# SafePTR: Token-Level Jailbreak Defense in Multimodal LLMs via Prune-then-Restore Mechanism

arXiv: https://arxiv.org/abs/2507.01513 · HTML: https://mlsanity.com/paper/2507.01513/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 20 of its 26 references that are in the map.
- First posted on arXiv 2 Jul 2025.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)
- LLM safety and security / [Vision-language and text-to-image safety](https://mlsanity.com/safety/vision-language-and-text-to-image-safety/)

## Builds on (most cited first, 20 of 20)

- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models](https://mlsanity.com/paper/2306.13394/) (2023)
- [MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities](https://mlsanity.com/paper/2308.02490/) (2023)
- [DeepSeek-VL: Towards Real-World Vision-Language Understanding](https://mlsanity.com/paper/2403.05525/) (2024)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts](https://mlsanity.com/paper/2311.05608/) (2023)
- [Visual Adversarial Examples Jailbreak Aligned Large Language Models](https://mlsanity.com/paper/2306.13213/) (2023)
- [JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks](https://mlsanity.com/paper/2404.03027/) (2024)
- [AdaShield: Safeguarding Multimodal Large Language Models from Structure-based Attack via Adaptive Shield Prompting](https://mlsanity.com/paper/2403.09513/) (2024)
- [Eyes Closed, Safety On: Protecting Multimodal LLMs via Image-to-Text Transformation](https://mlsanity.com/paper/2403.09572/) (2024)
- [MLLM-Protector: Ensuring MLLM's Safety without Hurting Performance](https://mlsanity.com/paper/2401.02906/) (2024)
- [ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time](https://mlsanity.com/paper/2410.06625/) (2024)
- [Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment](https://mlsanity.com/paper/2411.18688/) (2024)
- [CoCA: Regaining Safety-awareness of Multimodal Large Language Models with Constitutional Calibration](https://mlsanity.com/paper/2409.11365/) (2024)
- [The VLLM Safety Paradox: Dual Ease in Jailbreak Attack and Defense](https://mlsanity.com/paper/2411.08410/) (2024)
- [Cross-Modal Safety Mechanism Transfer in Large Vision-Language Models](https://mlsanity.com/paper/2410.12662/) (2024)

## Built on by (most cited first, 10 of 10)

- [TraceRouter: Robust Safety for Large Foundation Models via Path-Level Intervention](https://mlsanity.com/paper/2601.21900/) (2026)
- [SoK: Systematizing LLM Prompt Security: Taxonomies, Datasets, and Unified Evaluation of Attacks and Defenses](https://mlsanity.com/paper/2510.15476/) (2025)
- [Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory](https://mlsanity.com/paper/2603.15800/) (2026)
- [Risk Awareness Injection: Calibrating Vision-Language Models for Safety without Compromising Utility](https://mlsanity.com/paper/2602.03402/) (2026)
- [Feature-Space Smoothing: Certified Robustness of Deep Representations](https://mlsanity.com/paper/2601.16200/) (2026)
- [Localize and Neutralize: Gradient-guided Token Suppression against Visual Prompt Injection Attack](https://mlsanity.com/paper/2605.25194/) (2026)
- [Attack Ensembles Expose a Safety-Utility Trade-off in Black-Box Guard Defenses Against Encoded VLM Jailbreaks](https://mlsanity.com/paper/2607.26574/) (2026)
- [Moving the Safety Barrier: Dynamic Routing Adaptive Alignment Against White-Box Attacks](https://mlsanity.com/paper/2608.02674/) (2026)
- [Evolving Safety Landscape of Multi-modal Large Language Models: A Survey of Emerging Threats and Safeguards](https://mlsanity.com/paper/2608.07535/) (2026)
- [COMIC: Reference-Aware Safety Gating for Multimodal Large Language Models](https://mlsanity.com/paper/2608.17234/) (2026)

## Abstract (arXiv)

By incorporating visual inputs, Multimodal Large Language Models (MLLMs) extend LLMs to support visual reasoning. However, this integration also introduces new vulnerabilities, making MLLMs susceptible to multimodal jailbreak attacks and hindering their safe deployment.Existing defense methods, including Image-to-Text Translation, Safe Prompting, and Multimodal Safety Tuning, attempt to address this by aligning multimodal inputs with LLMs' built-in safeguards.Yet, they fall short in uncovering root causes of multimodal vulnerabilities, particularly how harmful multimodal tokens trigger jailbreak in MLLMs? Consequently, they remain vulnerable to text-driven multimodal jailbreaks, often exhibiting overdefensive behaviors and imposing heavy training overhead.To bridge this gap, we present an comprehensive analysis of where, how and which harmful multimodal tokens bypass safeguards in MLLMs. Surprisingly, we find that less than 1% tokens in early-middle layers are responsible for inducing unsafe behaviors, highlighting the potential of precisely removing a small subset of harmful tokens, without requiring safety tuning, can still effectively improve safety against jailbreaks. Motivated by this, we propose Safe Prune-then-Restore (SafePTR), an training-free defense framework that selectively prunes harmful tokens at vulnerable layers while restoring benign features at subsequent layers.Without incurring additional computational overhead, SafePTR significantly enhances the safety of MLLMs while preserving efficiency. Extensive evaluations across three MLLMs and five benchmarks demonstrate SafePTR's state-of-the-art performance in mitigating jailbreak risks without compromising utility.