The emergence of large language models (LLMs) has accelerated the development of self-driving laboratories (SDLs) for materials research. Despite their transformative potential, current SDL implementations rely on rigid, predefined protocols that limit their adaptability to dynamic experimental scenarios across different labs. A significant challenge persists in measuring how effectively AI agents can replicate the adaptive decision-making and experimental intuition of expert scientists. Here, we introduce AILA (Artificially Intelligent Lab Assistant), a framework that automates atomic force microscopy (AFM) through LLM-driven agents. Using AFM as an experimental testbed, we develop AFMBench-a comprehensive evaluation suite that challenges AI agents based on language models like GPT-4o and GPT-3.5 to perform tasks spanning the scientific workflow: from experimental design to results analysis. Our systematic assessment shows that state-of-the-art language models struggle even with basic tasks such as documentation retrieval, leading to a significant decline in performance in multi-agent coordination scenarios. Further, we observe that LLMs exhibit a tendency to not adhere to instructions or even divagate to additional tasks beyond the original request, raising serious concerns regarding safety alignment aspects of AI agents for SDLs. Finally, we demonstrate the application of AILA on increasingly complex experiments open-ended experiments: automated AFM calibration, high-resolution feature detection, and mechanical property measurement. Our findings emphasize the necessity for stringent benchmarking protocols before deploying AI agents as laboratory assistants across scientific disciplines.
Large Language Models (LLMs) demonstrate impressive zero-shot performance across a wide range of natural language processing tasks. Integrating various modality encoders further expands their capabilities, giving rise to Multimodal Large Language Models (MLLMs) that process not only text but also visual and auditory modality inputs. However, these advanced capabilities may also pose significant security risks, as models can be exploited to generate harmful or inappropriate content through jailbreak attack. While prior work has extensively explored how manipulating textual or visual modality inputs can circumvent safeguards in LLMs and MLLMs, the vulnerability of audio-specific Jailbreak on Large Audio-Language Models (LALMs) remains largely underexplored. To address this gap, we introduce \textbf{Jailbreak-AudioBench}, which consists of the Toolbox, curated Dataset, and comprehensive Benchmark. The Toolbox supports not only text-to-audio conversion but also various editing techniques for injecting audio hidden semantics. The curated Dataset provides diverse explicit and implicit jailbreak audio examples in both original and edited forms. Utilizing this dataset, we evaluate multiple state-of-the-art LALMs and establish the most comprehensive Jailbreak benchmark to date for audio modality. Finally, Jailbreak-AudioBench establishes a foundation for advancing future research on LALMs safety alignment by enabling the in-depth exposure of more powerful jailbreak threats, such as query-based audio editing, and by facilitating the development of effective defense mechanisms.
Ensuring safety alignment is a critical requirement for large language models (LLMs), particularly given increasing deployment in real-world applications. Despite considerable advancements, LLMs remain susceptible to jailbreak attacks, which exploit system vulnerabilities to circumvent safety measures and elicit harmful or inappropriate outputs. Furthermore, while adversarial training-based defense methods have shown promise, a prevalent issue is the unintended over-defense behavior, wherein models excessively reject benign queries, significantly undermining their practical utility. To address these limitations, we introduce LATPC, a Latent-space Adversarial Training with Post-aware Calibration framework. LATPC dynamically identifies safety-critical latent dimensions by contrasting harmful and benign inputs, enabling the adaptive construction of targeted refusal feature removal attacks. This mechanism allows adversarial training to concentrate on real-world jailbreak tactics that disguise harmful queries as benign ones. During inference, LATPC employs an efficient embedding-level calibration mechanism to minimize over-defense behaviors with negligible computational overhead. Experimental results across five types of disguise-based jailbreak attacks demonstrate that LATPC achieves a superior balance between safety and utility compared to existing defense frameworks. Further analysis demonstrates the effectiveness of leveraging safety-critical dimensions in developing robust defense methods against jailbreak attacks.
Text-to-image (T2I) models have been shown to be vulnerable to misuse, particularly in generating not-safe-for-work (NSFW) content, raising serious ethical concerns. In this work, we present PromptGuard, a novel content moderation technique that draws inspiration from the system prompt mechanism in large language models (LLMs) for safety alignment. Unlike LLMs, T2I models lack a direct interface for enforcing behavioral guidelines. Our key idea is to optimize a safety soft prompt that functions as an implicit system prompt within the T2I model's textual embedding space. This universal soft prompt (P*) directly moderates NSFW inputs, enabling safe yet realistic image generation without altering the inference efficiency or requiring proxy models. Extensive experiments across three datasets demonstrate that PromptGuard effectively mitigates NSFW content generation while preserving high-quality benign outputs. PromptGuard achieves 7.8 times faster than prior content moderation methods, surpassing eight state-of-the-art defenses with an optimal unsafe ratio down to 5.84%.
As advancements in large language models (LLMs) continue and the demand for personalized models increases, parameter-efficient fine-tuning (PEFT) methods (e.g., LoRA) will become essential due to their efficiency in reducing computation costs. However, recent studies have raised alarming concerns that LoRA fine-tuning could potentially compromise the safety alignment in LLMs, posing significant risks for the model owner. In this paper, we first investigate the underlying mechanism by analyzing the changes in safety alignment related features before and after fine-tuning. Then, we propose a fixed safety module calculated by safety data and a task-specific initialization for trainable parameters in low-rank adaptations, termed Safety-alignment preserved Low-Rank Adaptation (SaLoRA). Unlike previous LoRA methods and their variants, SaLoRA enables targeted modifications to LLMs without disrupting their original alignments. Our experiments show that SaLoRA outperforms various adapters-based approaches across various evaluation metrics in different fine-tuning tasks.
Large language models (LLMs) have made significant advancements across various tasks, but their safety alignment remain a major concern. Exploring jailbreak prompts can expose LLMs' vulnerabilities and guide efforts to secure them. Existing methods primarily design sophisticated instructions for the LLM to follow, or rely on multiple iterations, which could hinder the performance and efficiency of jailbreaks. In this work, we propose a novel jailbreak paradigm, Simple Assistive Task Linkage (SATA), which can effectively circumvent LLM safeguards and elicit harmful responses. Specifically, SATA first masks harmful keywords within a malicious query to generate a relatively benign query containing one or multiple [MASK] special tokens. It then employs a simple assistive task such as a masked language model task or an element lookup by position task to encode the semantics of the masked keywords. Finally, SATA links the assistive task with the masked query to jointly perform the jailbreak. Extensive experiments show that SATA achieves state-of-the-art performance and outperforms baselines by a large margin. Specifically, on AdvBench dataset, with mask language model (MLM) assistive task, SATA achieves an overall attack success rate (ASR) of 85% and harmful score (HS) of 4.57, and with element lookup by position (ELP) assistive task, SATA attains an overall ASR of 76% and HS of 4.43.
The security of Large Language Models (LLMs) has become an important research topic since the emergence of ChatGPT. Though there have been various effective methods to defend against jailbreak attacks, prefilling attacks remain an unsolved and popular threat against open-sourced LLMs. In-Context Learning (ICL) offers a computationally efficient defense against various jailbreak attacks, yet no effective ICL methods have been developed to counter prefilling attacks. In this paper, we: (1) show that ICL can effectively defend against prefilling jailbreak attacks by employing adversative sentence structures within demonstrations; (2) characterize the effectiveness of this defense through the lens of model size, number of demonstrations, over-defense, integration with other jailbreak attacks, and the presence of safety alignment. Given the experimental results and our analysis, we conclude that there is no free lunch for defending against prefilling jailbreak attacks with ICL. On the one hand, current safety alignment methods fail to mitigate prefilling jailbreak attacks, but adversative structures within ICL demonstrations provide robust defense across various model sizes and complex jailbreak attacks. On the other hand, LLMs exhibit similar over-defensiveness when utilizing ICL demonstrations with adversative structures, and this behavior appears to be independent of model size.
Text-to-image (T2I) models have become widespread, but their limited safety guardrails expose end users to harmful content and potentially allow for model misuse. Current safety measures are typically limited to text-based filtering or concept removal strategies, able to remove just a few concepts from the model's generative capabilities. In this work, we introduce SafetyDPO, a method for safety alignment of T2I models through Direct Preference Optimization (DPO). We enable the application of DPO for safety purposes in T2I models by synthetically generating a dataset of harmful and safe image-text pairs, which we call CoProV2. Using a custom DPO strategy and this dataset, we train safety experts, in the form of low-rank adaptation (LoRA) matrices, able to guide the generation process away from specific safety-related concepts. Then, we merge the experts into a single LoRA using a novel merging strategy for optimal scaling performance. This expert-based approach enables scalability, allowing us to remove 7 times more harmful concepts from T2I models compared to baselines. SafetyDPO consistently outperforms the state-of-the-art on many benchmarks and establishes new practices for safety alignment in T2I networks. Code and data will be shared at https://safetydpo.github.io/.
Although large language models (LLMs) achieve effective safety alignment at the time of release, they still face various safety challenges. A key issue is that fine-tuning often compromises the safety alignment of LLMs. To address this issue, we propose a method named \textbf{IRR} (\textbf{I}dentify, \textbf{R}emove, and \textbf{R}ecalibrate for Safety Realignment) that performs safety realignment for LLMs. The core of IRR is to identify and remove unsafe delta parameters from the fine-tuned models, while recalibrating the retained ones. We evaluate the effectiveness of IRR across various datasets, including both full fine-tuning and LoRA methods. Our results demonstrate that IRR significantly enhances the safety performance of fine-tuned models on safety benchmarks, such as harmful queries and jailbreak attacks, while maintaining their performance on downstream tasks. The source code is available at: \url{https://anonymous.4open.science/r/IRR-BD4F}.
Fine-tuning text-to-image diffusion models is widely used for personalization and adaptation for new domains. In this paper, we identify a critical vulnerability of fine-tuning: safety alignment methods designed to filter harmful content (e.g., nudity) can break down during fine-tuning, allowing previously suppressed content to resurface, even when using benign datasets. While this "fine-tuning jailbreaking" issue is known in large language models, it remains largely unexplored in text-to-image diffusion models. Our investigation reveals that standard fine-tuning can inadvertently undo safety measures, causing models to relearn harmful concepts that were previously removed and even exacerbate harmful behaviors. To address this issue, we present a novel but immediate solution called Modular LoRA, which involves training Safety Low-Rank Adaptation (LoRA) modules separately from Fine-Tuning LoRA components and merging them during inference. This method effectively prevents the re-learning of harmful content without compromising the model's performance on new tasks. Our experiments demonstrate that Modular LoRA outperforms traditional fine-tuning methods in maintaining safety alignment, offering a practical approach for enhancing the security of text-to-image diffusion models against potential attacks.
Large Language Models (LLMs) have transformed numerous fields by enabling advanced natural language interactions but remain susceptible to critical vulnerabilities, particularly jailbreak attacks. Current jailbreak techniques, while effective, often depend on input modifications, making them detectable and limiting their stealth and scalability. This paper presents Targeted Model Editing (TME), a novel white-box approach that bypasses safety filters by minimally altering internal model structures while preserving the model's intended functionalities. TME identifies and removes safety-critical transformations (SCTs) embedded in model matrices, enabling malicious queries to bypass restrictions without input modifications. By analyzing distinct activation patterns between safe and unsafe queries, TME isolates and approximates SCTs through an optimization process. Implemented in the D-LLM framework, our method achieves an average Attack Success Rate (ASR) of 84.86% on four mainstream open-source LLMs, maintaining high performance. Unlike existing methods, D-LLM eliminates the need for specific triggers or harmful response collections, offering a stealthier and more effective jailbreak strategy. This work reveals a covert and robust threat vector in LLM security and emphasizes the need for stronger safeguards in model safety alignment.
Recent studies have discovered that LLMs have serious privacy leakage concerns, where an LLM may be fooled into outputting private information under carefully crafted adversarial prompts. These risks include leaking system prompts, personally identifiable information, training data, and model parameters. Most existing red-teaming approaches for privacy leakage rely on humans to craft the adversarial prompts. A few automated methods are proposed for system prompt extraction, but they cannot be applied to more severe risks (e.g., training data extraction) and have limited effectiveness even for system prompt extraction. In this paper, we propose PrivAgent, a novel black-box red-teaming framework for LLM privacy leakage. We formulate different risks as a search problem with a unified attack goal. Our framework trains an open-source LLM through reinforcement learning as the attack agent to generate adversarial prompts for different target models under different risks. We propose a novel reward function to provide effective and fine-grained rewards for the attack agent. Finally, we introduce customizations to better fit our general framework to system prompt extraction and training data extraction. Through extensive evaluations, we first show that PrivAgent outperforms existing automated methods in system prompt leakage against six popular LLMs. Notably, our approach achieves a 100% success rate in extracting system prompts from real-world applications in OpenAI's GPT Store. We also show PrivAgent's effectiveness in extracting training data from an open-source LLM with a success rate of 5.9%. We further demonstrate PrivAgent's effectiveness in evading the existing guardrail defense and its helpfulness in enabling better safety alignment. Finally, we validate our customized designs through a detailed ablation study. We release our code here https://github.com/rucnyz/RedAgent.
In the rapidly evolving field of Large Language Models (LLMs), ensuring safety is a crucial and widely discussed topic. However, existing works often overlook the geo-diversity of cultural and legal standards across the world. To demonstrate the challenges posed by geo-diverse safety standards, we introduce SafeWorld, a novel benchmark specifically designed to evaluate LLMs' ability to generate responses that are not only helpful but also culturally sensitive and legally compliant across diverse global contexts. SafeWorld encompasses 2,342 test user queries, each grounded in high-quality, human-verified cultural norms and legal policies from 50 countries and 493 regions/races. On top of it, we propose a multi-dimensional automatic safety evaluation framework that assesses the contextual appropriateness, accuracy, and comprehensiveness of responses. Our evaluations reveal that current LLMs struggle to meet these criteria. To enhance LLMs' alignment with geo-diverse safety standards, we synthesize helpful preference pairs for Direct Preference Optimization (DPO) alignment training. The preference pair construction aims to encourage LLMs to behave appropriately and provide precise references to relevant cultural norms and policies when necessary. Our trained SafeWorldLM outperforms all competing models, including GPT-4o on all three evaluation dimensions by a large margin. Global human evaluators also note a nearly 20% higher winning rate in helpfulness and harmfulness evaluation. Our code and data can be found here: https://github.com/PlusLabNLP/SafeWorld.
Federated Parameter-Efficient Fine-Tuning (FedPEFT) has emerged as a promising paradigm for privacy-preserving and efficient adaptation of Pre-trained Language Models (PLMs) in Federated Learning (FL) settings. It preserves data privacy by keeping the data decentralized and training the model on local devices, ensuring that raw data never leaves the user's device. Moreover, the integration of PEFT methods such as LoRA significantly reduces the number of trainable parameters compared to fine-tuning the entire model, thereby minimizing communication costs and computational overhead. Despite its potential, the security implications of FedPEFT remain underexplored. This paper introduces a novel security threat to FedPEFT, termed PEFT-as-an-Attack (PaaA), which exposes how PEFT can be exploited as an attack vector to circumvent PLMs' safety alignment and generate harmful content in response to malicious prompts. Our evaluation of PaaA reveals that with less than 1% of the model's parameters set as trainable, and a small subset of clients acting maliciously, the attack achieves an approximate 80% attack success rate using representative PEFT methods such as LoRA. To mitigate this threat, we further investigate potential defense strategies, including Robust Aggregation Schemes (RASs) and Post-PEFT Safety Alignment (PPSA). However, our empirical analysis highlights the limitations of these defenses, i.e., even the most advanced RASs, such as DnC and ClippedClustering, struggle to defend against PaaA in scenarios with highly heterogeneous data distributions. Similarly, while PPSA can reduce attack success rates to below 10%, it severely degrades the model's accuracy on the target task. Our results underscore the urgent need for more effective defense mechanisms that simultaneously ensure security and maintain the performance of the FedPEFT paradigm.
With the widespread deployment of Multimodal Large Language Models (MLLMs) for visual-reasoning tasks, improving their safety has become crucial. Recent research indicates that despite training-time safety alignment, these models remain vulnerable to jailbreak attacks. In this work, we first highlight an important safety gap to describe that alignment achieved solely through safety training may be insufficient against jailbreak attacks. To address this vulnerability, we propose Immune, an inference-time defense framework that leverages a safe reward model through controlled decoding to defend against jailbreak attacks. Additionally, we provide a mathematical characterization of Immune, offering insights on why it improves safety against jailbreaks. Extensive evaluations on diverse jailbreak benchmarks using recent MLLMs reveal that Immune effectively enhances model safety while preserving the model's original capabilities. For instance, against text-based jailbreak attacks on LLaVA-1.6, Immune reduces the attack success rate by 57.82% and 16.78% compared to the base MLLM and state-of-the-art defense strategy, respectively.
Despite inheriting security measures from underlying language models, Vision-Language Models (VLMs) may still be vulnerable to safety alignment issues. Through empirical analysis, we uncover two critical findings: scenario-matched images can significantly amplify harmful outputs, and contrary to common assumptions in gradient-based attacks, minimal loss values do not guarantee optimal attack effectiveness. Building on these insights, we introduce MLAI (Multi-Loss Adversarial Images), a novel jailbreak framework that leverages scenario-aware image generation for semantic alignment, exploits flat minima theory for robust adversarial image selection, and employs multi-image collaborative attacks for enhanced effectiveness. Extensive experiments demonstrate MLAI's significant impact, achieving attack success rates of 77.75% on MiniGPT-4 and 82.80% on LLaVA-2, substantially outperforming existing methods by margins of 34.37% and 12.77% respectively. Furthermore, MLAI shows considerable transferability to commercial black-box VLMs, achieving up to 60.11% success rate. Our work reveals fundamental visual vulnerabilities in current VLMs safety mechanisms and underscores the need for stronger defenses. Warning: This paper contains potentially harmful example text.
The o1 system card identifies the o1 models as the most robust within OpenAI, with their defining characteristic being the progression from rapid, intuitive thinking to slower, more deliberate reasoning. This observation motivated us to investigate the influence of System-2 thinking patterns on model safety. In our preliminary research, we conducted safety evaluations of the o1 model, including complex jailbreak attack scenarios using adversarial natural language prompts and mathematical encoding prompts. Our findings indicate that the o1 model demonstrates relatively improved safety performance; however, it still exhibits vulnerabilities, particularly against jailbreak attacks employing mathematical encoding. Through detailed case analysis, we identified specific patterns in the o1 model's responses. We also explored the alignment of System-2 safety in open-source models using prompt engineering and supervised fine-tuning techniques. Experimental results show that some simple methods to encourage the model to carefully scrutinize user requests are beneficial for model safety. Additionally, we proposed a implementation plan for process supervision to enhance safety alignment. The implementation details and experimental results will be provided in future versions.
The remarkable capabilities of Large Language Models (LLMs) make them increasingly compelling for adoption in real-world healthcare applications. However, the risks associated with using LLMs in medical applications have not been systematically characterized. We propose using five key principles for safe and trustworthy medical AI: Truthfulness, Resilience, Fairness, Robustness, and Privacy, along with ten specific aspects. Under this comprehensive framework, we introduce a novel MedGuard benchmark with 1,000 expert-verified questions. Our evaluation of 11 commonly used LLMs shows that the current language models, regardless of their safety alignment mechanisms, generally perform poorly on most of our benchmarks, particularly when compared to the high performance of human physicians. Despite recent reports indicate that advanced LLMs like ChatGPT can match or even exceed human performance in various medical tasks, this study underscores a significant safety gap, highlighting the crucial need for human oversight and the implementation of AI safety guardrails.
The advent of large language models (LLMs) has spurred the development of numerous jailbreak techniques aimed at circumventing their security defenses against malicious attacks. An effective jailbreak approach is to identify a domain where safety generalization fails, a phenomenon known as mismatched generalization. In this paper, we introduce two novel jailbreak methods based on mismatched generalization: natural language games and custom language games, both of which effectively bypass the safety mechanisms of LLMs, with various kinds and different variants, making them hard to defend and leading to high attack rates. Natural language games involve the use of synthetic linguistic constructs and the actions intertwined with these constructs, such as the Ubbi Dubbi language. Building on this phenomenon, we propose the custom language games method: by engaging with LLMs using a variety of custom rules, we successfully execute jailbreak attacks across multiple LLM platforms. Extensive experiments demonstrate the effectiveness of our methods, achieving success rates of 93% on GPT-4o, 89% on GPT-4o-mini and 83% on Claude-3.5-Sonnet. Furthermore, to investigate the generalizability of safety alignments, we fine-tuned Llama-3.1-70B with the custom language games to achieve safety alignment within our datasets and found that when interacting through other language games, the fine-tuned models still failed to identify harmful content. This finding indicates that the safety alignment knowledge embedded in LLMs fails to generalize across different linguistic formats, thus opening new avenues for future research in this area.
Benefiting from the powerful capabilities of Large Language Models (LLMs), pre-trained visual encoder models connected to LLMs form Vision Language Models (VLMs). However, recent research shows that the visual modality in VLMs is highly vulnerable, allowing attackers to bypass safety alignment in LLMs through visually transmitted content, launching harmful attacks. To address this challenge, we propose a progressive concept-based alignment strategy, PSA-VLM, which incorporates safety modules as concept bottlenecks to enhance visual modality safety alignment. By aligning model predictions with specific safety concepts, we improve defenses against risky images, enhancing explainability and controllability while minimally impacting general performance. Our method is obtained through two-stage training. The low computational cost of the first stage brings very effective performance improvement, and the fine-tuning of the language model in the second stage further improves the safety performance. Our method achieves state-of-the-art results on popular VLM safety benchmark.