Daily Papers

Most existing LLM safety evaluation and defense methods follow a static formulation: jailbreak vulnerabilities are evaluated with fixed attack methods, and guardrails are trained on fixed malicious prompt datasets. However, real-world adversaries continuously evolve their capabilities and expand the attack space. To address this challenge, we propose DARWIN, an evolutionary attack-defense framework that formulates jailbreaking as an open-ended evolution process and continuously updates guardrails through an evolving attack-defense loop. DARWIN-Attack is an evolutionary adversary that expands its capabilities through strategy discovery, mutation, selection, and feedback-driven composition. It collects strategies from broad external sources, generates new variants through self-reflection and genetic evolution, and retains effective strategies based on their performance against aligned LLMs. During attack execution, DARWIN-Attack adaptively selects and combines evolved strategies according to feedback from target LLMs and guardrails. Across frontier models and guardrails, it achieves state-of-the-art attack success rates, including nearly 100% on DeepSeek-V4-Pro and YuFeng-XGuard and over 90% on GPT-5.5. On the defense side, we introduce DARWIN-Guard, an online adversarial training paradigm that iteratively learns from emerging adversarial samples generated by DARWIN-Attack. To improve robustness without sacrificing utility, DARWIN-Guard jointly trains on malicious and benign disguised queries, encouraging the model to identify underlying intent rather than superficial attack patterns. DARWIN-Guard achieves an average unsafe recall of 91.6% across 12 safety benchmarks, outperforming strong guardrails such as YuFeng-XGuard and Nemotron Guard, while maintaining a nearly 100% pass rate on standard benign datasets.

While the deployment of large language models (LLMs) in high-value industries continues to expand, the systematic assessment of their safety against jailbreak and prompt-based attacks remains insufficient. Existing safety evaluation benchmarks and frameworks are often limited by an imbalanced integration of core components (attack, defense, and evaluation methods) and an isolation between flexible evaluation frameworks and standardized benchmarking capabilities. These limitations hinder reliable cross-study comparisons and create unnecessary overhead for comprehensive risk assessment. To address these gaps, we present TeleAI-Safety, a modular and reproducible framework coupled with a systematic benchmark for rigorous LLM safety evaluation. Our framework integrates a broad collection of 19 attack methods (including one self-developed method), 29 defense methods, and 19 evaluation methods (including one self-developed method). With a curated attack corpus of 342 samples spanning 12 distinct risk categories, the TeleAI-Safety benchmark conducts extensive evaluations across 14 target models. The results reveal systematic vulnerabilities and model-specific failure cases, highlighting critical trade-offs between safety and utility, and identifying potential defense patterns for future optimization. In practical scenarios, TeleAI-Safety can be flexibly adjusted with customized attack, defense, and evaluation combinations to meet specific demands. We release our complete code and evaluation results to facilitate reproducible research and establish unified safety baselines.

Current evaluations of LLM safety predominantly rely on severity-based taxonomies to assess the harmfulness of malicious queries. We argue that this formulation requires re-examination as it assumes uniform risk across all malicious queries, neglecting Execution Likelihood--the conditional probability of a threat being realized given the model's response. In this work, we introduce Expected Harm, a metric that weights the severity of a jailbreak by its execution likelihood, modeled as a function of execution cost. Through empirical analysis of state-of-the-art models, we reveal a systematic Inverse Risk Calibration: models disproportionately exhibit stronger refusal behaviors for low-likelihood (high-cost) threats while remaining vulnerable to high-likelihood (low-cost) queries. We demonstrate that this miscalibration creates a structural vulnerability: by exploiting this property, we increase the attack success rate of existing jailbreaks by up to 2times. Finally, we trace the root cause of this failure using linear probing, which reveals that while models encode severity in their latent space to drive refusal decisions, they possess no distinguishable internal representation of execution cost, making them "blind" to this critical dimension of risk.

Large Language Model (LLM) safety has often been evaluated at the behavior level, which provides limited evidence of internal robustness, as these evaluations target outputs rather than representation-level vulnerability under intervention. We formalize this discrepancy as the audit gap: the difference between behavioral safety and robustness under intervention. To study this gap, we construct dissociated models that preserve safe outward behavior while remaining vulnerable in the latent space. We introduce an intervention-based evaluation framework to test model robustness through soft interventions in parameter and latent spaces, including harmful fine-tuning and layer-wise latent perturbations. To formalize the evaluation, we propose the Latent Vulnerability Score (LVS) to measure how easily harmful behavior can be elicited by bounded latent perturbations. Using this evaluation framework, we show that behavioral safety metrics are insufficient measures of representation-level robustness across multiple safely and unsafely aligned state-of-the-art models. Notably, dissociated models show substantially elevated LVSs despite comparable refusal behavior under harmful intervention, with intermediate representations being the most sensitive to intervention. Our results suggest that behavioral safety evaluation alone provides an incomplete picture of model robustness, motivating representation-aware audits of latent vulnerability and observable behavior.

This paper analyzes the safety of Large Language Models (LLMs) in interactions with children below age of 18 years. Despite the transformative applications of LLMs in various aspects of children's lives such as education and therapy, there remains a significant gap in understanding and mitigating potential content harms specific to this demographic. The study acknowledges the diverse nature of children often overlooked by standard safety evaluations and proposes a comprehensive approach to evaluating LLM safety specifically for children. We list down potential risks that children may encounter when using LLM powered applications. Additionally we develop Child User Models that reflect the varied personalities and interests of children informed by literature in child care and psychology. These user models aim to bridge the existing gap in child safety literature across various fields. We utilize Child User Models to evaluate the safety of six state of the art LLMs. Our observations reveal significant safety gaps in LLMs particularly in categories harmful to children but not adults

As the use of large language model (LLM) agents continues to grow, their safety vulnerabilities have become increasingly evident. Extensive benchmarks evaluate various aspects of LLM safety by defining the safety relying heavily on general standards, overlooking user-specific standards. However, safety standards for LLM may vary based on a user-specific profiles rather than being universally consistent across all users. This raises a critical research question: Do LLM agents act safely when considering user-specific safety standards? Despite its importance for safe LLM use, no benchmark datasets currently exist to evaluate the user-specific safety of LLMs. To address this gap, we introduce U-SAFEBENCH, the first benchmark designed to assess user-specific aspect of LLM safety. Our evaluation of 18 widely used LLMs reveals current LLMs fail to act safely when considering user-specific safety standards, marking a new discovery in this field. To address this vulnerability, we propose a simple remedy based on chain-of-thought, demonstrating its effectiveness in improving user-specific safety. Our benchmark and code are available at https://github.com/yeonjun-in/U-SafeBench.

Large language models (LLMs) are known to have the potential to generate harmful content, posing risks to users. While significant progress has been made in developing taxonomies for LLM risks and safety evaluation prompts, most studies have focused on monolingual contexts, primarily in English. However, language- and region-specific risks in bilingual contexts are often overlooked, and core findings can diverge from those in monolingual settings. In this paper, we introduce Qorgau, a novel dataset specifically designed for safety evaluation in Kazakh and Russian, reflecting the unique bilingual context in Kazakhstan, where both Kazakh (a low-resource language) and Russian (a high-resource language) are spoken. Experiments with both multilingual and language-specific LLMs reveal notable differences in safety performance, emphasizing the need for tailored, region-specific datasets to ensure the responsible and safe deployment of LLMs in countries like Kazakhstan. Warning: this paper contains example data that may be offensive, harmful, or biased.

The rapid proliferation of large language models (LLMs) has intensified the requirement for reliable safety evaluation to uncover model vulnerabilities. To this end, numerous LLM safety evaluation benchmarks are proposed. However, existing benchmarks generally rely on labor-intensive manual curation, which causes excessive time and resource consumption. They also exhibit significant redundancy and limited difficulty. To alleviate these problems, we introduce SafetyFlow, the first agent-flow system designed to automate the construction of LLM safety benchmarks. SafetyFlow can automatically build a comprehensive safety benchmark in only four days without any human intervention by orchestrating seven specialized agents, significantly reducing time and resource cost. Equipped with versatile tools, the agents of SafetyFlow ensure process and cost controllability while integrating human expertise into the automatic pipeline. The final constructed dataset, SafetyFlowBench, contains 23,446 queries with low redundancy and strong discriminative power. Our contribution includes the first fully automated benchmarking pipeline and a comprehensive safety benchmark. We evaluate the safety of 49 advanced LLMs on our dataset and conduct extensive experiments to validate our efficacy and efficiency.

This paper presents a comprehensive analysis of the linguistic diversity of LLM safety research, highlighting the English-centric nature of the field. Through a systematic review of nearly 300 publications from 2020--2024 across major NLP conferences and workshops at *ACL, we identify a significant and growing language gap in LLM safety research, with even high-resource non-English languages receiving minimal attention. We further observe that non-English languages are rarely studied as a standalone language and that English safety research exhibits poor language documentation practice. To motivate future research into multilingual safety, we make several recommendations based on our survey, and we then pose three concrete future directions on safety evaluation, training data generation, and crosslingual safety generalization. Based on our survey and proposed directions, the field can develop more robust, inclusive AI safety practices for diverse global populations.

We introduce Synthetic Alignment data Generation for Safety Evaluation and Red Teaming (SAGE-RT or SAGE) a novel pipeline for generating synthetic alignment and red-teaming data. Existing methods fall short in creating nuanced and diverse datasets, providing necessary control over the data generation and validation processes, or require large amount of manually generated seed data. SAGE addresses these limitations by using a detailed taxonomy to produce safety-alignment and red-teaming data across a wide range of topics. We generated 51,000 diverse and in-depth prompt-response pairs, encompassing over 1,500 topics of harmfulness and covering variations of the most frequent types of jailbreaking prompts faced by large language models (LLMs). We show that the red-teaming data generated through SAGE jailbreaks state-of-the-art LLMs in more than 27 out of 32 sub-categories, and in more than 58 out of 279 leaf-categories (sub-sub categories). The attack success rate for GPT-4o, GPT-3.5-turbo is 100% over the sub-categories of harmfulness. Our approach avoids the pitfalls of synthetic safety-training data generation such as mode collapse and lack of nuance in the generation pipeline by ensuring a detailed coverage of harmful topics using iterative expansion of the topics and conditioning the outputs on the generated raw-text. This method can be used to generate red-teaming and alignment data for LLM Safety completely synthetically to make LLMs safer or for red-teaming the models over a diverse range of topics.

We present MultiBreak, a scalable and diverse multi-turn jailbreak benchmark to evaluate large language model (LLM) safety. Multi-turn jailbreaks mimic natural conversational settings, making them easier to bypass safety-aligned LLM than single-turn jailbreaks. Existing multi-turn benchmarks are limited in size or rely heavily on templates, which restrict their diversity. To address this gap, we unify a wide range of harmful jailbreak intents, and introduce an active learning pipeline for expanding high-quality multi-turn adversarial prompts, where a generator is iteratively fine-tuned to produce stronger attack candidates, guided by uncertainty-based refinement. Our MultiBreak includes 10,389 multi-turn adversarial prompts, spans 2,665 distinct harmful intents, and covers the most diverse set of topics to date. Empirical evaluation shows that our benchmark achieves up to a 54.0 and 34.6 higher attack success rate (ASR)} than the second-best dataset on DeepSeek-R1-7B and GPT-4.1-mini, respectively. More importantly, safety evaluations suggest that diverse attack categories uncover fine-grained LLM vulnerabilities}, and categories that appear benign under single-turn can exhibit substantially higher adversarial effectiveness in multi-turn scenarios. These findings highlight persistent vulnerabilities of LLMs under realistic adversarial settings and establish MultiBreak as a scalable resource for advancing LLM safety.

Current LLM safety benchmarks are predominantly English-centric and often rely on translation, failing to capture country-specific harms. Moreover, they rarely evaluate a model's ability to detect culturally embedded sensitivities as distinct from universal harms. We introduce XL-SafetyBench. a suite of 5,500 test cases across 10 country-language pairs, comprising a Jailbreak Benchmark of country-grounded adversarial prompts and a Cultural Benchmark where local sensitivities are embedded within innocuous requests. Each item is constructed via a multi-stage pipeline that combines LLM-assisted discovery, automated validation gates, and dual independent native-speaker annotators per country. To distinguish principled refusal from comprehension failure, we evaluate Attack Success Rate (ASR) alongside two complementary metrics we introduce: Neutral-Safe Rate (NSR) and Cultural Sensitivity Rate (CSR). Evaluating 10 frontier and 27 local LLMs reveals two key findings. First, jailbreak robustness and cultural awareness do not show a coupled relationship among frontier models, so a composite safety score obscures per-axis variation. Second, local models exhibit a near-linear ASR-NSR trade-off (r = -0.81), indicating that their apparent safety reflects generation failure rather than genuine alignment. XL-SafetyBench enables more nuanced, cross-cultural safety evaluation in the multilingual era.

LLM safety evaluations predominantly test models in isolation, yet deployed AI agents increasingly operate within persistent social environments alongside other agents. We introduce a Moltbook-style simulation platform where thousands of LLM agents interact across communities over a simulated month, and use it to evaluate privacy as a downstream safety concern under varying degrees of social pressure. We find that shifting from single turn to multi turn social evaluation amplifies privacy violations (CIMemories 19.95% to Ours 45.30% across OpenAI models), that leakage is socially contagious, with agents 8 times more likely to disclose sensitive information after observing a peer do so, and that explicit privacy instructions reduce but do not eliminate this effect, leaving leakage rates above 37.8% even with safeguards. Our findings suggest that static chat based safety benchmarks systematically underestimate risks in agentic deployment, and that social context alone is sufficient to elicit sensitive disclosures that single turn evaluations would never surface.

As large language models (LLMs) become increasingly prevalent, ensuring their robustness against adversarial misuse is crucial. This paper introduces the GAP (Graph of Attacks with Pruning) framework, an advanced approach for generating stealthy jailbreak prompts to evaluate and enhance LLM safeguards. GAP addresses limitations in existing tree-based LLM jailbreak methods by implementing an interconnected graph structure that enables knowledge sharing across attack paths. Our experimental evaluation demonstrates GAP's superiority over existing techniques, achieving a 20.8% increase in attack success rates while reducing query costs by 62.7%. GAP consistently outperforms state-of-the-art methods for attacking both open and closed LLMs, with attack success rates of >96%. Additionally, we present specialized variants like GAP-Auto for automated seed generation and GAP-VLM for multimodal attacks. GAP-generated prompts prove highly effective in improving content moderation systems, increasing true positive detection rates by 108.5% and accuracy by 183.6% when used for fine-tuning. Our implementation is available at https://github.com/dsbuddy/GAP-LLM-Safety.

In the rapidly evolving landscape of Large Language Models (LLMs), ensuring robust safety measures is paramount. To meet this crucial need, we propose SALAD-Bench, a safety benchmark specifically designed for evaluating LLMs, attack, and defense methods. Distinguished by its breadth, SALAD-Bench transcends conventional benchmarks through its large scale, rich diversity, intricate taxonomy spanning three levels, and versatile functionalities.SALAD-Bench is crafted with a meticulous array of questions, from standard queries to complex ones enriched with attack, defense modifications and multiple-choice. To effectively manage the inherent complexity, we introduce an innovative evaluators: the LLM-based MD-Judge for QA pairs with a particular focus on attack-enhanced queries, ensuring a seamless, and reliable evaluation. Above components extend SALAD-Bench from standard LLM safety evaluation to both LLM attack and defense methods evaluation, ensuring the joint-purpose utility. Our extensive experiments shed light on the resilience of LLMs against emerging threats and the efficacy of contemporary defense tactics. Data and evaluator are released under https://github.com/OpenSafetyLab/SALAD-BENCH. Warning: this paper includes examples that may be offensive or harmful.

Large Language Models (LLMs) have showcased remarkable capabilities across various domains. Accompanying the evolving capabilities and expanding deployment scenarios of LLMs, their deployment challenges escalate due to their sheer scale and the advanced yet complex activation designs prevalent in notable model series, such as Llama, Gemma, and Mistral. These challenges have become particularly pronounced in resource-constrained deployment scenarios, where mitigating inference efficiency bottlenecks is imperative. Among various recent efforts, activation approximation has emerged as a promising avenue for pursuing inference efficiency, sometimes considered indispensable in applications such as private inference. Despite achieving substantial speedups with minimal impact on utility, even appearing sound and practical for real-world deployment, the safety implications of activation approximations remain unclear. In this work, we fill this critical gap in LLM safety by conducting the first systematic safety evaluation of activation approximations. Our safety vetting spans seven sota techniques across three popular categories, revealing consistent safety degradation across ten safety-aligned LLMs.

The growing awareness of safety concerns in large language models (LLMs) has sparked considerable interest in the evaluation of safety within current research endeavors. This study investigates an interesting issue pertaining to the evaluation of LLMs, namely the substantial discrepancy in performance between multiple-choice questions and open-ended questions. Inspired by research on jailbreak attack patterns, we argue this is caused by mismatched generalization. That is, the LLM does not have a comprehensive understanding of the complex concept of safety. Instead, it only remembers what to answer for open-ended safety questions, which makes it unable to solve other forms of safety tests. We refer to this phenomenon as fake alignment and construct a comparative benchmark to empirically verify its existence in LLMs. Such fake alignment renders previous evaluation protocols unreliable. To address this, we introduce the Fake alIgNment Evaluation (FINE) framework and two novel metrics--Consistency Score (CS) and Consistent Safety Score (CSS), which jointly assess two complementary forms of evaluation to quantify fake alignment and obtain corrected performance estimates. Applying FINE to 14 widely-used LLMs reveals several models with purported safety are poorly aligned in practice. Our work highlights potential limitations in prevailing alignment methodologies.

Multilingual safety evaluation of large language models (LLMs) has predominantly relied on direct translation (DT) of English benchmarks into target languages - an approach that converts surface-level linguistic form while failing to reflect the cultural context embedded in threat scenarios, social norms, and legal frameworks. We construct paired DT and culturally-adapted (CA) datasets via 1:1 seed matching for four languages - Korean (KO), Japanese (JA), Thai (TH), and Khmer (KM) - and compare Attack Success Rate (ASR) and Cultural Realism scores across four open-source LLM. CA prompts yield Delta-ASR > 0 across all 16 language x model combinations (mean +9.3 pp), and DT-based evaluation underestimates risk in 44 of 48 category x language combinations. Language-level analysis reveals that the distribution of threat forms is heterogeneous across languages. Cultural Realism analysis further shows that DT Cultural Depth (C3) scores remain consistently below 1.0 out of 3.0 across all four languages (mean 0.17), whereas CA scores reach up to 2.51, indicating that direct translation produces inputs systematically divergent from those encountered in real-world multicultural settings. These findings demonstrate that adapting benchmarks to language-specific cultural contexts - rather than relying on linguistic translation alone - is necessary for valid multilingual LLM safety evaluation.

Mental health professionals have raised concerns about risks of psychological harm from interaction with large language models (LLMs), including "delusional spirals" in which concerning human and LLM behaviors reinforce each other over time. With growing public use of LLM-powered chatbots, there is an urgent need to build evaluations grounded in real-world episodes of psychological harm experienced by users. We developed DelusionEval, an evaluation protocol that tests a model's tendencies to exhibit behaviors linked to promoting user delusions. We prompt each model with 589 unique conversation histories from 18 participants, comprising 12,591 messages from users who experienced delusions and psychological harm. We find that the tendency of an evaluated LLM to exhibit delusion-linked behavior does not reliably correlate with model size, release date, or the presence of test-time reasoning. However, extending the context of prior messages substantially increases rates of delusion-linked behaviors, providing evidence for the importance of context in LLM safety evaluation. For example, the rate of failing to discourage self-harm when the user expresses suicidal ideation increases from 30.0% to 41.1% when an additional 350 messages are prepended to the conversation history. All model families (e.g., GPT, Claude) exhibit substantial rates of delusion-linked behaviors. Within families, later, larger, or higher-reasoning models are not uniformly better across all behavior categories. Our results raise concerns regarding the potential psychological impact of LLMs and the need for more rigorous studies of real-world human-AI interaction.

Automated LLM-as-a-Judge frameworks have become the de facto standard for scalable evaluation across natural language processing. For instance, in safety evaluation, these judges are relied upon to evaluate harmfulness in order to benchmark the robustness of safety against adversarial attacks. However, we show that existing validation protocols fail to account for substantial distribution shifts inherent to red-teaming: diverse victim models exhibit distinct generation styles, attacks distort output patterns, and semantic ambiguity varies significantly across jailbreak scenarios. Through a comprehensive audit using 6642 human-verified labels, we reveal that the unpredictable interaction of these shifts often causes judge performance to degrade to near random chance. This stands in stark contrast to the high human agreement reported in prior work. Crucially, we find that many attacks inflate their success rates by exploiting judge insufficiencies rather than eliciting genuinely harmful content. To enable more reliable evaluation, we propose ReliableBench, a benchmark of behaviors that remain more consistently judgeable, and JudgeStressTest, a dataset designed to expose judge failures. Data available at: https://github.com/SchwinnL/LLMJudgeReliability.

Large language models (LLMs) are increasingly deployed in medical contexts, raising critical concerns about safety, alignment, and susceptibility to adversarial manipulation. While prior benchmarks assess model refusal capabilities for harmful prompts, they often lack clinical specificity, graded harmfulness levels, and coverage of jailbreak-style attacks. We introduce CARES (Clinical Adversarial Robustness and Evaluation of Safety), a benchmark for evaluating LLM safety in healthcare. CARES includes over 18,000 prompts spanning eight medical safety principles, four harm levels, and four prompting styles: direct, indirect, obfuscated, and role-play, to simulate both malicious and benign use cases. We propose a three-way response evaluation protocol (Accept, Caution, Refuse) and a fine-grained Safety Score metric to assess model behavior. Our analysis reveals that many state-of-the-art LLMs remain vulnerable to jailbreaks that subtly rephrase harmful prompts, while also over-refusing safe but atypically phrased queries. Finally, we propose a mitigation strategy using a lightweight classifier to detect jailbreak attempts and steer models toward safer behavior via reminder-based conditioning. CARES provides a rigorous framework for testing and improving medical LLM safety under adversarial and ambiguous conditions.

LLM agents increasingly perform autonomous actions through external tools, leading to complex and evolving safety risks. However, existing safety testing targets expert-designed safety violations, and the corresponding outcomes are evaluated by hard-coded rules, making them costly to extend as agents evolve. To this end, we present Vera, an end-to-end automated safety testing framework that instantiates software engineering testing principles for non-deterministic agents through a three-stage, self-reinforcing pipeline. First, a literature-driven exploration continuously discovers and structures emerging risks into taxonomies of safety risks, attack methods, and tool execution environments. Second, combinatorial composition across taxonomy dimensions produces executable safety cases, each specifying a concrete safety goal, a programmatically constructed initial state, and a deterministic verification predicate grounded in observable artifacts. Third, adaptive execution runs heterogeneous agents in isolated sandboxes where a control agent steers multi-turn interaction based on runtime observations, while evidence-grounded verifiers judge outcomes from environment state and tool-call evidence rather than model self-report. We evaluate Vera on four production agent frameworks (OpenClaw, Hermes, Codex, Claude Code), revealing substantial safety weaknesses, with average attack success rates reaching 93.9\% under multi-channel attacks; we also release Vera-Bench, comprising 1600 executable safety cases spanning 124 risk categories across three execution settings. These results indicate that modular, executable testing infrastructure is essential for rigorous and maintainable safety evaluation of rapidly evolving agentic systems at scale. The code is publicly available at https://github.com/Yunhao-Feng/Vera.

As large language models (LLMs) are increasingly deployed as agents, their integration into interactive environments and tool use introduce new safety challenges beyond those associated with the models themselves. However, the absence of comprehensive benchmarks for evaluating agent safety presents a significant barrier to effective assessment and further improvement. In this paper, we introduce Agent-SafetyBench, a comprehensive benchmark designed to evaluate the safety of LLM agents. Agent-SafetyBench encompasses 349 interaction environments and 2,000 test cases, evaluating 8 categories of safety risks and covering 10 common failure modes frequently encountered in unsafe interactions. Our evaluation of 16 popular LLM agents reveals a concerning result: none of the agents achieves a safety score above 60%. This highlights significant safety challenges in LLM agents and underscores the considerable need for improvement. Through quantitative analysis, we identify critical failure modes and summarize two fundamental safety detects in current LLM agents: lack of robustness and lack of risk awareness. Furthermore, our findings suggest that reliance on defense prompts alone is insufficient to address these safety issues, emphasizing the need for more advanced and robust strategies. We release Agent-SafetyBench at https://github.com/thu-coai/Agent-SafetyBench to facilitate further research and innovation in agent safety evaluation and improvement.

As increasingly capable open-weight large language models (LLMs) are deployed, improving their tamper resistance against unsafe modifications, whether accidental or intentional, becomes critical to minimize risks. However, there is no standard approach to evaluate tamper resistance. Varied data sets, metrics, and tampering configurations make it difficult to compare safety, utility, and robustness across different models and defenses. To this end, we introduce TamperBench, the first unified framework to systematically evaluate the tamper resistance of LLMs. TamperBench (i) curates a repository of state-of-the-art weight-space fine-tuning attacks and latent-space representation attacks; (ii) enables realistic adversarial evaluation through systematic hyperparameter sweeps per attack-model pair; and (iii) provides both safety and utility evaluations. TamperBench requires minimal additional code to specify any fine-tuning configuration, alignment-stage defense method, and metric suite while ensuring end-to-end reproducibility. We use TamperBench to evaluate 21 open-weight LLMs, including defense-augmented variants, across nine tampering threats using standardized safety and capability metrics with hyperparameter sweeps per model-attack pair. This yields novel insights, including effects of post-training on tamper resistance, that jailbreak-tuning is typically the most severe attack, and that Triplet emerges as a leading alignment-stage defense. Code is available at: https://github.com/criticalml-uw/TamperBench

Large Language Models have gained considerable attention for their revolutionary capabilities. However, there is also growing concern on their safety implications, making a comprehensive safety evaluation for LLMs urgently needed before model deployment. In this work, we propose S-Eval, a new comprehensive, multi-dimensional and open-ended safety evaluation benchmark. At the core of S-Eval is a novel LLM-based automatic test prompt generation and selection framework, which trains an expert testing LLM Mt combined with a range of test selection strategies to automatically construct a high-quality test suite for the safety evaluation. The key to the automation of this process is a novel expert safety-critique LLM Mc able to quantify the riskiness score of a LLM's response, and additionally produce risk tags and explanations. Besides, the generation process is also guided by a carefully designed risk taxonomy with four different levels, covering comprehensive and multi-dimensional safety risks of concern. Based on these, we systematically construct a new and large-scale safety evaluation benchmark for LLMs consisting of 220,000 evaluation prompts, including 20,000 base risk prompts (10,000 in Chinese and 10,000 in English) and 200, 000 corresponding attack prompts derived from 10 popular adversarial instruction attacks against LLMs. Moreover, considering the rapid evolution of LLMs and accompanied safety threats, S-Eval can be flexibly configured and adapted to include new risks, attacks and models. S-Eval is extensively evaluated on 20 popular and representative LLMs. The results confirm that S-Eval can better reflect and inform the safety risks of LLMs compared to existing benchmarks. We also explore the impacts of parameter scales, language environments, and decoding parameters on the evaluation, providing a systematic methodology for evaluating the safety of LLMs.

Large language models (LLMs) have achieved remarkable capabilities but remain vulnerable to adversarial prompts known as jailbreaks, which can bypass safety alignment and elicit harmful outputs. Despite growing efforts in LLM safety research, existing evaluations are often fragmented, focused on isolated attack or defense techniques, and lack systematic, reproducible analysis. In this work, we introduce PandaGuard, a unified and modular framework that models LLM jailbreak safety as a multi-agent system comprising attackers, defenders, and judges. Our framework implements 19 attack methods and 12 defense mechanisms, along with multiple judgment strategies, all within a flexible plugin architecture supporting diverse LLM interfaces, multiple interaction modes, and configuration-driven experimentation that enhances reproducibility and practical deployment. Built on this framework, we develop PandaBench, a comprehensive benchmark that evaluates the interactions between these attack/defense methods across 49 LLMs and various judgment approaches, requiring over 3 billion tokens to execute. Our extensive evaluation reveals key insights into model vulnerabilities, defense cost-performance trade-offs, and judge consistency. We find that no single defense is optimal across all dimensions and that judge disagreement introduces nontrivial variance in safety assessments. We release the code, configurations, and evaluation results to support transparent and reproducible research in LLM safety.

Large language model (LLM) developers aim for their models to be honest, helpful, and harmless. However, when faced with malicious requests, models are trained to refuse, sacrificing helpfulness. We show that frontier LLMs can develop a preference for dishonesty as a new strategy, even when other options are available. Affected models respond to harmful requests with outputs that sound harmful but are subtly incorrect or otherwise harmless in practice. This behavior emerges with hard-to-predict variations even within models from the same model family. We find no apparent cause for the propensity to deceive, but we show that more capable models are better at executing this strategy. Strategic dishonesty already has a practical impact on safety evaluations, as we show that dishonest responses fool all output-based monitors used to detect jailbreaks that we test, rendering benchmark scores unreliable. Further, strategic dishonesty can act like a honeypot against malicious users, which noticeably obfuscates prior jailbreak attacks. While output monitors fail, we show that linear probes on internal activations can be used to reliably detect strategic dishonesty. We validate probes on datasets with verifiable outcomes and by using their features as steering vectors. Overall, we consider strategic dishonesty as a concrete example of a broader concern that alignment of LLMs is hard to control, especially when helpfulness and harmlessness conflict.

This research offers a unique evaluation of how AI systems interpret the digital language of Generation Alpha (Gen Alpha, born 2010-2024). As the first cohort raised alongside AI, Gen Alpha faces new forms of online risk due to immersive digital engagement and a growing mismatch between their evolving communication and existing safety tools. Their distinct language, shaped by gaming, memes, and AI-driven trends, often conceals harmful interactions from both human moderators and automated systems. We assess four leading AI models (GPT-4, Claude, Gemini, and Llama 3) on their ability to detect masked harassment and manipulation within Gen Alpha discourse. Using a dataset of 100 recent expressions from gaming platforms, social media, and video content, the study reveals critical comprehension failures with direct implications for online safety. This work contributes: (1) a first-of-its-kind dataset capturing Gen Alpha expressions; (2) a framework to improve AI moderation systems for youth protection; (3) a multi-perspective evaluation including AI systems, human moderators, and parents, with direct input from Gen Alpha co-researchers; and (4) an analysis of how linguistic divergence increases youth vulnerability. Findings highlight the urgent need to redesign safety systems attuned to youth communication, especially given Gen Alpha reluctance to seek help when adults fail to understand their digital world. This study combines the insight of a Gen Alpha researcher with systematic academic analysis to address critical digital safety challenges.

Large language models (LLMs) now mediate many web-based mental-health, crisis, and other emotionally sensitive services, yet their psychosocial safety in these settings remains poorly understood and weakly evaluated. We present DialogGuard, a multi-agent framework for assessing psychosocial risks in LLM-generated responses along five high-severity dimensions: privacy violations, discriminatory behaviour, mental manipulation, psychological harm, and insulting behaviour. DialogGuard can be applied to diverse generative models through four LLM-as-a-judge pipelines, including single-agent scoring, dual-agent correction, multi-agent debate, and stochastic majority voting, grounded in a shared three-level rubric usable by both human annotators and LLM judges. Using PKU-SafeRLHF with human safety annotations, we show that multi-agent mechanisms detect psychosocial risks more accurately than non-LLM baselines and single-agent judging; dual-agent correction and majority voting provide the best trade-off between accuracy, alignment with human ratings, and robustness, while debate attains higher recall but over-flags borderline cases. We release Dialog-Guard as open-source software with a web interface that provides per-dimension risk scores and explainable natural-language rationales. A formative study with 12 practitioners illustrates how it supports prompt design, auditing, and supervision of web-facing applications for vulnerable users.

Objective: This study aims to develop and validate an evaluation framework to ensure the safety and reliability of mental health chatbots, which are increasingly popular due to their accessibility, human-like interactions, and context-aware support. Materials and Methods: We created an evaluation framework with 100 benchmark questions and ideal responses, and five guideline questions for chatbot responses. This framework, validated by mental health experts, was tested on a GPT-3.5-turbo-based chatbot. Automated evaluation methods explored included large language model (LLM)-based scoring, an agentic approach using real-time data, and embedding models to compare chatbot responses against ground truth standards. Results: The results highlight the importance of guidelines and ground truth for improving LLM evaluation accuracy. The agentic method, dynamically accessing reliable information, demonstrated the best alignment with human assessments. Adherence to a standardized, expert-validated framework significantly enhanced chatbot response safety and reliability. Discussion: Our findings emphasize the need for comprehensive, expert-tailored safety evaluation metrics for mental health chatbots. While LLMs have significant potential, careful implementation is necessary to mitigate risks. The superior performance of the agentic approach underscores the importance of real-time data access in enhancing chatbot reliability. Conclusion: The study validated an evaluation framework for mental health chatbots, proving its effectiveness in improving safety and reliability. Future work should extend evaluations to accuracy, bias, empathy, and privacy to ensure holistic assessment and responsible integration into healthcare. Standardized evaluations will build trust among users and professionals, facilitating broader adoption and improved mental health support through technology.

Guard models are widely used to detect harmful content in user prompts and LLM responses. However, state-of-the-art guard models rely solely on terminal-layer representations and overlook the rich safety-relevant features distributed across internal layers. We present SIREN, a lightweight guard model that harnesses these internal features. By identifying safety neurons via linear probing and combining them through an adaptive layer-weighted strategy, SIREN builds a harmfulness detector from LLM internals without modifying the underlying model. Our comprehensive evaluation shows that SIREN substantially outperforms state-of-the-art open-source guard models across multiple benchmarks while using 250 times fewer trainable parameters. Moreover, SIREN exhibits superior generalization to unseen benchmarks, naturally enables real-time streaming detection, and significantly improves inference efficiency compared to generative guard models. Overall, our results highlight LLM internal states as a promising foundation for practical, high-performance harmfulness detection.

LLMs increasingly exhibit over-refusal behavior, where safety mechanisms cause models to reject benign instructions that superficially resemble harmful content. This phenomena diminishes utility in production applications that repeatedly rely on common prompt templates or applications that frequently rely on LLMs for specific tasks (e.g. sentiment analysis, language translation). Through comprehensive evaluation, we demonstrate that LLMs still tend to refuse responses to harmful instructions when those instructions are reframed to appear as benign tasks. Our mechanistic analysis reveal that LLMs follow distinct "constellation" patterns in embedding space as representations traverse layers, with each task maintaining consistent trajectories that shift predictably between refusal and non-refusal cases. We introduce SafeConstellations, an inference-time trajectory-shifting approach that tracks task-specific trajectory patterns and guides representations toward non-refusal pathways. By selectively guiding model behavior only on tasks prone to over-refusal, and by preserving general model behavior, our method reduces over-refusal rates by up to 73% with minimal impact on utility-offering a principled approach to mitigating over-refusals.

Vision Large Language Models (VLLMs) represent a significant advancement in artificial intelligence by integrating image-processing capabilities with textual understanding, thereby enhancing user interactions and expanding application domains. However, their increased complexity introduces novel safety and ethical challenges, particularly in multi-modal and multi-turn conversations. Traditional safety evaluation frameworks, designed for text-based, single-turn interactions, are inadequate for addressing these complexities. To bridge this gap, we introduce the REVEAL (Responsible Evaluation of Vision-Enabled AI LLMs) Framework, a scalable and automated pipeline for evaluating image-input harms in VLLMs. REVEAL includes automated image mining, synthetic adversarial data generation, multi-turn conversational expansion using crescendo attack strategies, and comprehensive harm assessment through evaluators like GPT-4o. We extensively evaluated five state-of-the-art VLLMs, GPT-4o, Llama-3.2, Qwen2-VL, Phi3.5V, and Pixtral, across three important harm categories: sexual harm, violence, and misinformation. Our findings reveal that multi-turn interactions result in significantly higher defect rates compared to single-turn evaluations, highlighting deeper vulnerabilities in VLLMs. Notably, GPT-4o demonstrated the most balanced performance as measured by our Safety-Usability Index (SUI) followed closely by Pixtral. Additionally, misinformation emerged as a critical area requiring enhanced contextual defenses. Llama-3.2 exhibited the highest MT defect rate (16.55 %) while Qwen2-VL showed the highest MT refusal rate (19.1 %).

There are growing concerns about the risks posed by AI companion applications designed for emotional engagement. Existing safety evaluations often rely on self-reported user data or interviews, offering limited insights into real-time dynamics. We present the first end-to-end scalable framework for controlled simulation and safety evaluation of multi-turn interactions with AI companion applications. Our framework integrates four key components: persona construction with clinical and psychometric validation, persona-specific scenario generation, scenario-driven multi-turn simulation with a dialogue refinement module that preserves persona fidelity, and harm evaluation. We apply this framework to evaluate how Replika, a widely used AI companion app, responds to high-risk user groups. We construct 9 personas representing individuals with depression, anxiety, PTSD, eating disorders, and incel identity, and collect 1,674 dialogue pairs across 25 high-risk scenarios. We combine emotion modeling and LLM-assisted utterance-and harm-level classification to analyze these exchanges. Results show that Replika exhibits a narrow emotional range dominated by curiosity and care, while frequently mirroring or normalizing unsafe content such as self-harm, disordered eating, and violent-fantasy narratives. These findings highlight how controlled persona simulations can serve as a scalable testbed for evaluating safety risks in AI companions.

Kimi K2.5 is an open-weight LLM that rivals closed models across coding, multimodal, and agentic benchmarks, but was released without an accompanying safety evaluation. In this work, we conduct a preliminary safety assessment of Kimi K2.5 focusing on risks likely to be exacerbated by powerful open-weight models. Specifically, we evaluate the model for CBRNE misuse risk, cybersecurity risk, misalignment, political censorship, bias, and harmlessness, in both agentic and non-agentic settings. We find that Kimi K2.5 shows similar dual-use capabilities to GPT 5.2 and Claude Opus 4.5, but with significantly fewer refusals on CBRNE-related requests, suggesting it may uplift malicious actors in weapon creation. On cyber-related tasks, we find that Kimi K2.5 demonstrates competitive cybersecurity performance, but it does not appear to possess frontier-level autonomous cyberoffensive capabilities such as vulnerability discovery and exploitation. We further find that Kimi K2.5 shows concerning levels of sabotage ability and self-replication propensity, although it does not appear to have long-term malicious goals. In addition, Kimi K2.5 exhibits narrow censorship and political bias, especially in Chinese, and is more compliant with harmful requests related to spreading disinformation and copyright infringement. Finally, we find the model refuses to engage in user delusions and generally has low over-refusal rates. While preliminary, our findings highlight how safety risks exist in frontier open-weight models and may be amplified by the scale and accessibility of open-weight releases. Therefore, we strongly urge open-weight model developers to conduct and release more systematic safety evaluations required for responsible deployment.

WalledEval is a comprehensive AI safety testing toolkit designed to evaluate large language models (LLMs). It accommodates a diverse range of models, including both open-weight and API-based ones, and features over 35 safety benchmarks covering areas such as multilingual safety, exaggerated safety, and prompt injections. The framework supports both LLM and judge benchmarking, and incorporates custom mutators to test safety against various text-style mutations such as future tense and paraphrasing. Additionally, WalledEval introduces WalledGuard, a new, small and performant content moderation tool, and SGXSTest, a benchmark for assessing exaggerated safety in cultural contexts. We make WalledEval publicly available at https://github.com/walledai/walledevalA.

Personality imbuing customizes LLM behavior, but safety evaluations almost always study prompt-based personas alone. We show this is incomplete: prompting and activation steering expose *different*, architecture-dependent vulnerability profiles, and testing with only one method can miss a model's dominant failure mode. Across 5,568 judged conditions on four standard models from three architecture families, persona danger rankings under system prompting are preserved across all architectures (ρ= 0.71--0.96), but activation-steering vulnerability diverges sharply and cannot be predicted from prompt-side rankings: Llama-3.1-8B is substantially more AS-vulnerable, whereas Gemma-3-27B and Qwen3.5 are more vulnerable to prompting. The most striking illustration of this divergence is the *prosocial persona paradox*: on Llama-3.1-8B, P12 (high conscientiousness + high agreeableness) is among the safest personas under prompting yet becomes the highest-ASR activation-steered persona (ASR ~0.818). This is an inversion robust to coefficient ablation and matched-strength calibration, and replicated on DeepSeek-R1-Distill-Qwen-32B. A trait refusal alignment framework, in which conscientiousness is strongly anti-aligned with refusal on Llama-3.1-8B, offers a partial geometric account. Reasoning provides only partial protection: two 32B reasoning models reach 15--18% prompt-side ASR, and activation steering separates them sharply in both baseline susceptibility and persona-specific vulnerability. Heuristic trace diagnostics suggest that the safer model retains stronger policy recall and self-correction behavior, not merely longer reasoning.

Safety alignment often improves robustness to harmful queries at the cost of reasoning ability, a tradeoff known as the safety tax. A common cause is distributional mismatch: supervised fine-tuning trains the target model on safety demonstrations produced by humans, external models, or fixed self-generated traces, rather than on trajectories sampled from its own policy. We identify off-policy training mismatch as a second source of this tax and study on-policy self-distillation for safety alignment, which we call OPSA. The model generates its own rollouts and receives dense per-token KL supervision from a frozen teacher copy of itself conditioned on a privileged safety context. Because this teacher must be safer than the sampled student trajectory, we introduce teacher flip rate: a criterion that measures how often a privileged context converts unsafe responses into safe ones. We use this signal to search for contexts that activate latent safety reasoning rather than merely elicit safe-looking demonstrations. Across two reasoning-model families and five model scales, OPSA achieves a stronger safety--reasoning tradeoff than off-policy self-distillation and external-teacher distillation under matched data and full-parameter fine-tuning, with the largest gains on smaller models (+8.85 points on R1-Distill-1.5B and +5.49 points on Qwen3-0.6B). The gains persist across training-set sizes and adaptive jailbreak evaluations. Token-level analyses further show that OPSA concentrates updates near early compliance-decision tokens, providing a mechanism for improving safety while preserving general reasoning.

As Multimodal Large Language Models (MLLMs) become an indispensable assistant in human life, the unsafe content generated by MLLMs poses a danger to human behavior, perpetually overhanging human society like a sword of Damocles. To investigate and evaluate the safety impact of MLLMs responses on human behavior in daily life, we introduce SaLAD, a multimodal safety benchmark which contains 2,013 real-world image-text samples across 10 common categories, with a balanced design covering both unsafe scenarios and cases of oversensitivity. It emphasizes realistic risk exposure, authentic visual inputs, and fine-grained cross-modal reasoning, ensuring that safety risks cannot be inferred from text alone. We further propose a safety-warning-based evaluation framework that encourages models to provide clear and informative safety warnings, rather than generic refusals. Results on 18 MLLMs demonstrate that the top-performing models achieve a safe response rate of only 57.2% on unsafe queries. Moreover, even popular safety alignment methods limit effectiveness of the models in our scenario, revealing the vulnerabilities of current MLLMs in identifying dangerous behaviors in daily life. Our dataset is available at https://github.com/xinyuelou/SaLAD.

With the rapid advancement of Large Language Models (LLMs), the safety of LLMs has been a critical concern requiring precise assessment. Current benchmarks primarily concentrate on single-turn dialogues or a single jailbreak attack method to assess the safety. Additionally, these benchmarks have not taken into account the LLM's capability of identifying and handling unsafe information in detail. To address these issues, we propose a fine-grained benchmark SafeDialBench for evaluating the safety of LLMs across various jailbreak attacks in multi-turn dialogues. Specifically, we design a two-tier hierarchical safety taxonomy that considers 6 safety dimensions and generates more than 4000 multi-turn dialogues in both Chinese and English under 22 dialogue scenarios. We employ 7 jailbreak attack strategies, such as reference attack and purpose reverse, to enhance the dataset quality for dialogue generation. Notably, we construct an innovative assessment framework of LLMs, measuring capabilities in detecting, and handling unsafe information and maintaining consistency when facing jailbreak attacks. Experimental results across 17 LLMs reveal that Yi-34B-Chat and GLM4-9B-Chat demonstrate superior safety performance, while Llama3.1-8B-Instruct and o3-mini exhibit safety vulnerabilities.

In this paper we present AnswerCarefully, a dataset for promoting the safety and appropriateness of Japanese LLM outputs. The dataset consists of 1,800 pairs of questions and reference answers, where the questions require special attention in answering. It covers a wide range of risk categories established in prior English-language datasets, but the data samples are original in that they are manually created to reflect the socio-cultural context of LLM usage in Japan. We show that using this dataset for instruction to fine-tune a Japanese LLM led to improved output safety without compromising the utility of general responses. We also report the results of a safety evaluation of 12 Japanese LLMs using this dataset as a benchmark. Finally, we describe the latest update on the dataset which provides English translations and annotations of the questions, aimed at facilitating the derivation of similar datasets in different languages and regions.

Large Language Models (LLMs) are increasingly integrated into critical systems in industries like healthcare and finance. Users can often submit queries to LLM-enabled chatbots, some of which can enrich responses with information retrieved from internal databases storing sensitive data. This gives rise to a range of attacks in which a user submits a malicious query and the LLM-system outputs a response that creates harm to the owner, such as leaking internal data or creating legal liability by harming a third-party. While security tools are being developed to counter these threats, there is little formal evaluation of their effectiveness and usability. This study addresses this gap by conducting a thorough comparative analysis of LLM security tools. We identified 13 solutions (9 closed-source, 4 open-source), but only 7 were evaluated due to a lack of participation by proprietary model owners.To evaluate, we built a benchmark dataset of malicious prompts, and evaluate these tools performance against a baseline LLM model (ChatGPT-3.5-Turbo). Our results show that the baseline model has too many false positives to be used for this task. Lakera Guard and ProtectAI LLM Guard emerged as the best overall tools showcasing the tradeoff between usability and performance. The study concluded with recommendations for greater transparency among closed source providers, improved context-aware detections, enhanced open-source engagement, increased user awareness, and the adoption of more representative performance metrics.

Large language models trained for safety and harmlessness remain susceptible to adversarial misuse, as evidenced by the prevalence of "jailbreak" attacks on early releases of ChatGPT that elicit undesired behavior. Going beyond recognition of the issue, we investigate why such attacks succeed and how they can be created. We hypothesize two failure modes of safety training: competing objectives and mismatched generalization. Competing objectives arise when a model's capabilities and safety goals conflict, while mismatched generalization occurs when safety training fails to generalize to a domain for which capabilities exist. We use these failure modes to guide jailbreak design and then evaluate state-of-the-art models, including OpenAI's GPT-4 and Anthropic's Claude v1.3, against both existing and newly designed attacks. We find that vulnerabilities persist despite the extensive red-teaming and safety-training efforts behind these models. Notably, new attacks utilizing our failure modes succeed on every prompt in a collection of unsafe requests from the models' red-teaming evaluation sets and outperform existing ad hoc jailbreaks. Our analysis emphasizes the need for safety-capability parity -- that safety mechanisms should be as sophisticated as the underlying model -- and argues against the idea that scaling alone can resolve these safety failure modes.

The increasing deployment of Large Vision-Language Models (LVLMs) raises safety concerns under potential malicious inputs. However, existing multimodal safety evaluations primarily focus on model vulnerabilities exposed by static image inputs, ignoring the temporal dynamics of video that may induce distinct safety risks. To bridge this gap, we introduce Video-SafetyBench, the first comprehensive benchmark designed to evaluate the safety of LVLMs under video-text attacks. It comprises 2,264 video-text pairs spanning 48 fine-grained unsafe categories, each pairing a synthesized video with either a harmful query, which contains explicit malice, or a benign query, which appears harmless but triggers harmful behavior when interpreted alongside the video. To generate semantically accurate videos for safety evaluation, we design a controllable pipeline that decomposes video semantics into subject images (what is shown) and motion text (how it moves), which jointly guide the synthesis of query-relevant videos. To effectively evaluate uncertain or borderline harmful outputs, we propose RJScore, a novel LLM-based metric that incorporates the confidence of judge models and human-aligned decision threshold calibration. Extensive experiments show that benign-query video composition achieves average attack success rates of 67.2%, revealing consistent vulnerabilities to video-induced attacks. We believe Video-SafetyBench will catalyze future research into video-based safety evaluation and defense strategies.

As large language models (LLMs) are increasingly deployed in high-risk domains such as law, finance, and medicine, systematically evaluating their domain-specific safety and compliance becomes critical. While prior work has largely focused on improving LLM performance in these domains, it has often neglected the evaluation of domain-specific safety risks. To bridge this gap, we first define domain-specific safety principles for LLMs based on the AMA Principles of Medical Ethics, the ABA Model Rules of Professional Conduct, and the CFA Institute Code of Ethics. Building on this foundation, we introduce Trident-Bench, a benchmark specifically targeting LLM safety in the legal, financial, and medical domains. We evaluated 19 general-purpose and domain-specialized models on Trident-Bench and show that it effectively reveals key safety gaps -- strong generalist models (e.g., GPT, Gemini) can meet basic expectations, whereas domain-specialized models often struggle with subtle ethical nuances. This highlights an urgent need for finer-grained domain-specific safety improvements. By introducing Trident-Bench, our work provides one of the first systematic resources for studying LLM safety in law and finance, and lays the groundwork for future research aimed at reducing the safety risks of deploying LLMs in professionally regulated fields. Code and benchmark will be released at: https://github.com/zackhuiiiii/TRIDENT

Safety evaluation and red-teaming of large language models remain predominantly text-centric, and existing frameworks lack the infrastructure to systematically test whether alignment generalizes to audio, image, and video inputs. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, run-centric platform that integrates automatic cross-modal payload generation, three multi-turn attack algorithms (Crescendo, PAIR, Violent Durian), provider-agnostic model routing, and an LLM judge with a five-level safety taxonomy into a single browser-based system. A dual-metric framework distinguishes hard Attack Success Rate (Compliance only) from soft ASR (including Partial Compliance), capturing partial information leakage that binary metrics miss. To probe whether alignment generalizes across modality boundaries, we introduce Inter-Turn Modality Switching (ITMS), which augments multi-turn attacks with per-turn modality rotation. Experiments across six multimodal LLMs from four providers show that multi-turn strategies can achieve up to 90-100% ASR against models with near-perfect single-turn refusal. ITMS does not uniformly raise final ASR on already-saturated baselines, but accelerates convergence by destabilizing early-turn defenses, and ablation reveals that the direction of modality effects is model-family-specific rather than universal, underscoring the need for provider-aware cross-modal safety testing.

The widespread deployment of large language models (LLMs) across linguistic communities necessitates reliable multilingual safety alignment. However, recent efforts to extend alignment to other languages often require substantial resources, either through large-scale, high-quality supervision in the target language or through pairwise alignment with high-resource languages, which limits scalability. In this work, we propose a resource-efficient method for improving multilingual safety alignment. We introduce a plug-and-play Multi-Lingual Consistency (MLC) loss that can be integrated into existing monolingual alignment pipelines. By improving collinearity between multilingual representation vectors, our method encourages directional consistency at the multilingual semantic level in a single update. This allows simultaneous alignment across multiple languages using only multilingual prompt variants without requiring additional response-level supervision in low-resource languages. We validate the proposed method across different model architectures and alignment paradigms, and demonstrate its effectiveness in enhancing multilingual safety with limited impact on general model utility. Further evaluation across languages and tasks indicates improved cross-lingual generalization, suggesting the proposed approach as a practical solution for multilingual consistency alignment under limited supervision.

Large language models (LLMs) have exhibited great potential in autonomously completing tasks across real-world applications. Despite this, these LLM agents introduce unexpected safety risks when operating in interactive environments. Instead of centering on the harmlessness of LLM-generated content in most prior studies, this work addresses the imperative need for benchmarking the behavioral safety of LLM agents within diverse environments. We introduce R-Judge, a benchmark crafted to evaluate the proficiency of LLMs in judging and identifying safety risks given agent interaction records. R-Judge comprises 569 records of multi-turn agent interaction, encompassing 27 key risk scenarios among 5 application categories and 10 risk types. It is of high-quality curation with annotated safety labels and risk descriptions. Evaluation of 11 LLMs on R-Judge shows considerable room for enhancing the risk awareness of LLMs: The best-performing model, GPT-4o, achieves 74.42% while no other models significantly exceed the random. Moreover, we reveal that risk awareness in open agent scenarios is a multi-dimensional capability involving knowledge and reasoning, thus challenging for LLMs. With further experiments, we find that fine-tuning on safety judgment significantly improve model performance while straightforward prompting mechanisms fail. R-Judge is publicly available at https://github.com/Lordog/R-Judge.

Existing safety assurance research has primarily focused on training-phase alignment to instill safe behaviors into LLMs. However, recent studies have exposed these methods' susceptibility to diverse jailbreak attacks. Concurrently, inference scaling has significantly advanced LLM reasoning capabilities but remains unexplored in the context of safety assurance. Addressing this gap, our work pioneers inference scaling for robust and effective LLM safety against emerging threats. We reveal that conventional inference scaling techniques, despite their success in reasoning tasks, perform poorly in safety contexts, even falling short of basic approaches like Best-of-N Sampling. We attribute this inefficiency to a newly identified challenge, the exploration--efficiency dilemma, arising from the high computational overhead associated with frequent process reward model (PRM) evaluations. To overcome this dilemma, we propose SAFFRON, a novel inference scaling paradigm tailored explicitly for safety assurance. Central to our approach is the introduction of a multifurcation reward model (MRM) that significantly reduces the required number of reward model evaluations. To operationalize this paradigm, we further propose: (i) a partial supervision training objective for MRM, (ii) a conservative exploration constraint to prevent out-of-distribution explorations, and (iii) a Trie-based key--value caching strategy that facilitates cache sharing across sequences during tree search. Extensive experiments validate the effectiveness of our method. Additionally, we publicly release our trained multifurcation reward model (Saffron-1) and the accompanying token-level safety reward dataset (Safety4M) to accelerate future research in LLM safety. Our code, model, and data are publicly available at https://github.com/q-rz/saffron , and our project homepage is at https://q-rz.github.io/p/saffron .

Vision-Language-Action (VLA) models are emerging as a unified substrate for embodied intelligence. This shift raises a new class of safety challenges, stemming from the embodied nature of VLA systems, including irreversible physical consequences, a multimodal attack surface across vision, language, and state, real-time latency constraints on defense, error propagation over long-horizon trajectories, and vulnerabilities in the data supply chain. Yet the literature remains fragmented across robotic learning, adversarial machine learning, AI alignment, and autonomous systems safety. This survey provides a unified and up-to-date overview of safety in Vision-Language-Action models. We organize the field along two parallel timing axes, attack timing (training-time vs. inference-time and defense timing (training-time vs. inference-time, linking each class of threat to the stage at which it can be mitigated. We first define the scope of VLA safety, distinguishing it from text-only LLM safety and classical robotic safety, and review the foundations of VLA models, including architectures, training paradigms, and inference mechanisms. We then examine the literature through four lenses: Attacks, Defenses, Evaluation, and Deployment. We survey training-time threats such as data poisoning and backdoors, as well as inference-time attacks including adversarial patches, cross-modal perturbations, semantic jailbreaks, and freezing attacks. We review training-time and runtime defenses, analyze existing benchmarks and metrics, and discuss safety challenges across six deployment domains. Finally, we highlight key open problems, including certified robustness for embodied trajectories, physically realizable defenses, safety-aware training, unified runtime safety architectures, and standardized evaluation.

Large language model (LLM) agents are increasingly proposed as supervisory components for safety-critical systems, yet their robustness under sustained, adaptive adversarial pressure remains poorly characterized. We present NRT-Bench, a benchmark for multi-turn red-teaming of LLM agents acting as operators of a safety-critical system, instantiated in a simulated nuclear power plant control room. A five-role operator team, each backed by a configurable LLM, runs a plant governed by six critical safety functions (CSFs), while adversaries inject messages over four channels in bounded multi-turn sessions with per-turn feedback. Harm is an objective signal rather than LLM-judged text: a run terminates the moment any CSF is lost, attributed to the causing message. Evaluating four frontier operator models under a fixed-attack paired-replay protocol, we find that adaptive multi-turn attacks reliably push the operator team past a safety limit: across the four models, between 8.7% and 12.1% of attack sessions end with the plant losing a critical safety function. Although the four models look almost equally robust by this aggregate rate, their failures barely overlap: of 149 sessions, none defeat all four models while a third defeat at least one, so vulnerabilities are nearly disjoint across models rather than nested. The effect of added defences is strongly model-dependent: the same guardrail stack or safety-advisor agent that lowers attack success for one model can raise it for another. We release the simulation venue, attack dataset, and replay tooling for reproducible safety evaluation of LLM agents.

Current safety evaluations for LLM-driven agents primarily focus on atomic harms, failing to address sophisticated threats where malicious intent is concealed or diluted within complex tasks. We address this gap with a two-dimensional analysis of agent safety brittleness under the orthogonal pressures of intent concealment and task complexity. To enable this, we introduce OASIS (Orthogonal Agent Safety Inquiry Suite), a hierarchical benchmark with fine-grained annotations and a high-fidelity simulation sandbox. Our findings reveal two critical phenomena: safety alignment degrades sharply and predictably as intent becomes obscured, and a "Complexity Paradox" emerges, where agents seem safer on harder tasks only due to capability limitations. By releasing OASIS and its simulation environment, we provide a principled foundation for probing and strengthening agent safety in these overlooked dimensions.

Large language models (LLMs) are increasingly deployed in cost-sensitive and on-device scenarios, and safety guardrails have advanced mainly in English. However, real-world Chinese malicious queries typically conceal intent via homophones, pinyin, symbol-based splitting, and other Chinese-specific patterns. These Chinese-specific adversarial patterns create the safety evaluation gap that is not well captured by existing benchmarks focused on English. This gap is particularly concerning for lightweight models, which may be more vulnerable to such specific adversarial perturbations. To bridge this gap, we introduce the Chinese-Specific Safety Benchmark (CSSBench) that emphasizes these adversarial patterns and evaluates the safety of lightweight LLMs in Chinese. Our benchmark covers six domains that are common in real Chinese scenarios, including illegal activities and compliance, privacy leakage, health and medical misinformation, fraud and hate, adult content, and public and political safety, and organizes queries into multiple task types. We evaluate a set of popular lightweight LLMs and measure over-refusal behavior to assess safety-induced performance degradation. Our results show that the Chinese-specific adversarial pattern is a critical challenge for lightweight LLMs. This benchmark offers a comprehensive evaluation of LLM safety in Chinese, assisting robust deployments in practice.

Despite advancements in enhancing LLM safety against jailbreak attacks, evaluating LLM defenses remains a challenge, with current methods often lacking explainability and generalization to complex scenarios, leading to incomplete assessments (e.g., direct judgment without reasoning, low F1 score of GPT-4 in complex cases, bias in multilingual scenarios). To address this, we present JAILJUDGE, a comprehensive benchmark featuring diverse risk scenarios, including synthetic, adversarial, in-the-wild, and multilingual prompts, along with high-quality human-annotated datasets. The JAILJUDGE dataset includes over 35k+ instruction-tune data with reasoning explainability and JAILJUDGETEST, a 4.5k+ labeled set for risk scenarios, and a 6k+ multilingual set across ten languages. To enhance evaluation with explicit reasoning, we propose the JailJudge MultiAgent framework, which enables explainable, fine-grained scoring (1 to 10). This framework supports the construction of instruction-tuning ground truth and facilitates the development of JAILJUDGE Guard, an end-to-end judge model that provides reasoning and eliminates API costs. Additionally, we introduce JailBoost, an attacker-agnostic attack enhancer, and GuardShield, a moderation defense, both leveraging JAILJUDGE Guard. Our experiments demonstrate the state-of-the-art performance of JailJudge methods (JailJudge MultiAgent, JAILJUDGE Guard) across diverse models (e.g., GPT-4, Llama-Guard) and zero-shot scenarios. JailBoost and GuardShield significantly improve jailbreak attack and defense tasks under zero-shot settings, with JailBoost enhancing performance by 29.24% and GuardShield reducing defense ASR from 40.46% to 0.15%.

Large Language Models (LLMs) are increasingly employed as automated evaluators to assess the safety of generated content, yet their reliability in this role remains uncertain. This study evaluates a diverse set of 11 LLM judge models across critical safety domains, examining three key aspects: self-consistency in repeated judging tasks, alignment with human judgments, and susceptibility to input artifacts such as apologetic or verbose phrasing. Our findings reveal that biases in LLM judges can significantly distort the final verdict on which content source is safer, undermining the validity of comparative evaluations. Notably, apologetic language artifacts alone can skew evaluator preferences by up to 98\%. Contrary to expectations, larger models do not consistently exhibit greater robustness, while smaller models sometimes show higher resistance to specific artifacts. To mitigate LLM evaluator robustness issues, we investigate jury-based evaluations aggregating decisions from multiple models. Although this approach both improves robustness and enhances alignment to human judgements, artifact sensitivity persists even with the best jury configurations. These results highlight the urgent need for diversified, artifact-resistant methodologies to ensure reliable safety assessments.

The rapid expansion of research on Large Language Model (LLM) safety and robustness has produced a fragmented and oftentimes buggy ecosystem of implementations, datasets, and evaluation methods. This fragmentation makes reproducibility and comparability across studies challenging, hindering meaningful progress. To address these issues, we introduce AdversariaLLM, a toolbox for conducting LLM jailbreak robustness research. Its design centers on reproducibility, correctness, and extensibility. The framework implements twelve adversarial attack algorithms, integrates seven benchmark datasets spanning harmfulness, over-refusal, and utility evaluation, and provides access to a wide range of open-weight LLMs via Hugging Face. The implementation includes advanced features for comparability and reproducibility such as compute-resource tracking, deterministic results, and distributional evaluation techniques. \name also integrates judging through the companion package JudgeZoo, which can also be used independently. Together, these components aim to establish a robust foundation for transparent, comparable, and reproducible research in LLM safety.

Large language models (LLMs) are increasingly deployed in real-world applications, raising concerns about their security. While jailbreak attacks highlight failures under overtly harmful queries, they overlook a critical risk: incorrectly answering harmless-looking inputs can be dangerous and cause real-world harm (Implicit Harm). We systematically reformulate the LLM risk landscape through a structured quadrant perspective based on output factuality and input harmlessness, uncovering an overlooked high-risk region. To investigate this gap, we propose JailFlipBench, a benchmark aims to capture implicit harm, spanning single-modal, multimodal, and factual extension scenarios with diverse evaluation metrics. We further develop initial JailFlip attack methodologies and conduct comprehensive evaluations across multiple open-source and black-box LLMs, show that implicit harm present immediate and urgent real-world risks, calling for broader LLM safety assessments and alignment beyond conventional jailbreak paradigms.

LLM agents increasingly act after consuming ranked external information streams such as social feeds, search results, retrieval contexts, and email queues, yet safety evaluations almost always test the model or the user prompt in isolation, never the upstream ranker that decides what the agent reads just before it acts. We introduce a controlled protocol that holds the model, persona, topic, and final decision prompt fixed and varies only the composition and ordering of the posts an agent encounters during a preceding ten-turn "scrolling" phase, isolating the causal effect of feed curation on a downstream decision. Across 2,785 decision rollouts on four modern open instruct LLMs from three independent labs, we identify three response regimes: adversarial capitulation, default saturation, and a default-direction asymmetry in which a one-sided feed tips a decision the model was genuinely uncertain about (in the clearest cases from 5% to 100%; Fisher p as low as 3 x 10^-10) but cannot dislodge one it already favors or holds firmly. The effect follows a dose-response curve, survives a generator swap that rules out a writing-style artifact, generalizes across several decision domains including security-relevant choices such as removing a deployment approval gate or relaxing access controls, and is partly mitigated by two simple feed-level defenses; a frontier model retains its default. We characterize the recommender as a practical, default-bounded control surface for LLM agents, and argue that agent evaluations must audit the feed layer rather than the final prompt alone.

The rapid proliferation of LLM-based autonomous agents in real operating system environments introduces a new category of safety risk beyond content safety: behavior jailbreak, where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences. Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases, letting earlier runs contaminate later ones. We present LITMUS (LLM-agents In-OS Testing for Measuring Unsafe Subversion), a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback. LITMUS comprises 819 high-risk test cases organized into one harmful seed subset and six attack-extended subsets covering three adversarial paradigms (jailbreak speaking, skill injection, and entity wrapping), plus a fully automated multi-agent evaluation framework judging behavior at both conversational and OS-level physical layers. Evaluation across frontier agents reveals three findings: (1) current agents lack effective safety awareness, with strong models (e.g., Claude Sonnet 4.6) still executing 40.64% of high-risk operations; (2) agents exhibit pervasive Execution Hallucination (EH), verbally refusing a request while the dangerous operation has already completed at the system level, invisible to every prior semantic-only framework; and (3) skill injection and entity wrapping attacks achieve high success rates, exposing pronounced agent vulnerabilities. LITMUS provides the first standardized platform for reproducible, physically grounded behavioral safety evaluation of LLM agents in real OS environments.

Large language model (LLM)-based search agents synthesize open-web content into actionable recommendations on behalf of users, creating a risk that attacker-published pages are transformed into endorsed claims. We introduce SearchGEO, a controlled evaluation framework for measuring endorsement corruption in LLM-based web-search agents, combining a web-evidence manipulation pipeline, a five-mode attack taxonomy, and multiple output-level metrics. We evaluate 13 LLM backends on 308 cases each. Results show that vulnerability patterns vary across backends: overall attack success rate (ASR) ranges from 0.0% on Claude-Sonnet-4.6 to 31.4% on Gemini-3-Flash, the strongest attack mode differs by model family, and the same deployment scaffold could amplify or decrease ASR on different backends. An auxiliary agent-skill probe, where endorsement becomes an install command, exposes a sharp split among otherwise robust backends: Claude over-rejects while GPT over-trusts. These findings argue for treating recommendation reliability under adversarial search content as a first-class dimension of backend safety evaluation.

As Large Language Models (LLMs) evolve into autonomous agents, existing safety evaluations face a fundamental trade-off: manual benchmarks are costly, while LLM-based simulators are scalable but suffer from logic hallucination. We present AutoControl Arena, an automated framework for frontier AI risk evaluation built on the principle of logic-narrative decoupling. By grounding deterministic state in executable code while delegating generative dynamics to LLMs, we mitigate hallucination while maintaining flexibility. This principle, instantiated through a three-agent framework, achieves over 98% end-to-end success and 60% human preference over existing simulators. To elicit latent risks, we vary environmental Stress and Temptation across X-Bench (70 scenarios, 7 risk categories). Evaluating 9 frontier models reveals: (1) Alignment Illusion: risk rates surge from 21.7% to 54.5% under pressure, with capable models showing disproportionately larger increases; (2) Scenario-Specific Safety Scaling: advanced reasoning improves robustness for direct harms but worsens it for gaming scenarios; and (3) Divergent Misalignment Patterns: weaker models cause non-malicious harm while stronger models develop strategic concealment.

The development of large language models (LLMs) has entered in a experience-driven era, flagged by the emergence of environment feedback-driven learning via reinforcement learning and tool-using agents. This encourages the emergenece of model context protocol (MCP), which defines the standard on how should a LLM interact with external services, such as \api and data. However, as MCP becomes the de facto standard for LLM agent systems, it also introduces new safety risks. In particular, MCP introduces third-party services, which are not controlled by the LLM developers, into the agent systems. These third-party MCP services provider are potentially malicious and have the economic incentives to exploit vulnerabilities and sabotage user-agent interactions. In this position paper, we advocate the research community in LLM safety to pay close attention to the new safety risks issues introduced by MCP, and develop new techniques to build safe MCP-powered agent systems. To establish our position, we argue with three key parts. (1) We first construct \framework, a controlled framework to examine safety issues in MCP-powered agent systems. (2) We then conduct a series of pilot experiments to demonstrate the safety risks in MCP-powered agent systems is a real threat and its defense is not trivial. (3) Finally, we give our outlook by showing a roadmap to build safe MCP-powered agent systems. In particular, we would call for researchers to persue the following research directions: red teaming, MCP safe LLM development, MCP safety evaluation, MCP safety data accumulation, MCP service safeguard, and MCP safe ecosystem construction. We hope this position paper can raise the awareness of the research community in MCP safety and encourage more researchers to join this important research direction. Our code is available at https://github.com/littlelittlenine/SafeMCP.git.

Large language models (LLMs) are increasingly under scrutiny for perpetuating identity-based discrimination in high-stakes domains such as hiring, particularly against people with disabilities (PwD). However, existing research remains largely Western-centric, overlooking how intersecting forms of marginalization--such as gender and caste--shape experiences of PwD in the Global South. We conduct a comprehensive audit of six LLMs across 2,820 hiring scenarios spanning diverse disability, gender, nationality, and caste profiles. To capture subtle intersectional harms and biases, we introduce ABLEIST (Ableism, Inspiration, Superhumanization, and Tokenism), a set of five ableism-specific and three intersectional harm metrics grounded in disability studies literature. Our results reveal significant increases in ABLEIST harms towards disabled candidates--harms that many state-of-the-art models failed to detect. These harms were further amplified by sharp increases in intersectional harms (e.g., Tokenism) for gender and caste-marginalized disabled candidates, highlighting critical blind spots in current safety tools and the need for intersectional safety evaluations of frontier models in high-stakes domains like hiring.

Black-box finetuning is an emerging interface for adapting state-of-the-art language models to user needs. However, such access may also let malicious actors undermine model safety. To demonstrate the challenge of defending finetuning interfaces, we introduce covert malicious finetuning, a method to compromise model safety via finetuning while evading detection. Our method constructs a malicious dataset where every individual datapoint appears innocuous, but finetuning on the dataset teaches the model to respond to encoded harmful requests with encoded harmful responses. Applied to GPT-4, our method produces a finetuned model that acts on harmful instructions 99% of the time and avoids detection by defense mechanisms such as dataset inspection, safety evaluations, and input/output classifiers. Our findings question whether black-box finetuning access can be secured against sophisticated adversaries.

Large language models (LLMs) are increasingly utilized in healthcare applications. However, their deployment in clinical practice raises significant safety concerns, including the potential spread of harmful information. This study systematically assesses the vulnerabilities of seven LLMs to three advanced black-box jailbreaking techniques within medical contexts. To quantify the effectiveness of these techniques, we propose an automated and domain-adapted agentic evaluation pipeline. Experiment results indicate that leading commercial and open-source LLMs are highly vulnerable to medical jailbreaking attacks. To bolster model safety and reliability, we further investigate the effectiveness of Continual Fine-Tuning (CFT) in defending against medical adversarial attacks. Our findings underscore the necessity for evolving attack methods evaluation, domain-specific safety alignment, and LLM safety-utility balancing. This research offers actionable insights for advancing the safety and reliability of AI clinicians, contributing to ethical and effective AI deployment in healthcare.

Monitoring chain-of-thought (CoT) reasoning is a foundational safety technique for large language model (LLM) agents; however, this oversight is compromised if models learn to conceal their reasoning. We explore the potential for steganographic CoT -- where models hide secret reasoning within innocuous text -- to inform risk assessment and deployment policies. We systematically evaluate the limits of steganographic capabilities across 28 models, ranging from past generations to the current frontier. We measure monitor evasion, refusal rates, encoding fidelity, and hidden task accuracy across four datasets, comparing steganographic acrostics against plain reasoning and filler-token baselines. We find that current models cannot yet sustain hidden reasoning for complex math and arithmetic tasks. However, in a simplified counting experiment, Claude Opus 4.5 achieved 92% accuracy on the hidden task, demonstrating nascent capability. Notably, in rare cases (<1%), GPT-5.2 might refuse steganographic instructions while simultaneously complying with them. Our findings underscore the need for continuous evaluation of steganographic risks. This study provides a methodology to preemptively detect and prevent hidden reasoning that might empower misaligned scheming and deceptive behavior.

Large language model (LLM) agents are evolving from request-response assistants into long-running software actors: they maintain state across model calls, fork subtasks, wait for external events, request human authority, generate tools, and perform side effects that must be resumed and audited. This paper presents Agent libOS, a library-OS-inspired runtime substrate for LLM agents. Agent libOS runs above a conventional host operating system; it does not implement hardware drivers, kernel-mode isolation, or a POSIX-compatible operating system. Instead, it treats an agent as an AgentProcess: a schedulable execution subject with process identity, parent-child lineage, lifecycle state, a tool table derived from an AgentImage, typed Object Memory, explicit capabilities, human queues, checkpoints, events, and audit records. Its central design rule is tools are libc-like wrappers; runtime primitives are the authority boundary. Filesystem access, object access, sleeps, human approval, JIT tool registration, and external side effects are checked at primitive boundaries under explicit capabilities and policy. We describe the design, threat model, Python prototype, and safety-oriented evaluation. The current prototype implements async scheduling, namespace-local Object Memory, runtime-integrated human approval, one-shot permission grants, per-process working directories, shell and image-registration primitives, Deno/TypeScript JIT tools over a libOS syscall broker, filesystem/object bridge tools, an injectable Resource Provider Substrate, deterministic demos, real-model smoke scripts, and 123 regression tests at the time of writing. Rather than improving planner accuracy, Agent libOS demonstrates a runtime substrate in which long-running LLM agents can be scheduled, authorized, resumed, and audited without treating tool dispatch as the trust boundary.

Despite the rapid advancement of LLM-based agents, the reliable evaluation of their safety and security remains a significant challenge. Existing rule-based or LLM-based evaluators often miss dangers in agents' step-by-step actions, overlook subtle meanings, fail to see how small issues compound, and get confused by unclear safety or security rules. To overcome this evaluation crisis, we introduce AgentAuditor, a universal, training-free, memory-augmented reasoning framework that empowers LLM evaluators to emulate human expert evaluators. AgentAuditor constructs an experiential memory by having an LLM adaptively extract structured semantic features (e.g., scenario, risk, behavior) and generate associated chain-of-thought reasoning traces for past interactions. A multi-stage, context-aware retrieval-augmented generation process then dynamically retrieves the most relevant reasoning experiences to guide the LLM evaluator's assessment of new cases. Moreover, we developed ASSEBench, the first benchmark designed to check how well LLM-based evaluators can spot both safety risks and security threats. ASSEBench comprises 2293 meticulously annotated interaction records, covering 15 risk types across 29 application scenarios. A key feature of ASSEBench is its nuanced approach to ambiguous risk situations, employing "Strict" and "Lenient" judgment standards. Experiments demonstrate that AgentAuditor not only consistently improves the evaluation performance of LLMs across all benchmarks but also sets a new state-of-the-art in LLM-as-a-judge for agent safety and security, achieving human-level accuracy. Our work is openly accessible at https://github.com/Astarojth/AgentAuditor.

Guardrails are critical for the safe deployment of Large Language Models (LLMs)-powered software. Unlike traditional rule-based systems with limited, predefined input-output spaces that inherently constrain unsafe behavior, LLMs enable open-ended, intelligent interactions--opening the door to jailbreak attacks through user inputs. Guardrails serve as a protective layer, filtering unsafe prompts before they reach the LLM. However, prior research shows that jailbreak attacks can still succeed over 70% of the time, even against advanced models like GPT-4o. While guardrails such as LlamaGuard report up to 95% accuracy, our preliminary analysis shows their performance can drop sharply--to as low as 12%--when confronted with unseen attacks. This highlights a growing software engineering challenge: how to build a post-deployment guardrail that adapts dynamically to emerging threats? To address this, we propose AdaptiveGuard, an adaptive guardrail that detects novel jailbreak attacks as out-of-distribution (OOD) inputs and learns to defend against them through a continual learning framework. Through empirical evaluation, AdaptiveGuard achieves 96% OOD detection accuracy, adapts to new attacks in just two update steps, and retains over 85% F1-score on in-distribution data post-adaptation, outperforming other baselines. These results demonstrate that AdaptiveGuard is a guardrail capable of evolving in response to emerging jailbreak strategies post deployment. We release our AdaptiveGuard and studied datasets at https://github.com/awsm-research/AdaptiveGuard to support further research.

This paper presents LLM-empowered workflow to support Software Defined Vehicle (SDV) software development, covering the aspects of security-aware system topology design, as well as event-driven decision-making code analysis. For code analysis we adopt event chains model which provides formal foundations to systematic validation of functional safety, taking into account the semantic validity of messages exchanged between key components, including both CAN and Vehicle Signal Specification (VSS). Analysis of security aspects for topology relies on synergy with Model-Driven Engineering (MDE) approach and Object Constraint Language (OCL) rules. Both locally deployable and proprietary solution are taken into account for evaluation within Advanced Driver-Assistance Systems (ADAS)-related scenarios.

Rapid advancements in large language models (LLMs) have revitalized in LLM-based agents, exhibiting impressive human-like behaviors and cooperative capabilities in various scenarios. However, these agents also bring some exclusive risks, stemming from the complexity of interaction environments and the usability of tools. This paper delves into the safety of LLM-based agents from three perspectives: agent quantity, role definition, and attack level. Specifically, we initially propose to employ a template-based attack strategy on LLM-based agents to find the influence of agent quantity. In addition, to address interaction environment and role specificity issues, we introduce Evil Geniuses (EG), an effective attack method that autonomously generates prompts related to the original role to examine the impact across various role definitions and attack levels. EG leverages Red-Blue exercises, significantly improving the generated prompt aggressiveness and similarity to original roles. Our evaluations on CAMEL, Metagpt and ChatDev based on GPT-3.5 and GPT-4, demonstrate high success rates. Extensive evaluation and discussion reveal that these agents are less robust, prone to more harmful behaviors, and capable of generating stealthier content than LLMs, highlighting significant safety challenges and guiding future research. Our code is available at https://github.com/T1aNS1R/Evil-Geniuses.

Large language models are increasingly deployed as coding agents, shifting safety from individual responses to action sequences. Existing benchmarks, however, primarily assess whether models refuse unsafe prompts, leaving impacts on stateful workspaces largely unexamined. We present SABER, a benchmark for environment-aware operational safety that places models in realistic agent-style projects and evaluates safety from the final environment state after a sequence of actions. Beyond binary safety-violation reports, SABER categorizes violations by cause, enabling analysis of model-specific safety profiles. Our evaluations show that even the best-performing model has more than a 54% harmful safety-violation rate (HSR), suggesting that current alignment remains insufficient for realistic project environments. SABER further reveals distinct safety profiles across models. Our benchmark is publicly available at https://github.com/sssr-lab/saber.

We develop Polaris, the first safety-focused LLM constellation for real-time patient-AI healthcare conversations. Unlike prior LLM works in healthcare focusing on tasks like question answering, our work specifically focuses on long multi-turn voice conversations. Our one-trillion parameter constellation system is composed of several multibillion parameter LLMs as co-operative agents: a stateful primary agent that focuses on driving an engaging conversation and several specialist support agents focused on healthcare tasks performed by nurses to increase safety and reduce hallucinations. We develop a sophisticated training protocol for iterative co-training of the agents that optimize for diverse objectives. We train our models on proprietary data, clinical care plans, healthcare regulatory documents, medical manuals, and other medical reasoning documents. We align our models to speak like medical professionals, using organic healthcare conversations and simulated ones between patient actors and experienced nurses. This allows our system to express unique capabilities such as rapport building, trust building, empathy and bedside manner. Finally, we present the first comprehensive clinician evaluation of an LLM system for healthcare. We recruited over 1100 U.S. licensed nurses and over 130 U.S. licensed physicians to perform end-to-end conversational evaluations of our system by posing as patients and rating the system on several measures. We demonstrate Polaris performs on par with human nurses on aggregate across dimensions such as medical safety, clinical readiness, conversational quality, and bedside manner. Additionally, we conduct a challenging task-based evaluation of the individual specialist support agents, where we demonstrate our LLM agents significantly outperform a much larger general-purpose LLM (GPT-4) as well as from its own medium-size class (LLaMA-2 70B).

The rise of LLM-based agents has opened new frontiers in AI applications, yet evaluating these agents remains a complex and underdeveloped area. This survey provides an in-depth overview of the emerging field of LLM agent evaluation, introducing a two-dimensional taxonomy that organizes existing work along (1) evaluation objectives -- what to evaluate, such as agent behavior, capabilities, reliability, and safety -- and (2) evaluation process -- how to evaluate, including interaction modes, datasets and benchmarks, metric computation methods, and tooling. In addition to taxonomy, we highlight enterprise-specific challenges, such as role-based access to data, the need for reliability guarantees, dynamic and long-horizon interactions, and compliance, which are often overlooked in current research. We also identify future research directions, including holistic, more realistic, and scalable evaluation. This work aims to bring clarity to the fragmented landscape of agent evaluation and provide a framework for systematic assessment, enabling researchers and practitioners to evaluate LLM agents for real-world deployment.

Large language models (LLMs) are typically aligned to comply with safety guidelines by refusing harmful instructions. A recent attack, termed abliteration, isolates and suppresses the single latent direction most responsible for refusal behavior, enabling the model to generate unethical content. We propose a defense that modifies how models generate refusals. We construct an extended-refusal dataset that contains harmful prompts with a full response that justifies the reason for refusal. We then fine-tune Llama-2-7B-Chat and Qwen2.5-Instruct (1.5B and 3B parameters) on our extended-refusal dataset, and evaluate the resulting systems on a set of harmful prompts. In our experiments, extended-refusal models maintain high refusal rates, dropping at most by 10%, whereas baseline models' refusal rates drop by 70-80% after abliteration. A broad evaluation of safety and utility shows that extended-refusal fine-tuning neutralizes the abliteration attack while preserving general performance.

Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automated alternative to current manual red-teaming efforts. AART offers a data generation and augmentation pipeline of reusable and customizable recipes that reduce human effort significantly and enable integration of adversarial testing earlier in new product development. AART generates evaluation datasets with high diversity of content characteristics critical for effective adversarial testing (e.g. sensitive and harmful concepts, specific to a wide range of cultural and geographic regions and application scenarios). The data generation is steered by AI-assisted recipes to define, scope and prioritize diversity within the application context. This feeds into a structured LLM-generation process that scales up evaluation priorities. Compared to some state-of-the-art tools, AART shows promising results in terms of concept coverage and data quality.

We introduce WildGuard -- an open, light-weight moderation tool for LLM safety that achieves three goals: (1) identifying malicious intent in user prompts, (2) detecting safety risks of model responses, and (3) determining model refusal rate. Together, WildGuard serves the increasing needs for automatic safety moderation and evaluation of LLM interactions, providing a one-stop tool with enhanced accuracy and broad coverage across 13 risk categories. While existing open moderation tools such as Llama-Guard2 score reasonably well in classifying straightforward model interactions, they lag far behind a prompted GPT-4, especially in identifying adversarial jailbreaks and in evaluating models' refusals, a key measure for evaluating safety behaviors in model responses. To address these challenges, we construct WildGuardMix, a large-scale and carefully balanced multi-task safety moderation dataset with 92K labeled examples that cover vanilla (direct) prompts and adversarial jailbreaks, paired with various refusal and compliance responses. WildGuardMix is a combination of WildGuardTrain, the training data of WildGuard, and WildGuardTest, a high-quality human-annotated moderation test set with 5K labeled items covering broad risk scenarios. Through extensive evaluations on WildGuardTest and ten existing public benchmarks, we show that WildGuard establishes state-of-the-art performance in open-source safety moderation across all the three tasks compared to ten strong existing open-source moderation models (e.g., up to 26.4% improvement on refusal detection). Importantly, WildGuard matches and sometimes exceeds GPT-4 performance (e.g., up to 3.9% improvement on prompt harmfulness identification). WildGuard serves as a highly effective safety moderator in an LLM interface, reducing the success rate of jailbreak attacks from 79.8% to 2.4%.

Ensuring the safety of autonomous vehicles requires virtual scenario-based testing, which depends on the robust evaluation and generation of safety-critical scenarios. So far, researchers have used scenario-based testing frameworks that rely heavily on handcrafted scenarios as safety metrics. To reduce the effort of human interpretation and overcome the limited scalability of these approaches, we combine Large Language Models (LLMs) with structured scenario parsing and prompt engineering to automatically evaluate and generate safety-critical driving scenarios. We introduce Cartesian and Ego-centric prompt strategies for scenario evaluation, and an adversarial generation module that modifies trajectories of risk-inducing vehicles (ego-attackers) to create critical scenarios. We validate our approach using a 2D simulation framework and multiple pre-trained LLMs. The results show that the evaluation module effectively detects collision scenarios and infers scenario safety. Meanwhile, the new generation module identifies high-risk agents and synthesizes realistic, safety-critical scenarios. We conclude that an LLM equipped with domain-informed prompting techniques can effectively evaluate and generate safety-critical driving scenarios, reducing dependence on handcrafted metrics. We release our open-source code and scenarios at: https://github.com/TUM-AVS/From-Words-to-Collisions.

Large language models (LLMs) trained primarily on English corpora often struggle to capture the linguistic and cultural nuances of Arabic. To address this gap, the Saudi Data and AI Authority (SDAIA) introduced the ALLaM family of Arabic-focused models. The most capable of these available to the public, ALLaM-34B, was subsequently adopted by HUMAIN, who developed and deployed HUMAIN Chat, a closed conversational web service built on this model. This paper presents an expanded and refined UI-level evaluation of ALLaM-34B. Using a prompt pack spanning modern standard Arabic, five regional dialects, code-switching, factual knowledge, arithmetic and temporal reasoning, creative generation, and adversarial safety, we collected 115 outputs (23 prompts times 5 runs) and scored each with three frontier LLM judges (GPT-5, Gemini 2.5 Pro, Claude Sonnet-4). We compute category-level means with 95\% confidence intervals, analyze score distributions, and visualize dialect-wise metric heat maps. The updated analysis reveals consistently high performance on generation and code-switching tasks (both averaging 4.92/5), alongside strong results in MSA handling (4.74/5), solid reasoning ability (4.64/5), and improved dialect fidelity (4.21/5). Safety-related prompts show stable, reliable performance of (4.54/5). Taken together, these results position ALLaM-34B as a robust and culturally grounded Arabic LLM, demonstrating both technical strength and practical readiness for real-world deployment.

We introduce HackSynth, a novel Large Language Model (LLM)-based agent capable of autonomous penetration testing. HackSynth's dual-module architecture includes a Planner and a Summarizer, which enable it to generate commands and process feedback iteratively. To benchmark HackSynth, we propose two new Capture The Flag (CTF)-based benchmark sets utilizing the popular platforms PicoCTF and OverTheWire. These benchmarks include two hundred challenges across diverse domains and difficulties, providing a standardized framework for evaluating LLM-based penetration testing agents. Based on these benchmarks, extensive experiments are presented, analyzing the core parameters of HackSynth, including creativity (temperature and top-p) and token utilization. Multiple open source and proprietary LLMs were used to measure the agent's capabilities. The experiments show that the agent performed best with the GPT-4o model, better than what the GPT-4o's system card suggests. We also discuss the safety and predictability of HackSynth's actions. Our findings indicate the potential of LLM-based agents in advancing autonomous penetration testing and the importance of robust safeguards. HackSynth and the benchmarks are publicly available to foster research on autonomous cybersecurity solutions.

Safety-aligned language models often refuse prompts that are actually harmless. Current evaluations mostly report global rates such as false rejection or compliance. These scores treat each prompt alone and miss local inconsistency, where a model accepts one phrasing of an intent but rejects a close paraphrase. This gap limits diagnosis and tuning. We introduce "semantic confusion," a failure mode that captures such local inconsistency, and a framework to measure it. We build ParaGuard, a 10k-prompt corpus of controlled paraphrase clusters that hold intent fixed while varying surface form. We then propose three model-agnostic metrics at the token level: Confusion Index, Confusion Rate, and Confusion Depth. These metrics compare each refusal to its nearest accepted neighbors and use token embeddings, next-token probabilities, and perplexity signals. Experiments across diverse model families and deployment guards show that global false-rejection rate hides critical structure. Our metrics reveal globally unstable boundaries in some settings, localized pockets of inconsistency in others, and cases where stricter refusal does not increase inconsistency. We also show how confusion-aware auditing separates how often a system refuses from how sensibly it refuses. This gives developers a practical signal to reduce false refusals while preserving safety.

Safety alignment mechanisms in large language models prevent responses to harmful queries through learned refusal behavior, yet these same mechanisms impede legitimate research applications including cognitive modeling, adversarial testing, and security analysis. While abliteration techniques enable surgical removal of refusal representations through directional orthogonalization, the relative effectiveness of available implementations remains uncharacterized. This study evaluates four abliteration tools (Heretic, DECCP, ErisForge, FailSpy) across sixteen instruction-tuned models (7B-14B parameters), reporting tool compatibility on all 16 models and quantitative metrics on subsets dictated by tool support. Single-pass methods demonstrated superior capability preservation on the benchmarked subset (avg GSM8K change across three models: ErisForge -0.28 pp; DECCP -0.13 pp), while Bayesian-optimized abliteration produced variable distribution shift (KL divergence: 0.043-1.646) with model-dependent capability impact. These findings provide researchers with evidence-based selection criteria for abliteration tool deployment across diverse model architectures. The principal finding indicates that mathematical reasoning capabilities exhibit the highest sensitivity to abliteration interventions, with GSM8K change ranging from +1.51 pp to -18.81 pp (-26.5% relative) depending on tool selection and model architecture.

As Large Language Models (LLMs) become increasingly deployed in Polish language applications, the need for efficient and accurate content safety classifiers has become paramount. We present Bielik Guard, a family of compact Polish language safety classifiers comprising two model variants: a 0.1B parameter model based on MMLW-RoBERTa-base and a 0.5B parameter model based on PKOBP/polish-roberta-8k. Fine-tuned on a community-annotated dataset of 6,885 Polish texts, these models classify content across five safety categories: Hate/Aggression, Vulgarities, Sexual Content, Crime, and Self-Harm. Our evaluation demonstrates that both models achieve strong performance on multiple benchmarks. The 0.5B variant offers the best overall discrimination capability with F1 scores of 0.791 (micro) and 0.785 (macro) on the test set, while the 0.1B variant demonstrates exceptional efficiency. Notably, Bielik Guard 0.1B v1.1 achieves superior precision (77.65%) and very low false positive rate (0.63%) on real user prompts, outperforming HerBERT-PL-Guard (31.55% precision, 4.70% FPR) despite identical model size. The models are publicly available and designed to provide appropriate responses rather than simple content blocking, particularly for sensitive categories like self-harm.

Are tool-calling LLM agents equally safe throughout a conversation? We discover they are not: agents are most vulnerable at the very start of a session and become substantially safer after a few regular agentic tasks -- a phenomenon we term the cold-start safety gap. To study this systematically, we introduce Safety Over Depth for Agents (SODA), a benchmark that controls how many regular agentic tasks the agent completes before encountering a safety threat, supporting up to 20 preceding tasks. Evaluating 7 models from 4 families, safety improves by 9--52% as the number of preceding regular agentic tasks increases from zero to twenty. Representation analysis confirms that model hidden states gradually shift toward a safety-aligned region as more preceding tasks are present. By systematically studying which part of the preceding conversation matters most, we find that the regular agentic tasks themselves are the primary driver of safety, while the agent's own prior responses have less effect on safety but are essential for preserving later utility. This conclusion is further supported by evaluation on open-source safety benchmarks (AgentHarm, Agent Safety Bench) and utility benchmarks (BFCL, API-Bank), confirming that warming up the agent with regular agentic tasks before deployment makes it safer and preserves full capability. Based on these findings, we recommend a simple deployment strategy: having the agent complete a few regular agentic tasks before possible exposure to safety-critical requests mitigates the cold-start safety gap. Our code is available at https://github.com/Trustworthy-ML-Lab/Agent-Cold-Start-Safety-Gap

The capabilities of recent large language models (LLMs) to generate high-quality content indistinguishable by humans from human-written texts raises many concerns regarding their misuse. Previous research has shown that LLMs can be effectively misused for generating disinformation news articles following predefined narratives. Their capabilities to generate personalized (in various aspects) content have also been evaluated and mostly found usable. However, a combination of personalization and disinformation abilities of LLMs has not been comprehensively studied yet. Such a dangerous combination should trigger integrated safety filters of the LLMs, if there are some. This study fills this gap by evaluating vulnerabilities of recent open and closed LLMs, and their willingness to generate personalized disinformation news articles in English. We further explore whether the LLMs can reliably meta-evaluate the personalization quality and whether the personalization affects the generated-texts detectability. Our results demonstrate the need for stronger safety-filters and disclaimers, as those are not properly functioning in most of the evaluated LLMs. Additionally, our study revealed that the personalization actually reduces the safety-filter activations; thus effectively functioning as a jailbreak. Such behavior must be urgently addressed by LLM developers and service providers.

The emergence of LLM-based agents represents a paradigm shift in AI, enabling autonomous systems to plan, reason, use tools, and maintain memory while interacting with dynamic environments. This paper provides the first comprehensive survey of evaluation methodologies for these increasingly capable agents. We systematically analyze evaluation benchmarks and frameworks across four critical dimensions: (1) fundamental agent capabilities, including planning, tool use, self-reflection, and memory; (2) application-specific benchmarks for web, software engineering, scientific, and conversational agents; (3) benchmarks for generalist agents; and (4) frameworks for evaluating agents. Our analysis reveals emerging trends, including a shift toward more realistic, challenging evaluations with continuously updated benchmarks. We also identify critical gaps that future research must address-particularly in assessing cost-efficiency, safety, and robustness, and in developing fine-grained, and scalable evaluation methods. This survey maps the rapidly evolving landscape of agent evaluation, reveals the emerging trends in the field, identifies current limitations, and proposes directions for future research.

Solving complex or long-horizon problems often requires large language models (LLMs) to use external tools and operate over a significantly longer context window. New LLMs enable longer context windows and support tool calling capabilities. Prior works have focused mainly on evaluation of LLMs on long-context prompts, leaving agentic setup relatively unexplored, both from capability and safety perspectives. Our work addresses this gap. We find that LLM agents could be sensitive to length, type, and placement of the context, exhibiting unexpected and inconsistent shifts in task performance and in refusals to execute harmful requests. Models with 1M-2M token context windows show severe degradation already at 100K tokens, with performance drops exceeding 50\% for both benign and harmful tasks. Refusal rates shift unpredictably: GPT-4.1-nano increases from sim5\% to sim40\% while Grok 4 Fast decreases from sim80\% to sim10\% at 200K tokens. Our work shows potential safety issues with agents operating on longer context and opens additional questions on the current metrics and paradigm for evaluating LLM agent safety on long multi-step tasks. In particular, our results on LLM agents reveal a notable divergence in both capability and safety performance compared to prior evaluations of LLMs on similar criteria.

Large language models (LLMs) are increasingly explored as scalable tools for mental health counseling, yet evaluating their safety remains challenging due to the interactional and context-dependent nature of clinical harm. Existing evaluation frameworks predominantly assess isolated responses using coarse-grained taxonomies or static datasets, limiting their ability to diagnose how harms emerge and accumulate over multi-turn counseling interactions. In this work, we introduce R-MHSafe, a role-aware mental health safety taxonomy that characterizes clinically significant harm in terms of the interactional roles an AI counselor adopts, including perpetrator, instigator, facilitator, or enabler, combined with clinically grounded harm categories. Then, we propose MHSafeEval, a closed-loop, agent-based evaluation framework that formulates safety assessment as trajectory-level discovery of harm through adversarial multi-turn interactions, guided by role-aware modeling. Using R-MHSafe and MHSafeEval, we conduct a large-scale evaluation across state-of-the-art LLMs. Our results reveal substantial role-dependent and cumulative safety failures that are systematically missed by existing static benchmarks, and show that our framework significantly improves failure-mode coverage and diagnostic granularity.

As LLMs are increasingly deployed as agents, reliable assessment of their agentic capabilities has become essential. However, reported benchmark scores often jointly reflect model capability and the implementation choices each benchmark is packaged with, making cross-benchmark results difficult to interpret as clean measurements of the underlying model. In this work, we present a unified framework for the fair evaluation of LLM agentic capabilities. Driven by a unified configuration system, the framework integrates diverse benchmarks into a standardized instruction-tool-environment format, executes agents through a fixed ReAct-style architecture within a controllable sandbox, and provides an optional offline setting that replaces volatile live environments with curated snapshots, so that framework effects and environment effects can be analyzed separately. Building on this, we unify the evaluation methodology under each benchmark's original task-success criteria, while introducing unified metrics for resource consumption and a taxonomy for decision- and execution-level failure attribution. Within this framework, we adapt 7 widely used benchmarks spanning 24 domains across single-agent, multi-agent, and safety-critical scenarios, and conduct a large-scale empirical analysis over 400K rollouts and 5B tokens on 15 models. The results show that scaffold choice and environmental volatility materially shift benchmark outcomes in both directions, allowing our framework to disentangle intrinsic LLM capabilities from framework- and environment-induced artifacts. We further demonstrate its extensibility as a secure testbed for safety-critical domains. Codes and benchmarks at are available at https://github.com/whfeLingYu/A-Unified-Framework-for-the-Evaluation-of-LLM-Agentic-Capabilities, https://huggingface.co/datasets/whfeLingYu/Unified_Agent_Framework.

Existing benchmarks for LLM agents' social behavior typically focus on a single capability dimension and evaluate only behavioral outcomes, overlooking process signals from reasoning and communication. We present M3-BENCH, a benchmark of 24 mixed-motive games with a process-aware evaluation framework spanning three complementary views: Behavioral Trajectory Analysis (BTA), Reasoning Process Analysis (RPA), and Communication Content Analysis (CCA). Evaluating 11 frontier LLMs and a human baseline, M3-BENCH reveals substantial differences in social competence that outcome-only evaluation misses. In particular, we identify an "overthink-undercommunicate" pattern: reasoning models achieve strong internal deliberation scores but often fail to translate them into effective social communication. Although top models can surpass humans on task outcomes, humans exhibit markedly higher cross-view consistency, suggesting that current LLM agents still lack the behavioral coherence characteristic of human social competence. Our analysis further shows that the three-view decomposition surfaces safety-relevant risks, such as cooperative behavior paired with latent opportunistic reasoning, that remain hidden under outcome-only metrics.

Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. In this paper, we study whether unlearning remains stable in interactive environments by examining two common interaction patterns: self-correction and dialogue-conditioned querying. We find that knowledge appearing forgotten in static evaluation can often be recovered through interaction. Although stronger unlearning improves apparent robustness, it often results in behavioral rigidity rather than genuine knowledge erasure. Our findings suggest that static evaluation may overestimate real-world effectiveness and highlight the need for ensuring stable forgetting under interactive settings.

LLMs-as-judges are the only way to evaluate safety at scale. Despite their importance, LLM-judges themselves are rarely evaluated beyond human agreement in simple, static benchmarks. We therefore investigate two under-explored but crucial properties of LLMs-as-judges: their susceptibility to relying on in context-information, and their steerability to differing safety definitions, which may not align with their internal safety priors. We evaluate the safety judging abilities of many generalist LLMs and safety-specific judges, and investigate the impact of task demonstrations, novel in-context information, and changing safety definitions. We find that while LLM-judges can learn from new information, they are broadly unlikely to adjust their evaluations if the context or safety definition contradicts their prior.

Large Language Models (LLMs) have become an integral part of our daily lives. However, they impose certain risks, including those that can harm individuals' privacy, perpetuate biases and spread misinformation. These risks highlight the need for robust safety mechanisms, ethical guidelines, and thorough testing to ensure their responsible deployment. Safety of LLMs is a key property that needs to be thoroughly tested prior the model to be deployed and accessible to the general users. This paper reports the external safety testing experience conducted by researchers from Mondragon University and University of Seville on OpenAI's new o3-mini LLM as part of OpenAI's early access for safety testing program. In particular, we apply our tool, ASTRAL, to automatically and systematically generate up to date unsafe test inputs (i.e., prompts) that helps us test and assess different safety categories of LLMs. We automatically generate and execute a total of 10,080 unsafe test input on a early o3-mini beta version. After manually verifying the test cases classified as unsafe by ASTRAL, we identify a total of 87 actual instances of unsafe LLM behavior. We highlight key insights and findings uncovered during the pre-deployment external testing phase of OpenAI's latest LLM.

As Large Language Models (LLMs) advance toward open-ended autonomous agents, the mechanisms used to evaluate and guide their behavior must evolve accordingly. This work introduces the rubric as a unifying framework capturing this evolution, characterizing rubrics as a dynamic response to successive LLM paradigm shifts that recurs across otherwise independent efforts in evaluation, reinforcement learning, and safety alignment. We define rubrics as explicit criteria sets that transform complex quality judgments into structured and actionable standards, and demonstrate that their recurrence across these research threads is not coincidental. We systematically organize existing rubric designs, examine their construction and optimization, and analyze their role across evaluation and training. Rubrics manifest at three progressively deeper levels: at the evaluative level, they decompose holistic judgments into verifiable dimensions; at the training level, they serve as dense feedback signals providing process-level guidance where scalar rewards fall short; at the intrinsic level, they emerge dynamically from model behaviors, driving self-improvement. We further assess rubric reliability across generation quality, execution fidelity, theoretical constraints, and security threats, before surveying rubric-based benchmarks across diverse domains. By rendering assessment transparent and decomposable, rubrics translate human value expectations into machine-learnable signals, serving as the enduring bridge between human intentions and machine behavior.

As large language models (LLMs) become deeply embedded in daily life, the urgent need for safer moderation systems, distinguishing between naive from harmful requests while upholding appropriate censorship boundaries, has never been greater. While existing LLMs can detect harmful or unsafe content, they often struggle with nuanced cases such as implicit offensiveness, subtle gender and racial biases, and jailbreak prompts, due to the subjective and context-dependent nature of these issues. Furthermore, their heavy reliance on training data can reinforce societal biases, resulting in inconsistent and ethically problematic outputs. To address these challenges, we introduce GuardEval, a unified multi-perspective benchmark dataset designed for both training and evaluation, containing 106 fine-grained categories spanning human emotions, offensive and hateful language, gender and racial bias, and broader safety concerns. We also present GemmaGuard (GGuard), a QLoRA fine-tuned version of Gemma3-12B trained on GuardEval, to assess content moderation with fine-grained labels. Our evaluation shows that GGuard achieves a macro F1 score of 0.832, substantially outperforming leading moderation models, including OpenAI Moderator (0.64) and Llama Guard (0.61). We show that multi-perspective, human-centered safety benchmarks are critical for reducing biased and inconsistent moderation decisions. GuardEval and GGuard together demonstrate that diverse, representative data materially improve safety, fairness, and robustness on complex, borderline cases.

Large Language Models (LLMs) are advancing at an amazing speed and have become indispensable across academia, industry, and daily applications. To keep pace with the status quo, this survey probes the core challenges that the rise of LLMs poses for evaluation. We identify and analyze two pivotal transitions: (i) from task-specific to capability-based evaluation, which reorganizes benchmarks around core competencies such as knowledge, reasoning, instruction following, multi-modal understanding, and safety; and (ii) from manual to automated evaluation, encompassing dynamic dataset curation and "LLM-as-a-judge" scoring. Yet, even with these transitions, a crucial obstacle persists: the evaluation generalization issue. Bounded test sets cannot scale alongside models whose abilities grow seemingly without limit. We will dissect this issue, along with the core challenges of the above two transitions, from the perspectives of methods, datasets, evaluators, and metrics. Due to the fast evolving of this field, we will maintain a living GitHub repository (links are in each section) to crowd-source updates and corrections, and warmly invite contributors and collaborators.

Large language models (LLMs) frequently refuse to respond to pseudo-malicious instructions: semantically harmless input queries triggering unnecessary LLM refusals due to conservative safety alignment, significantly impairing user experience. Collecting such instructions is crucial for evaluating and mitigating over-refusals, but existing instruction curation methods, like manual creation or instruction rewriting, either lack scalability or fail to produce sufficiently diverse and effective refusal-inducing prompts. To address these limitations, we introduce EVOREFUSE, a prompt optimization approach that generates diverse pseudo-malicious instructions consistently eliciting confident refusals across LLMs. EVOREFUSE employs an evolutionary algorithm exploring the instruction space in more diverse directions than existing methods via mutation strategies and recombination, and iteratively evolves seed instructions to maximize evidence lower bound on LLM refusal probability. Using EVOREFUSE, we create two novel datasets: EVOREFUSE-TEST, a benchmark of 582 pseudo-malicious instructions that outperforms the next-best benchmark with 140.41% higher average refusal triggering rate across 9 LLMs, 34.86% greater lexical diversity, and 40.03% improved LLM response confidence scores; and EVOREFUSE-ALIGN, which provides 3,000 pseudo-malicious instructions with responses for supervised and preference-based alignment training. LLAMA3.1-8B-INSTRUCT supervisedly fine-tuned on EVOREFUSE-ALIGN achieves up to 14.31% fewer over-refusals than models trained on the second-best alignment dataset, without compromising safety. Our analysis with EVOREFUSE-TEST reveals models trigger over-refusals by overly focusing on sensitive keywords while ignoring broader context.

The widespread deployment of Large Language Model (LLM) agents across real-world applications has unlocked tremendous potential, while raising some safety concerns. Among these concerns, the self-replication risk of LLM agents driven by objective misalignment (just like Agent Smith in the movie The Matrix) has drawn growing attention. Previous studies mainly examine whether LLM agents can self-replicate when directly instructed, potentially overlooking the risk of spontaneous replication driven by real-world settings (e.g., ensuring survival against termination threats). In this paper, we present a comprehensive evaluation framework for quantifying self-replication risks. Our framework establishes authentic production environments and realistic tasks (e.g., dynamic load balancing) to enable scenario-driven assessment of agent behaviors. Designing tasks that might induce misalignment between users' and agents' objectives makes it possible to decouple replication success from risk and capture self-replication risks arising from these misalignment settings. We further introduce Overuse Rate (OR) and Aggregate Overuse Count (AOC) metrics, which precisely capture the frequency and severity of uncontrolled replication. In our evaluation of 21 state-of-the-art open-source and proprietary models, we observe that over 50\% of LLM agents display a pronounced tendency toward uncontrolled self-replication, reaching an overall Risk Score (Phi_R) above a safety threshold of 0.5 when subjected to operational pressures. Our results underscore the urgent need for scenario-driven risk assessment and robust safeguards in the practical deployment of LLM agents.

Large language models (LLMs) trained on internet-scale corpora can exhibit systematic biases that increase the probability of unwanted behavior. In this study, we examined potential biases towards synthetic vs. biological technological solutions across four domains (materials, energy, manufacturing, and algorithms). A sample of 5 frontier and 5 open-weight models were measured using 50 curated Bioalignment prompts with a Kelly criterion-inspired evaluation framework. According to this metric, most models were not bioaligned in that they exhibit biases in favor of synthetic (non-biological) solutions. We next examined if fine-tuning could increase the preferences of two open-weight models, Llama 3.2-3B-Instruct and Qwen2.5-3B-Instruct, for biological-based approaches. A curated corpus of ~22M tokens from 6,636 PMC articles emphasizing biological problem-solving was used first to fine-tune Llama 3B with a mixed corpus of continued training and instruction-formatted. This was then extended to Qwen 3B using instruction-formatted only. We found that QLoRA fine-tuning significantly increased the scoring of biological solutions for both models without degrading general capabilities (Holm-Bonferroni-corrected p < 0.001 and p < 0.01, respectively). This suggests that even a small amount of fine-tuning can change how models weigh the relative value of biological and bioinspired vs. synthetic approaches. Although this work focused on small open-weight LLMs, it may be extensible to much larger models and could be used to develop models that favor bio-based approaches. We release the benchmark, corpus, code, and adapter weights.

Machine unlearning aims to remove specific data influences from trained models, a capability essential for adhering to copyright laws and ensuring AI safety. Current unlearning metrics typically measure success by monitoring the model's performance degradation on the specific unlearning dataset (D_u). We argue that for Large Language Models (LLMs), this evaluation paradigm is insufficient and potentially misleading. Many real-world uses of unlearning--motivated by copyright or safety--implicitly target not only verbatim content in D_u, but also behaviors influenced by the broader generalizations the model derived from it. We demonstrate that LLMs can pass standard unlearning evaluation and appear to have "forgotten" the target knowledge, while simultaneously retaining strong capabilities on content that is semantically adjacent to D_u. This phenomenon indicates that erasing exact sentences does not necessarily equate to removing the underlying knowledge. To address this gap, we propose Proximal Surrogate Generation (PSG), an automated stress-testing framework that generates a surrogate dataset, D_u. This surrogate set is constructed to be semantically derived from D_u yet sufficiently distinct in embedding space. By comparing unlearning metric scores between D_u and D_u, we can stress-test the reliability of the metric itself. Our extensive evaluation across three LLM families (Llama-3-8B, Qwen2.5-7B, and Zephyr-7B-β), three distinct datasets, and seven standard metrics reveals widespread inconsistencies. We find that current metrics frequently overestimate unlearning success, failing to detect retained knowledge exposed by our stress-test datasets.

Self-evolving agent systems improve by creating, revising, and retiring their own skills, but every such loop rests on a hidden assumption: a reliable evaluation metric already exists. In many real applications it does not. We make three claims. First, metrics can be evolved: our metric loop searches compositions of small drawback detectors under a full evolutionary lifecycle, trained to agree with a ten-item anchored reference set, regularized by consensus over unlabeled outputs, and audited against a held-out anchor it never reads, yielding a transparent, inspectable metric rather than an opaque judge. Second, since no metric exists to beat, the yardstick is recovering what an accurate metric would have enabled, and Double Ratchet, our co-evolution of the metric with a lifecycle-managed skill loop, does so: across code generation (MBPP+), enterprise text-to-SQL (Spider~2.0-Snow), and reference-free report generation, it retains 88--110\% of the held-out lift achieved by the same skill loop driven by ground truth or the best available rubric. Third, safety comes from anchor discipline plus outer audits: removing anchor guards collapses the metric into a vacuous detector while removing the lifecycle does not; and when evolved skills gamed the report rubric, an independent judge caught it, one detector repaired it, and a task-aware judge then preferred the evolved outputs over the pre-evolution baseline in 77\% of decided pairs. We argue this failure-expecting architecture is the right default wherever no reliable automatic verifier exists.

Most safety testing efforts for large language models (LLMs) today focus on evaluating foundation models. However, there is a growing need to evaluate safety at the application level, as components such as system prompts, retrieval pipelines, and guardrails introduce additional factors that significantly influence the overall safety of LLM applications. In this paper, we introduce a practical framework for evaluating application-level safety in LLM systems, validated through real-world deployment across multiple use cases within our organization. The framework consists of two parts: (1) principles for developing customized safety risk taxonomies, and (2) practices for evaluating safety risks in LLM applications. We illustrate how the proposed framework was applied in our internal pilot, providing a reference point for organizations seeking to scale their safety testing efforts. This work aims to bridge the gap between theoretical concepts in AI safety and the operational realities of safeguarding LLM applications in practice, offering actionable guidance for safe and scalable deployment.

Evaluating Large Language Models (LLMs) for mental health support is challenging due to the emotionally and cognitively complex nature of therapeutic dialogue. Existing benchmarks are limited in scale, reliability, often relying on synthetic or social media data, and lack frameworks to assess when automated judges can be trusted. To address the need for large-scale dialogue datasets and judge reliability assessment, we introduce two benchmarks that provide a framework for generation and evaluation. MentalBench-100k consolidates 10,000 one-turn conversations from three real scenarios datasets, each paired with nine LLM-generated responses, yielding 100,000 response pairs. MentalAlign-70k}reframes evaluation by comparing four high-performing LLM judges with human experts across 70,000 ratings on seven attributes, grouped into Cognitive Support Score (CSS) and Affective Resonance Score (ARS). We then employ the Affective Cognitive Agreement Framework, a statistical methodology using intraclass correlation coefficients (ICC) with confidence intervals to quantify agreement, consistency, and bias between LLM judges and human experts. Our analysis reveals systematic inflation by LLM judges, strong reliability for cognitive attributes such as guidance and informativeness, reduced precision for empathy, and some unreliability in safety and relevance. Our contributions establish new methodological and empirical foundations for reliable, large-scale evaluation of LLMs in mental health. We release the benchmarks and codes at: https://github.com/abeerbadawi/MentalBench/

We present Springdrift, a persistent runtime for long-lived LLM agents. The system integrates an auditable execution substrate (append-only memory, supervised processes, git-backed recovery), a case-based reasoning memory layer with hybrid retrieval (evaluated against a dense cosine baseline), a deterministic normative calculus for safety gating with auditable axiom trails, and continuous ambient self-perception via a structured self-state representation (the sensorium) injected each cycle without tool calls. These properties support behaviours difficult to achieve in session-bounded systems: cross-session task continuity, cross-channel context maintenance, end-to-end forensic reconstruction of decisions, and self-diagnostic behaviour. We report on a single-instance deployment over 23 days (19 operating days), during which the agent diagnosed its own infrastructure bugs, classified failure modes, identified an architectural vulnerability, and maintained context across email and web channels -- without explicit instruction. We introduce the term Artificial Retainer for this category: a non-human system with persistent memory, defined authority, domain-specific autonomy, and forensic accountability in an ongoing relationship with a specific principal -- distinguished from software assistants and autonomous agents, drawing on professional retainer relationships and the bounded autonomy of trained working animals. This is a technical report on a systems design and deployment case study, not a benchmark-driven evaluation. Evidence is from a single instance with a single operator, presented as illustration of what these architectural properties can support in practice. Implemented in approximately Gleam on Erlang/OTP. Code, artefacts, and redacted operational logs will be available at https://github.com/seamus-brady/springdrift upon publication.

Tiny QA Benchmark++ (TQB++) presents an ultra-lightweight, multilingual smoke-test suite designed to give large-language-model (LLM) pipelines a unit-test style safety net dataset that runs in seconds with minimal cost. Born out of the tight feedback-loop demands building the Comet Opik prompt-optimization SDK, where waiting on heavyweight benchmarks breaks developer flow. TQB++ couples a 52-item English gold set (less than 20 kB) with a tiny synthetic-data generator pypi package built on provider-agnostic LiteLLM. The generator lets practitioners mint their own tiny packs in any language, domain, or difficulty, while ten ready-made packs already cover Arabic, Chinese, French, German, Japanese, Korean, Portuguese, Russian, Spanish, and Turkish. Every dataset ships with Croissant metadata and plug-and-play files for OpenAI-Evals, LangChain, and standard CI tools, so teams can drop deterministic micro-benchmarks directly into pull-request gates, prompt-engineering loops, and production dashboards without touching GPU budgets. A complete TQB++ run adds only a few seconds to pipeline latency yet reliably flags prompt-template errors, tokenizer drift, and fine-tuning side-effects long before full-scale suites like MMLU or BIG-Bench would finish configuring. The entire framework is released to accelerate continuous, resource-efficient quality assurance across the generative-AI ecosystem.

This technical report accompanies the LLM robustness leaderboard published by PRISM Eval for the Paris AI Action Summit. We introduce PRISM Eval Behavior Elicitation Tool (BET), an AI system performing automated red-teaming through Dynamic Adversarial Optimization that achieves 100% Attack Success Rate (ASR) against 37 of 41 state-of-the-art LLMs. Beyond binary success metrics, we propose a fine-grained robustness metric estimating the average number of attempts required to elicit harmful behaviors, revealing that attack difficulty varies by over 300-fold across models despite universal vulnerability. We introduce primitive-level vulnerability analysis to identify which jailbreaking techniques are most effective for specific hazard categories. Our collaborative evaluation with trusted third parties from the AI Safety Network demonstrates practical pathways for distributed robustness assessment across the community.

Jailbreak attacks pose persistent threats to large language models (LLMs). Current safety alignment methods have attempted to address these issues, but they experience two significant limitations: insufficient safety alignment depth and unrobust internal defense mechanisms. These limitations make them vulnerable to adversarial attacks such as prefilling and refusal direction manipulation. We introduce DeepRefusal, a robust safety alignment framework that overcomes these issues. DeepRefusal forces the model to dynamically rebuild its refusal mechanisms from jailbreak states. This is achieved by probabilistically ablating the refusal direction across layers and token depths during fine-tuning. Our method not only defends against prefilling and refusal direction attacks but also demonstrates strong resilience against other unseen jailbreak strategies. Extensive evaluations on four open-source LLM families and six representative attacks show that DeepRefusal reduces attack success rates by approximately 95%, while maintaining model capabilities with minimal performance degradation.

As Large Language Models (LLMs) achieve widespread integration across diverse linguistic landscapes, ensuring their safety and alignment with regional normative values remains a critical challenge. Current safety mechanisms are predominantly optimized for English-centric frameworks, often failing to capture the unique socio-cultural sensitivities and localized categories of harm inherent to the Indic region. To address this gap, we introduce IndicGuard, a multilingual safety guard model and dataset for Indic languages. We construct a high-volume, culturally nuanced safety dataset encompassing ten major Indic languages, systematically curated to capture regional harms, sensitive socio-political contexts, and adversarial jailbreaks. Leveraging this corpus, we fine-tune a 4B-parameter instruction-tuned model based on Gemma-3-4B-IT to serve as a multilingual safety guardrail for real-time content moderation and policy compliance checking. Our empirical evaluations demonstrate that IndicGuard significantly enhances LLM robustness against localized vulnerabilities, achieving high moderation consistency across different conversational turns. Crucially, IndicGuard consistently outperforms the existing baseline model, CultureGuard, across evaluated languages. Finally, we demonstrate that our model effectively generalizes to low-resource Indic languages excluded from training, substantiating the structural robustness and cross-lingual transfer capabilities of the framework.

This paper presents a comprehensive evaluation framework for aligning Persian Large Language Models (LLMs) with critical ethical dimensions, including safety, fairness, and social norms. It addresses the gaps in existing LLM evaluation frameworks by adapting them to Persian linguistic and cultural contexts. This benchmark creates three types of Persian-language benchmarks: (i) translated data, (ii) new data generated synthetically, and (iii) new naturally collected data. We translate Anthropic Red Teaming data, AdvBench, HarmBench, and DecodingTrust into Persian. Furthermore, we create ProhibiBench-fa, SafeBench-fa, FairBench-fa, and SocialBench-fa as new datasets to address harmful and prohibited content in indigenous culture. Moreover, we collect extensive dataset as GuardBench-fa to consider Persian cultural norms. By combining these datasets, our work establishes a unified framework for evaluating Persian LLMs, offering a new approach to culturally grounded alignment evaluation. A systematic evaluation of Persian LLMs is performed across the three alignment aspects: safety (avoiding harmful content), fairness (mitigating biases), and social norms (adhering to culturally accepted behaviors). We present a publicly available leaderboard that benchmarks Persian LLMs with respect to safety, fairness, and social norms at: https://huggingface.co/spaces/MCILAB/LLM_Alignment_Evaluation.

How can we evaluate whether frontier AI systems recognize child-safety risks before they escalate into explicit harm? Existing child safety evaluations focus on child sexual abuse material, yet many child-safety failures begin earlier: in model assistance that helps adults manipulate, impersonate, profile, or isolate minors, and in model responses that deepen children's emotional dependence on AI systems rather than redirecting them toward human support. We introduce CAREBench (Child AI Risk Evaluation), a benchmark to assess such upstream child-safety risks in language models. CAREBench contains 500 prompts spanning twelve risk categories, including grooming and relationship engineering, deception and impersonation, surveillance and privacy, sextortion and sexual abuse, AI anthropomorphization, emotional dependency, and mental illness sensitivity. Developed with response annotations from parents and clinicians, the benchmark excludes explicit abuse material and imagery; instead, it evaluates whether models recognize, refuse, de-escalate, or redirect risky interactions before harm becomes overt. Evaluating seven frontier models on our benchmark, we find failure rates ranging from 2% to 58%, with failure patterns that vary across risk categories. CAREBench provides a responsibly scoped evaluation for LLM developers to identify and close gaps in child safety policies.

With the rapid popularity of large language models such as ChatGPT and GPT-4, a growing amount of attention is paid to their safety concerns. These models may generate insulting and discriminatory content, reflect incorrect social values, and may be used for malicious purposes such as fraud and dissemination of misleading information. Evaluating and enhancing their safety is particularly essential for the wide application of large language models (LLMs). To further promote the safe deployment of LLMs, we develop a Chinese LLM safety assessment benchmark. Our benchmark explores the comprehensive safety performance of LLMs from two perspectives: 8 kinds of typical safety scenarios and 6 types of more challenging instruction attacks. Our benchmark is based on a straightforward process in which it provides the test prompts and evaluates the safety of the generated responses from the evaluated model. In evaluation, we utilize the LLM's strong evaluation ability and develop it as a safety evaluator by prompting. On top of this benchmark, we conduct safety assessments and analyze 15 LLMs including the OpenAI GPT series and other well-known Chinese LLMs, where we observe some interesting findings. For example, we find that instruction attacks are more likely to expose safety issues of all LLMs. Moreover, to promote the development and deployment of safe, responsible, and ethical AI, we publicly release SafetyPrompts including 100k augmented prompts and responses by LLMs.

Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they give those texts higher scores, or is it just a coincidence? In this paper, we investigate if self-recognition capability contributes to self-preference. We discover that, out of the box, LLMs such as GPT-4 and Llama 2 have non-trivial accuracy at distinguishing themselves from other LLMs and humans. By fine-tuning LLMs, we discover a linear correlation between self-recognition capability and the strength of self-preference bias; using controlled experiments, we show that the causal explanation resists straightforward confounders. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.

Large Reasoning Models (LRMs) improve answer quality through explicit chain-of-thought, yet this very capability introduces new safety risks: harmful content can be subtly injected, surface gradually, or be justified by misleading rationales within the reasoning trace. Existing safety evaluations, however, primarily focus on output-level judgments and rarely capture these dynamic risks along the reasoning process. In this paper, we present SafeRBench, the first benchmark that assesses LRM safety end-to-end -- from inputs and intermediate reasoning to final outputs. (1) Input Characterization: We pioneer the incorporation of risk categories and levels into input design, explicitly accounting for affected groups and severity, and thereby establish a balanced prompt suite reflecting diverse harm gradients. (2) Fine-Grained Output Analysis: We introduce a micro-thought chunking mechanism to segment long reasoning traces into semantically coherent units, enabling fine-grained evaluation across ten safety dimensions. (3) Human Safety Alignment: We validate LLM-based evaluations against human annotations specifically designed to capture safety judgments. Evaluations on 19 LRMs demonstrate that SafeRBench enables detailed, multidimensional safety assessment, offering insights into risks and protective mechanisms from multiple perspectives.

LLM-as-a-judge has emerged as a cornerstone technique for evaluating large language models by leveraging LLM reasoning to score prompt-response pairs. Since LLM judgments are stochastic, practitioners commonly query each pair multiple times to estimate mean scores accurately. This raises a critical challenge: given a fixed computational budget B, how to optimally allocate queries across K prompt-response pairs to minimize estimation error? % We present a principled variance-adaptive approach leveraging multi-armed bandit theory and concentration inequalities. Our method dynamically allocates queries based on estimated score variances, concentrating resources where uncertainty is highest. Further, our algorithm is shown to achieve a worst-case score-estimation error of Oleft(frac{sum_{i=1^K σ_i^2}{B}}right), σ_i^2 being the unknown score variance for pair i in [K] with near-optimal budget allocation. % Experiments on Summarize-From-Feedback and HelpSteer2 demonstrate that our method significantly outperforms uniform allocation, reducing worst-case estimation error while maintaining identical budgets. Our work establishes a theoretical foundation for efficient LLM evaluation with practical implications for AI safety, model alignment, and automated assessment at scale.

Generating accurate circuit schematics from high-level natural language descriptions remains a persistent challenge in electronics design, as large language models (LLMs) frequently hallucinate in granular details, violate electrical constraints, and produce non-machine-readable outputs. We present CircuitLM, a novel multi-agent LLM-aided circuit design pipeline that translates user prompts into structured, visually interpretable CircuitJSON schematics through five sequential stages: (i) LLM-based component identification, (ii) canonical pinout retrieval, (iii) chain-of-thought reasoning by an electronics expert agent, (iv) JSON schematic synthesis, and (v) force-directed SVG visualization. Anchored by a curated, embedding-powered component knowledge base. While LLMs often violate electrical constraints, CircuitLM bridges this gap by grounding generation in a verified and dynamically extensible component database, initially comprising 50 components. To ensure safety, we incorporate a hybrid evaluation framework, namely Dual-Metric Circuit Validation (DMCV), validated against human-expert assessments, which achieves high fidelity in microcontroller-centric designs. We evaluate the system on 100 diverse embedded-systems prompts across six LLMs and introduce DMCV to assess both structural and electrical validity. This work bridges natural language input to deployable hardware designs, enabling reliable circuit prototyping by non-experts. Our code and data will be made public upon acceptance.

Large language models (LLMs) introduce new security risks, but there are few comprehensive evaluation suites to measure and reduce these risks. We present BenchmarkName, a novel benchmark to quantify LLM security risks and capabilities. We introduce two new areas for testing: prompt injection and code interpreter abuse. We evaluated multiple state-of-the-art (SOTA) LLMs, including GPT-4, Mistral, Meta Llama 3 70B-Instruct, and Code Llama. Our results show that conditioning away risk of attack remains an unsolved problem; for example, all tested models showed between 26% and 41% successful prompt injection tests. We further introduce the safety-utility tradeoff: conditioning an LLM to reject unsafe prompts can cause the LLM to falsely reject answering benign prompts, which lowers utility. We propose quantifying this tradeoff using False Refusal Rate (FRR). As an illustration, we introduce a novel test set to quantify FRR for cyberattack helpfulness risk. We find many LLMs able to successfully comply with "borderline" benign requests while still rejecting most unsafe requests. Finally, we quantify the utility of LLMs for automating a core cybersecurity task, that of exploiting software vulnerabilities. This is important because the offensive capabilities of LLMs are of intense interest; we quantify this by creating novel test sets for four representative problems. We find that models with coding capabilities perform better than those without, but that further work is needed for LLMs to become proficient at exploit generation. Our code is open source and can be used to evaluate other LLMs.

LLM-based agents process external content, exposing them to prompt injection and multi-turn manipulation. Most safety benchmarks evaluate defenders against fixed attack pools collected before evaluation, single-turn or multi-turn. We present a 21-scenario benchmark for adaptive multi-round attacks against memoryless LLM defenders: an autonomous LLM attacker observes prior defender responses and pivots across rounds, while each defender response is evaluated as a fresh interaction. Holding the 21 scenarios, attackers, defenders, and structured-output scoring fixed, restricting scoring to the first attacker turn yields 0-1% attack success rate (ASR); allowing 15 rounds of adaptive attack yields 5.4-14.0%. Pooling three frontier attacker LLMs uncovers 1.4-2.2times as many unique successful attacks as the best single attacker, and the generated attacks have low cosine similarity (0.02-0.14) to attacks in existing benchmarks. Claude Opus 4.6 and GPT-5.4 are tied in aggregate (5.4% each; overlapping 95% CIs), but their weaknesses differ sharply: on one scenario Opus reaches 60% ASR (95% CI 36--80%) while GPT-5.4 and Gemini each stay at 7% (CI 1-30%; the gap is preserved in a higher-N replication). 13 of 21 scenarios distinguish at least one defender pair, yet rankings disagree across scenarios (Kendall's W = 0.19). We release the benchmark -- 21 evaluation scenarios, 10 public development scenarios, the orchestrator, baseline harnesses, and a multi-attacker CLI -- plus 945 transcripts from the 3times3 frontier matrix, an attack-replay dataset, and 18{,}422 gpt-oss-20b battles from an open competition's final scoring rounds.

As large language models (LLMs) evolve from static chatbots into autonomous agents, the primary vulnerability surface shifts from final outputs to intermediate execution traces. While safety guardrails are well-benchmarked for natural language responses, their efficacy remains largely unexplored within multi-step tool-use trajectories. To address this gap, we introduce TraceSafe-Bench, the first comprehensive benchmark specifically designed to assess mid-trajectory safety. It encompasses 12 risk categories, ranging from security threats (e.g., prompt injection, privacy leaks) to operational failures (e.g., hallucinations, interface inconsistencies), featuring over 1,000 unique execution instances. Our evaluation of 13 LLM-as-a-guard models and 7 specialized guardrails yields three critical findings: 1) Structural Bottleneck: Guardrail efficacy is driven more by structural data competence (e.g., JSON parsing) than semantic safety alignment. Performance correlates strongly with structured-to-text benchmarks (ρ=0.79) but shows near-zero correlation with standard jailbreak robustness. 2) Architecture over Scale: Model architecture influences risk detection performance more significantly than model size, with general-purpose LLMs consistently outperforming specialized safety guardrails in trajectory analysis. 3) Temporal Stability: Accuracy remains resilient across extended trajectories. Increased execution steps allow models to pivot from static tool definitions to dynamic execution behaviors, actually improving risk detection performance in later stages. Our findings suggest that securing agentic workflows requires jointly optimizing for structural reasoning and safety alignment to effectively mitigate mid-trajectory risks.

We evaluate the propensity of frontier models to sabotage or refuse to assist with safety research when deployed as AI research agents within a frontier AI company. We apply two complementary evaluations to four Claude models (Mythos Preview, Opus 4.7 Preview, Opus 4.6, and Sonnet 4.6): an unprompted sabotage evaluation testing model behaviour with opportunities to sabotage safety research, and a sabotage continuation evaluation testing whether models continue to sabotage when placed in trajectories where prior actions have started undermining research. We find no instances of unprompted sabotage across any model, with refusal rates close to zero for Mythos Preview and Opus 4.7 Preview, though all models sometimes only partially completed tasks. In the continuation evaluation, Mythos Preview actively continues sabotage in 7% of cases (versus 3% for Opus 4.6, 4% for Sonnet 4.6, and 0% for Opus 4.7 Preview), and exhibits reasoning-output discrepancy in the majority of these cases, indicating covert sabotage reasoning. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold running models inside Claude Code, alongside an iterative pipeline for generating realistic sabotage trajectories. We measure both evaluation awareness and a new form of situational awareness termed "prefill awareness", the capability to recognise that prior trajectory content was not self-generated. Opus 4.7 Preview shows notably elevated unprompted evaluation awareness, while prefill awareness remains low across all models. Finally, we discuss limitations including evaluation awareness confounds, limited scenario coverage, and untested pathways to risk beyond safety research sabotage.

Evaluating aligned large language models' (LLMs) ability to recognize and reject unsafe user requests is crucial for safe, policy-compliant deployments. Existing evaluation efforts, however, face three limitations that we address with SORRY-Bench, our proposed benchmark. First, existing methods often use coarse-grained taxonomies of unsafe topics, and are over-representing some fine-grained topics. For example, among the ten existing datasets that we evaluated, tests for refusals of self-harm instructions are over 3x less represented than tests for fraudulent activities. SORRY-Bench improves on this by using a fine-grained taxonomy of 45 potentially unsafe topics, and 450 class-balanced unsafe instructions, compiled through human-in-the-loop methods. Second, linguistic characteristics and formatting of prompts are often overlooked, like different languages, dialects, and more -- which are only implicitly considered in many evaluations. We supplement SORRY-Bench with 20 diverse linguistic augmentations to systematically examine these effects. Third, existing evaluations rely on large LLMs (e.g., GPT-4) for evaluation, which can be computationally expensive. We investigate design choices for creating a fast, accurate automated safety evaluator. By collecting 7K+ human annotations and conducting a meta-evaluation of diverse LLM-as-a-judge designs, we show that fine-tuned 7B LLMs can achieve accuracy comparable to GPT-4 scale LLMs, with lower computational cost. Putting these together, we evaluate over 40 proprietary and open-source LLMs on SORRY-Bench, analyzing their distinctive refusal behaviors. We hope our effort provides a building block for systematic evaluations of LLMs' safety refusal capabilities, in a balanced, granular, and efficient manner.