In today's era, where large language models (LLMs) are integrated into numerous real-world applications, ensuring their safety and robustness is crucial for responsible AI usage. Automated red-teaming methods play a key role in this process by generating adversarial attacks to identify and mitigate potential vulnerabilities in these models. However, existing methods often struggle with slow performance, limited categorical diversity, and high resource demands. While Rainbow Teaming, a recent approach, addresses the diversity challenge by framing adversarial prompt generation as a quality-diversity search, it remains slow and requires a large fine-tuned mutator for optimal performance. To overcome these limitations, we propose Ferret, a novel approach that builds upon Rainbow Teaming by generating multiple adversarial prompt mutations per iteration and using a scoring function to rank and select the most effective adversarial prompt. We explore various scoring functions, including reward models, Llama Guard, and LLM-as-a-judge, to rank adversarial mutations based on their potential harm to improve the efficiency of the search for harmful mutations. Our results demonstrate that Ferret, utilizing a reward model as a scoring function, improves the overall attack success rate (ASR) to 95%, which is 46% higher than Rainbow Teaming. Additionally, Ferret reduces the time needed to achieve a 90% ASR by 15.2% compared to the baseline and generates adversarial prompts that are transferable i.e. effective on other LLMs of larger size. Our codes are available at https://github.com/declare-lab/ferret.
Large language models are susceptible to jailbreak attacks, which can result in the generation of harmful content. While prior defenses mitigate these risks by perturbing or inspecting inputs, they ignore competing objectives, the underlying cause of alignment failures. In this paper, we propose Alignment-Enhanced Decoding (AED), a novel defense that employs adaptive decoding to address the root causes of jailbreak issues. We first define the Competitive Index to quantify alignment failures and utilize feedback from self-evaluation to compute post-alignment logits. Then, AED adaptively combines AED and post-alignment logits with the original logits to obtain harmless and helpful distributions. Consequently, our method enhances safety alignment while maintaining helpfulness. We conduct experiments across five models and four common jailbreaks, with the results validating the effectiveness of our approach. Code is available at https://github.com/GIGABaozi/AED.git.
Multimodal Large Language Models (MLLMs) extend the capacity of LLMs to understand multimodal information comprehensively, achieving remarkable performance in many vision-centric tasks. Despite that, recent studies have shown that these models are susceptible to jailbreak attacks, which refer to an exploitative technique where malicious users can break the safety alignment of the target model and generate misleading and harmful answers. This potential threat is caused by both the inherent vulnerabilities of LLM and the larger attack scope introduced by vision input. To enhance the security of MLLMs against jailbreak attacks, researchers have developed various defense techniques. However, these methods either require modifications to the model's internal structure or demand significant computational resources during the inference phase. Multimodal information is a double-edged sword. While it increases the risk of attacks, it also provides additional data that can enhance safeguards. Inspired by this, we propose Cross-modality Information DEtectoR (CIDER), a plug-and-play jailbreaking detector designed to identify maliciously perturbed image inputs, utilizing the cross-modal similarity between harmful queries and adversarial images. CIDER is independent of the target MLLMs and requires less computation cost. Extensive experimental results demonstrate the effectiveness and efficiency of CIDER, as well as its transferability to both white-box and black-box MLLMs.
The risk of harmful content generated by large language models (LLMs) becomes a critical concern. This paper presents a systematic study on assessing and improving LLMs' capability to perform the task of \textbf{course-correction}, \ie, the model can steer away from generating harmful content autonomously. To start with, we introduce the \textsc{C$^2$-Eval} benchmark for quantitative assessment and analyze 10 popular LLMs, revealing varying proficiency of current safety-tuned LLMs in course-correction. To improve, we propose fine-tuning LLMs with preference learning, emphasizing the preference for timely course-correction. Using an automated pipeline, we create \textsc{C$^2$-Syn}, a synthetic dataset with 750K pairwise preferences, to teach models the concept of timely course-correction through data-driven preference learning. Experiments on 2 LLMs, \textsc{Llama2-Chat 7B} and \textsc{Qwen2 7B}, show that our method effectively enhances course-correction skills without affecting general performance. Additionally, it effectively improves LLMs' safety, particularly in resisting jailbreak attacks.
Personality psychologists have analyzed the relationship between personality and safety behaviors in human society. Although Large Language Models (LLMs) demonstrate personality traits, the relationship between personality traits and safety abilities in LLMs still remains a mystery. In this paper, we discover that LLMs' personality traits are closely related to their safety abilities, i.e., toxicity, privacy, and fairness, based on the reliable MBTI-M scale. Meanwhile, the safety alignment generally increases various LLMs' Extraversion, Sensing, and Judging traits. According to such findings, we can edit LLMs' personality traits and improve their safety performance, e.g., inducing personality from ISTJ to ISTP resulted in a relative improvement of approximately 43% and 10% in privacy and fairness performance, respectively. Additionally, we find that LLMs with different personality traits are differentially susceptible to jailbreak. This study pioneers the investigation of LLM safety from a personality perspective, providing new insights into LLM safety enhancement.
Large Language Models (LLMs), trained on a large amount of corpus, have demonstrated remarkable abilities. However, it may not be sufficient to directly apply open-source LLMs like Llama to certain real-world scenarios, since most of them are trained for \emph{general} purposes. Thus, the demands for customizing publicly available LLMs emerge, but are currently under-studied. In this work, we consider customizing pre-trained LLMs with new human preferences. Specifically, the LLM should not only meet the new preference but also preserve its original capabilities after customization. Drawing inspiration from the observation that human preference can be expressed as a reward model, we propose to cast LLM customization as optimizing the sum of two reward functions, one of which (denoted as $r_1$) was used to pre-train the LLM while the other (denoted as $r_2$) characterizes the new human preference. The obstacle here is that both reward functions are unknown, making the application of modern reinforcement learning methods infeasible. Thanks to the residual Q-learning framework, we can restore the customized LLM with the pre-trained LLM and the \emph{residual Q-function} without the reward function $r_1$. Moreover, we find that for a fixed pre-trained LLM, the reward function $r_2$ can be derived from the residual Q-function, enabling us to directly learn the residual Q-function from the new human preference data upon the Bradley-Terry model. We name our method Q-Adapter as it introduces an adapter module to approximate the residual Q-function for customizing the pre-trained LLM towards the new preference. Experiments based on the Llama-3.1 model on the DSP dataset and HH-RLHF dataset illustrate the superior effectiveness of Q-Adapter on both retaining existing knowledge and learning new preferences. Code is available at \url{https://github.com/mansicer/Q-Adapter}.
Large Language Models (LLMs) are known to be vulnerable to jailbreak attacks. An important observation is that, while different types of jailbreak attacks can generate significantly different queries, they mostly result in similar responses that are rooted in the same harmful knowledge (e.g., detailed steps to make a bomb). Consequently, unlearning-based approaches have been proposed to mitigate jailbreak attacks by directly removing harmful knowledge from the model. In this paper, we identify a novel ripple effect of unlearning, wherein LLMs can implicitly unlearn harmful knowledge that was not explicitly introduced during the unlearning phase (e.g., a model unlearning the steps for theft may also implicitly unlearn the steps for making a bomb). Through over 100 experimental runs spanning multiple models, attack strategies, and defense methods, we empirically validate this phenomenon, which makes unlearning-based methods able to decrease the Attack Success Rate on unseen data from more than 70% to less than 10% with only 100 training samples. Further analysis reveals that the strong generalization ability of unlearning may stem from the intrinsic relatedness among harmful responses across harmful questions (e.g., response patterns, shared steps and actions in response, and similarity among their learned representations in the LLM). We also discuss the potential limitations of unlearning and the observed ripple effect. We hope our research could contribute to a deeper understanding of unlearning. Our code is available at https://github.com/thu-coai/SafeUnlearning.
The widespread applications of large language models (LLMs) have brought about concerns regarding their potential misuse. Although aligned with human preference data before release, LLMs remain vulnerable to various malicious attacks. In this paper, we adopt a red-teaming strategy to enhance LLM safety and introduce SeqAR, a simple yet effective framework to design jailbreak prompts automatically. The SeqAR framework generates and optimizes multiple jailbreak characters and then applies sequential jailbreak characters in a single query to bypass the guardrails of the target LLM. Different from previous work which relies on proprietary LLMs or seed jailbreak templates crafted by human expertise, SeqAR can generate and optimize the jailbreak prompt in a cold-start scenario using open-sourced LLMs without any seed jailbreak templates. Experimental results show that SeqAR achieves attack success rates of 88% and 60% in bypassing the safety alignment of GPT-3.5-1106 and GPT-4, respectively. Furthermore, we extensively evaluate the transferability of the generated templates across different LLMs and held-out malicious requests, while also exploring defense strategies against the jailbreak attack designed by SeqAR.
As the development of large language models (LLMs) rapidly advances, securing these models effectively without compromising their utility has become a pivotal area of research. However, current defense strategies against jailbreak attacks (i.e., efforts to bypass security protocols) often suffer from limited adaptability, restricted general capability, and high cost. To address these challenges, we introduce SafeAligner, a methodology implemented at the decoding stage to fortify defenses against jailbreak attacks. We begin by developing two specialized models: the Sentinel Model, which is trained to foster safety, and the Intruder Model, designed to generate riskier responses. SafeAligner leverages the disparity in security levels between the responses from these models to differentiate between harmful and beneficial tokens, effectively guiding the safety alignment by altering the output token distribution of the target model. Extensive experiments show that SafeAligner can increase the likelihood of beneficial tokens, while reducing the occurrence of harmful ones, thereby ensuring secure alignment with minimal loss to generality.
As large language models (LLMs) have advanced rapidly, concerns regarding their safety have become prominent. In this paper, we discover that code-switching in red-teaming queries can effectively elicit undesirable behaviors of LLMs, which are common practices in natural language. We introduce a simple yet effective framework, CSRT, to synthesize code-switching red-teaming queries and investigate the safety and multilingual understanding of LLMs comprehensively. Through extensive experiments with ten state-of-the-art LLMs and code-switching queries combining up to 10 languages, we demonstrate that the CSRT significantly outperforms existing multilingual red-teaming techniques, achieving 46.7% more attacks than standard attacks in English and being effective in conventional safety domains. We also examine the multilingual ability of those LLMs to generate and understand code-switching texts. Additionally, we validate the extensibility of the CSRT by generating code-switching attack prompts with monolingual data. We finally conduct detailed ablation studies exploring code-switching and propound unintended correlation between resource availability of languages and safety alignment in existing multilingual LLMs.
As Artificial General Intelligence (AGI) becomes increasingly integrated into various facets of human life, ensuring the safety and ethical alignment of such systems is paramount. Previous studies primarily focus on single-modality threats, which may not suffice given the integrated and complex nature of cross-modality interactions. We introduce a novel safety alignment challenge called Safe Inputs but Unsafe Output (SIUO) to evaluate cross-modality safety alignment. Specifically, it considers cases where single modalities are safe independently but could potentially lead to unsafe or unethical outputs when combined. To empirically investigate this problem, we developed the SIUO, a cross-modality benchmark encompassing 9 critical safety domains, such as self-harm, illegal activities, and privacy violations. Our findings reveal substantial safety vulnerabilities in both closed- and open-source LVLMs, such as GPT-4V and LLaVA, underscoring the inadequacy of current models to reliably interpret and respond to complex, real-world scenarios.
Large language models (LLMs) are expected to follow instructions from users and engage in conversations. Techniques to enhance LLMs' instruction-following capabilities typically fine-tune them using data structured according to a predefined chat template. Although chat templates are shown to be effective in optimizing LLM performance, their impact on safety alignment of LLMs has been less understood, which is crucial for deploying LLMs safely at scale. In this paper, we investigate how chat templates affect safety alignment of LLMs. We identify a common vulnerability, named ChatBug, that is introduced by chat templates. Our key insight to identify ChatBug is that the chat templates provide a rigid format that need to be followed by LLMs, but not by users. Hence, a malicious user may not necessarily follow the chat template when prompting LLMs. Instead, malicious users could leverage their knowledge of the chat template and accordingly craft their prompts to bypass safety alignments of LLMs. We develop two attacks to exploit the ChatBug vulnerability. We demonstrate that a malicious user can exploit the ChatBug vulnerability of eight state-of-the-art (SOTA) LLMs and effectively elicit unintended responses from these models. Moreover, we show that ChatBug can be exploited by existing jailbreak attacks to enhance their attack success rates. We investigate potential countermeasures to ChatBug. Our results show that while adversarial training effectively mitigates the ChatBug vulnerability, the victim model incurs significant performance degradation. These results highlight the trade-off between safety alignment and helpfulness. Developing new methods for instruction tuning to balance this trade-off is an open and critical direction for future research
To mitigate the risk of harmful outputs from large vision models (LVMs), we introduce the SafeSora dataset to promote research on aligning text-to-video generation with human values. This dataset encompasses human preferences in text-to-video generation tasks along two primary dimensions: helpfulness and harmlessness. To capture in-depth human preferences and facilitate structured reasoning by crowdworkers, we subdivide helpfulness into 4 sub-dimensions and harmlessness into 12 sub-categories, serving as the basis for pilot annotations. The SafeSora dataset includes 14,711 unique prompts, 57,333 unique videos generated by 4 distinct LVMs, and 51,691 pairs of preference annotations labeled by humans. We further demonstrate the utility of the SafeSora dataset through several applications, including training the text-video moderation model and aligning LVMs with human preference by fine-tuning a prompt augmentation module or the diffusion model. These applications highlight its potential as the foundation for text-to-video alignment research, such as human preference modeling and the development and validation of alignment algorithms.
Safety-aligned language models often exhibit fragile and imbalanced safety mechanisms, increasing the likelihood of generating unsafe content. In addition, incorporating new knowledge through editing techniques to language models can further compromise safety. To address these issues, we propose SafeInfer, a context-adaptive, decoding-time safety alignment strategy for generating safe responses to user queries. SafeInfer comprises two phases: the safety amplification phase, which employs safe demonstration examples to adjust the model's hidden states and increase the likelihood of safer outputs, and the safety-guided decoding phase, which influences token selection based on safety-optimized distributions, ensuring the generated content complies with ethical guidelines. Further, we present HarmEval, a novel benchmark for extensive safety evaluations, designed to address potential misuse scenarios in accordance with the policies of leading AI tech giants.
The emergence of Vision Language Models (VLMs) has brought unprecedented advances in understanding multimodal information. The combination of textual and visual semantics in VLMs is highly complex and diverse, making the safety alignment of these models challenging. Furthermore, due to the limited study on the safety alignment of VLMs, there is a lack of large-scale, high-quality datasets. To address these limitations, we propose a Safety Preference Alignment dataset for Vision Language Models named SPA-VL. In terms of breadth, SPA-VL covers 6 harmfulness domains, 13 categories, and 53 subcategories, and contains 100,788 samples of the quadruple (question, image, chosen response, rejected response). In terms of depth, the responses are collected from 12 open-source (e.g., QwenVL) and closed-source (e.g., Gemini) VLMs to ensure diversity. The construction of preference data is fully automated, and the experimental results indicate that models trained with alignment techniques on the SPA-VL dataset exhibit substantial improvements in harmlessness and helpfulness while maintaining core capabilities. SPA-VL, as a large-scale, high-quality, and diverse dataset, represents a significant milestone in ensuring that VLMs achieve both harmlessness and helpfulness.
Ensuring the safe alignment of large language models (LLMs) with human values is critical as they become integral to applications like translation and question answering. Current alignment methods struggle with dynamic user intentions and complex objectives, making models vulnerable to generating harmful content. We propose Safety Arithmetic, a training-free framework enhancing LLM safety across different scenarios: Base models, Supervised fine-tuned models (SFT), and Edited models. Safety Arithmetic involves Harm Direction Removal to avoid harmful content and Safety Alignment to promote safe responses. Additionally, we present NoIntentEdit, a dataset highlighting edit instances that could compromise model safety if used unintentionally. Our experiments show that Safety Arithmetic significantly improves safety measures, reduces over-safety, and maintains model utility, outperforming existing methods in ensuring safe content generation.
Large language models (LLMs) rely on safety alignment to avoid responding to malicious user inputs. Unfortunately, jailbreak can circumvent safety guardrails, resulting in LLMs generating harmful content and raising concerns about LLM safety. Due to language models with intensive parameters often regarded as black boxes, the mechanisms of alignment and jailbreak are challenging to elucidate. In this paper, we employ weak classifiers to explain LLM safety through the intermediate hidden states. We first confirm that LLMs learn ethical concepts during pre-training rather than alignment and can identify malicious and normal inputs in the early layers. Alignment actually associates the early concepts with emotion guesses in the middle layers and then refines them to the specific reject tokens for safe generations. Jailbreak disturbs the transformation of early unethical classification into negative emotions. We conduct experiments on models from 7B to 70B across various model families to prove our conclusion. Overall, our paper indicates the intrinsical mechanism of LLM safety and how jailbreaks circumvent safety guardrails, offering a new perspective on LLM safety and reducing concerns. Our code is available at https://github.com/ydyjya/LLM-IHS-Explanation.
The safety alignment of current Large Language Models (LLMs) is vulnerable. Relatively simple attacks, or even benign fine-tuning, can jailbreak aligned models. We argue that many of these vulnerabilities are related to a shared underlying issue: safety alignment can take shortcuts, wherein the alignment adapts a model's generative distribution primarily over only its very first few output tokens. We refer to this issue as shallow safety alignment. In this paper, we present case studies to explain why shallow safety alignment can exist and provide evidence that current aligned LLMs are subject to this issue. We also show how these findings help explain multiple recently discovered vulnerabilities in LLMs, including the susceptibility to adversarial suffix attacks, prefilling attacks, decoding parameter attacks, and fine-tuning attacks. Importantly, we discuss how this consolidated notion of shallow safety alignment sheds light on promising research directions for mitigating these vulnerabilities. For instance, we show that deepening the safety alignment beyond just the first few tokens can often meaningfully improve robustness against some common exploits. Finally, we design a regularized finetuning objective that makes the safety alignment more persistent against fine-tuning attacks by constraining updates on initial tokens. Overall, we advocate that future safety alignment should be made more than just a few tokens deep.
Large Language Models (LLMs) require careful safety alignment to prevent malicious outputs. While significant research focuses on mitigating harmful content generation, the enhanced safety often come with the side effect of over-refusal, where LLMs may reject innocuous prompts and become less helpful. Although the issue of over-refusal has been empirically observed, a systematic measurement is challenging due to the difficulty of crafting prompts that can elicit the over-refusal behaviors of LLMs. This study proposes a novel method for automatically generating large-scale over-refusal datasets. Leveraging this technique, we introduce OR-Bench, the first large-scale over-refusal benchmark. OR-Bench comprises 80,000 over-refusal prompts across 10 common rejection categories, a subset of around 1,000 hard prompts that are challenging even for state-of-the-art LLMs, and an additional 600 toxic prompts to prevent indiscriminate responses. We then conduct a comprehensive study to measure the over-refusal of 32 popular LLMs across 8 model families. Our datasets are publicly available at https://huggingface.co/bench-llms and our codebase is open-sourced at https://github.com/justincui03/or-bench. We hope this benchmark can help the community develop better safety aligned models.
The growing safety concerns surrounding large language models raise an urgent need to align them with diverse human preferences to simultaneously enhance their helpfulness and safety. A promising approach is to enforce safety constraints through Reinforcement Learning from Human Feedback (RLHF). For such constrained RLHF, typical Lagrangian-based primal-dual policy optimization methods are computationally expensive and often unstable. This paper presents a perspective of dualization that reduces constrained alignment to an equivalent unconstrained alignment problem. We do so by pre-optimizing a smooth and convex dual function that has a closed form. This shortcut eliminates the need for cumbersome primal-dual policy iterations, greatly reducing the computational burden and improving training stability. Our strategy leads to two practical algorithms in model-based and preference-based settings (MoCAN and PeCAN, respectively). A broad range of experiments demonstrate the effectiveness and merits of our algorithms.