# MTSA: Multi-turn Safety Alignment for LLMs through Multi-round Red-teaming

arXiv: https://arxiv.org/abs/2505.17147 · HTML: https://mlsanity.com/paper/2505.17147/

## Key facts

- Cited by 32 papers in the map (data through 14 Oct 2026).
- Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 39 of its 52 references that are in the map.
- First posted on arXiv 22 May 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)

## Builds on (most cited first, 30 of 39)

- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (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)
- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations](https://mlsanity.com/paper/2312.06674/) (2023)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [A General Theoretical Paradigm to Understand Learning from Human Preferences](https://mlsanity.com/paper/2310.12036/) (2023)
- [BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset](https://mlsanity.com/paper/2307.04657/) (2023)
- [Enhancing Chat Language Models by Scaling High-quality Instructional Conversations](https://mlsanity.com/paper/2305.14233/) (2023)
- [AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback](https://mlsanity.com/paper/2305.14387/) (2023)
- [Safe RLHF: Safe Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2310.12773/) (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)
- [GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Cipher](https://mlsanity.com/paper/2308.06463/) (2023)
- [Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts](https://mlsanity.com/paper/2406.12845/) (2024)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment](https://mlsanity.com/paper/2308.09662/) (2023)
- [On Second Thought, Let's Not Think Step by Step! Bias and Toxicity in Zero-Shot Reasoning](https://mlsanity.com/paper/2212.08061/) (2022)
- [MART: Improving LLM Safety with Multi-round Automatic Red-Teaming](https://mlsanity.com/paper/2311.07689/) (2023)
- [LLM Defenses Are Not Robust to Multi-Turn Human Jailbreaks Yet](https://mlsanity.com/paper/2408.15221/) (2024)
- [Language Models are Homer Simpson! Safety Re-Alignment of Fine-tuned Language Models through Task Arithmetic](https://mlsanity.com/paper/2402.11746/) (2024)
- [Chain of Attack: a Semantic-Driven Contextual Multi-Turn attacker for LLM](https://mlsanity.com/paper/2405.05610/) (2024)
- [REBEL: Reinforcement Learning via Regressing Relative Rewards](https://mlsanity.com/paper/2404.16767/) (2024)
- [Speak Out of Turn: Safety Vulnerability of Large Language Models in Multi-turn Dialogue](https://mlsanity.com/paper/2402.17262/) (2024)
- [CoSafe: Evaluating Large Language Model Safety in Multi-Turn Dialogue Coreference](https://mlsanity.com/paper/2406.17626/) (2024)
- [Distract Large Language Models for Automatic Jailbreak Attack](https://mlsanity.com/paper/2403.08424/) (2024)
- [Multimodal Reasoning with Multimodal Knowledge Graph](https://mlsanity.com/paper/2406.02030/) (2024)
- [MRJ-Agent: An Effective Jailbreak Agent for Multi-Round Dialogue](https://mlsanity.com/paper/2411.03814/) (2024)

## Built on by (most cited first, 30 of 32)

- [CounselBench: A Large-Scale Expert Evaluation and Adversarial Benchmarking of Large Language Models in Mental Health Question Answering](https://mlsanity.com/paper/2506.08584/) (2025)
- [MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety](https://mlsanity.com/paper/2602.01539/) (2026)
- [TROJail: Trajectory-Level Optimization for Multi-Turn Large Language Model Jailbreaks with Process Rewards](https://mlsanity.com/paper/2512.07761/) (2025)
- [Multi-objective Large Language Model Alignment with Hierarchical Experts](https://mlsanity.com/paper/2505.20925/) (2025)
- [STAC: When Innocent Tools Form Dangerous Chains for LLM Agents](https://mlsanity.com/paper/2509.25624/) (2025)
- [Tree-based Dialogue Reinforced Policy Optimization for Red-Teaming Attacks](https://mlsanity.com/paper/2510.02286/) (2025)
- [DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs](https://mlsanity.com/paper/2602.16935/) (2026)
- [Safeguarding LLM Agents against Long-Horizon Threats via Shadow Memory](https://mlsanity.com/paper/2605.03228/) (2026)
- [Eliciting Behaviors in Multi-Turn Conversations](https://mlsanity.com/paper/2512.23701/) (2025)
- [Do No Harm: Exposing Hidden Vulnerabilities of LLMs via Persona-based Client Simulation Attack in Psychological Counseling](https://mlsanity.com/paper/2604.04842/) (2026)
- [Backdoors in RLVR: Jailbreak Backdoors in LLMs From Verifiable Reward](https://mlsanity.com/paper/2604.09748/) (2026)
- [Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking](https://mlsanity.com/paper/2605.08778/) (2026)
- [Agent Reinforcement Learning via Pivotal-Aware Self-Feedback Retry](https://mlsanity.com/paper/2607.03702/) (2026)
- [Attention-Aware GNN-based Input Defense against Multi-Turn LLM Jailbreak](https://mlsanity.com/paper/2507.07146/) (2025)
- [Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization](https://mlsanity.com/paper/2605.10067/) (2026)
- [Investigating and Alleviating Harm Amplification in LLM Interactions](https://mlsanity.com/paper/2606.02423/) (2026)
- [Structuring the Space of Sociotechnical Alignment](https://mlsanity.com/paper/2607.01250/) (2026)
- [JMedEthicBench: A Multi-Turn Conversational Benchmark for Evaluating Medical Safety in Japanese Large Language Models](https://mlsanity.com/paper/2601.01627/) (2026)
- [Evaluating Temporal Consistency in Multi-Turn Language Models](https://mlsanity.com/paper/2604.23051/) (2026)
- [PsychJail: Exploring Psychological Jailbreaks via Multi-Turn Persuasion of LLM Policies](https://mlsanity.com/paper/2608.23028/) (2026)
- [Decoupling Safety into Orthogonal Subspace: Cost-Efficient and Performance-Preserving Alignment for Large Language Models](https://mlsanity.com/paper/2510.09004/) (2025)
- [State-Dependent Refusal and Learned Incapacity in RLHF-Aligned Language Models](https://mlsanity.com/paper/2512.13762/) (2025)
- [RoTRAG: Rule of Thumb Reasoning for Conversation Harm Detection with Retrieval-Augmented Generation](https://mlsanity.com/paper/2604.17301/) (2026)
- [One Turn Too Late: Learning When to Intervene Against Multi-Turn Malicious Intent](https://mlsanity.com/paper/2605.05630/) (2026)
- [Mitigating Many-shot Jailbreak Attacks with One Single Demonstration](https://mlsanity.com/paper/2605.08277/) (2026)
- [SafeSpec: Fast and Safe LLM via Dynamic Reflective Sampling](https://mlsanity.com/paper/2606.19755/) (2026)
- [MJ: Multi-turn LLM Jailbreaking via Decomposed Credit Assignment](https://mlsanity.com/paper/2607.11070/) (2026)
- [Robust Critics: Defending LLMs Against Multi-Turn Attacks](https://mlsanity.com/paper/2607.20472/) (2026)
- [AI Alignment through a Game-theoretic Lens: A Survey](https://mlsanity.com/paper/2608.27910/) (2026)
- [CoDeL: Co-Evolutionary Defense against Indirect Prompt Injection in LLM-based Agents](https://mlsanity.com/paper/2609.34463/) (2026)

## Abstract (arXiv)

The proliferation of jailbreak attacks against large language models (LLMs) highlights the need for robust security measures. However, in multi-round dialogues, malicious intentions may be hidden in interactions, leading LLMs to be more prone to produce harmful responses. In this paper, we propose the Multi-Turn Safety Alignment (ourapproach) framework, to address the challenge of securing LLMs in multi-round interactions. It consists of two stages: In the thought-guided attack learning stage, the red-team model learns about thought-guided multi-round jailbreak attacks to generate adversarial prompts. In the adversarial iterative optimization stage, the red-team model and the target model continuously improve their respective capabilities in interaction. Furthermore, we introduce a multi-turn reinforcement learning algorithm based on future rewards to enhance the robustness of safety alignment. Experimental results show that the red-team model exhibits state-of-the-art attack capabilities, while the target model significantly improves its performance on safety benchmarks.