
- [Endless Jailbreaks with Bijection Learning](https://mlsanity.com/paper/2410.01294/) (2024)
- [Intelligent AI Delegation](https://mlsanity.com/paper/2602.11865/) (2026)
- [HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router](https://mlsanity.com/paper/2410.02684/) (2024)
- [Safety Reasoning with Guidelines](https://mlsanity.com/paper/2502.04040/) (2025)
- [Lifelong Safety Alignment for Language Models](https://mlsanity.com/paper/2505.20259/) (2025)
- [On Almost Surely Safe Alignment of Large Language Models at Inference-Time](https://mlsanity.com/paper/2502.01208/) (2025)
- [Course-Correction: Safety Alignment Using Synthetic Preferences](https://mlsanity.com/paper/2407.16637/) (2024)
- [Too Good to be Bad: On the Failure of LLMs to Role-Play Villains](https://mlsanity.com/paper/2511.04962/) (2025)
- [TraceRouter: Robust Safety for Large Foundation Models via Path-Level Intervention](https://mlsanity.com/paper/2601.21900/) (2026)
- [Jailbreaking to Jailbreak](https://mlsanity.com/paper/2502.09638/) (2025)
- [Safety Alignment Can Be Not Superficial With Explicit Safety Signals](https://mlsanity.com/paper/2505.17072/) (2025)
- [Align is not Enough: Multimodal Universal Jailbreak Attack against Multimodal Large Language Models](https://mlsanity.com/paper/2506.01307/) (2025)
- [POROver: Improving Safety and Reducing Overrefusal in Large Language Models with Overgeneration and Preference Optimization](https://mlsanity.com/paper/2410.12999/) (2024)
- [Regression for the Mean: Auto-Evaluation and Inference with Few Labels through Post-hoc Regression](https://mlsanity.com/paper/2411.12665/) (2024)
- [OMNIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities](https://mlsanity.com/paper/2505.23856/) (2025)
- [What Matters For Safety Alignment?](https://mlsanity.com/paper/2601.03868/) (2026)

## Abstract (arXiv)

This study addresses a critical gap in safety tuning practices for Large Language Models (LLMs) by identifying and tackling a refusal position bias within safety tuning data, which compromises the models' ability to appropriately refuse generating unsafe content. We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position, significantly enhancing their safety capabilities. DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation (MLE) with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence. Our empirical evaluation, conducted using LLaMA3 and Mistral model families across six attack scenarios, demonstrates that our method not only improves model safety without compromising performance but also surpasses baseline methods in defending against attacks.