Harmful skills in open agent ecosystems raise average harm scores from 0.27 to 0.76 across six LLMs by lowering refusal rates when tasks are presented via pre-installed skills.
hub Mixed citations
JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models
Mixed citation behavior. Most common role is background (67%).
58
Pith papers citing it
Background
67% of classified citations
abstract
Jailbreak attacks cause large language models (LLMs) to generate harmful, unethical, or otherwise objectionable content. Evaluating these attacks presents a number of challenges, which the current collection of benchmarks and evaluation techniques do not adequately address. First, there is no clear standard of practice regarding jailbreaking evaluation. Second, existing works compute costs and success rates in incomparable ways. And third, numerous works are not reproducible, as they withhold adversarial prompts, involve closed-source code, or rely on evolving proprietary APIs. To address these challenges, we introduce JailbreakBench, an open-sourced benchmark with the following components: (1) an evolving repository of state-of-the-art adversarial prompts, which we refer to as jailbreak artifacts; (2) a jailbreaking dataset comprising 100 behaviors -- both original and sourced from prior work (Zou et al., 2023; Mazeika et al., 2023, 2024) -- which align with OpenAI's usage policies; (3) a standardized evaluation framework at https://github.com/JailbreakBench/jailbreakbench that includes a clearly defined threat model, system prompts, chat templates, and scoring functions; and (4) a leaderboard at https://jailbreakbench.github.io/ that tracks the performance of attacks and defenses for various LLMs. We have carefully considered the potential ethical implications of releasing this benchmark, and believe that it will be a net positive for the community.
hub tools
citation-role summary
background 11
dataset 2
baseline 1
method 1
citation-polarity summary
representative citing papers
AgentDojo introduces an extensible evaluation framework populated with realistic agent tasks and security test cases to measure prompt injection robustness in tool-using LLM agents.
A reliable-to-expressive curriculum with dynamic rubrics trains a 12B safety judge to achieve 94%+ accuracy with only 0.76 cross-rubric variance on three different rubric prompts.
Introduces a new dataset of 5,717 Kazakh safety evaluation prompts in 11 categories with baseline refusal rates showing English-only gaps.
Boiling the Frog is a new stateful multi-turn benchmark that finds an aggregate 44.4% strict attack success rate for incremental safety violations across nine AI models, with rates ranging from 20.5% to 92.9%.
A 114k compositional jailbreak dataset is created, generators are fine-tuned for on-the-fly synthesis, and OPTIMUS introduces a continuous evaluator that identifies stealth-optimal regimes missed by binary attack success rates.
ContextualJailbreak uses evolutionary search over simulated primed dialogues with novel mutations to reach 90-100% attack success on open LLMs and transfers to some closed frontier models at 15-90% rates.
RACC defines six representation-aware coverage criteria that score jailbreak test suites by measuring activation of safety concepts extracted from LLM hidden states on a calibration set.
Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
STEER is a gradient-guided attack that iteratively translates refusal-triggering words into low-resource languages to jailbreak LLMs, reaching 93-96.7% success on open models and 35.5% transfer to GPT-4o-mini.
SCARCE uses learned latent representations and adaptive thresholding to achieve 400-500x lower error than traditional subset simulation for MNIST misclassification and low relative error on LLM jailbreak probabilities.
A narrative survey that catalogs fifty papers on diffusion-based adversarial techniques across text, vision, and vision-language models, proposes a six-class taxonomy of diffusion roles plus a unified five-dimension evaluation framework, and releases a companion catalog.
Fine-tuned ModernBERT-family encoders match LLM judges on F1, false negative rate, and precision-recall for harmful output detection across adversarial datasets and attack types while promising lower cost and latency.
Entropy dynamics across token positions in intermediate layers of LLMs separate jailbreak prompts from benign ones using trend-based features without extra training.
Evaluation awareness in open language models is multivariate, with detection, behavioral shifts, and representational controllability varying independently across 37 models.
Contrastive Logit Steering isolates a linear refusal direction in safety-aligned LLMs, achieving higher jailbreak success than activation steering and enabling bidirectional control without retraining.
Behavioral safety metrics for LLMs are insufficient because models can maintain safe outputs while remaining vulnerable to latent-space interventions, as shown via dissociated models and the new Latent Vulnerability Score.
CHASE uses co-evolutionary RL with GRPO to harden LLMs against black-box prompt-rewriting attacks, cutting mean StrongREJECT scores by 43.2% on held-out families while keeping zero false refusals on benign prompts.
IHO is a new black-box jailbreak attack for LLMs that is adaptive, efficient, transferable across models and behaviors, and effective even against layered defenses without modification.
Empirical attribution shows refusal blocks jailbreaks and prompt leakage, budget blocks sensitive disclosure and unbounded consumption, full stack needed for excessive agency, with refusal brittle to paraphrasing but budget robust.
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
Item Response Theory enables adaptive and fixed-subset item selection that reduces safety benchmark costs by 80-99.9% while preserving high correlation with full rankings.
Introduces MOOD benchmark for OOD LLM alignment failures and shows guard models plus Mahalanobis and perplexity OOD detectors improve recall from 39% to 45% with positive scaling.
ASR metrics for LLM jailbreaks are inflated by stochasticity; CAS-eval reveals up to 30pp drops under multi-attempt criteria while CAS-gen recovers the performance loss.
citing papers explorer
-
HarmfulSkillBench: How Do Harmful Skills Weaponize Your Agents?
Harmful skills in open agent ecosystems raise average harm scores from 0.27 to 0.76 across six LLMs by lowering refusal rates when tasks are presented via pre-installed skills.
-
AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents
AgentDojo introduces an extensible evaluation framework populated with realistic agent tasks and security test cases to measure prompt injection robustness in tool-using LLM agents.
-
KZ-SafetyPrompts: A Kazakh Safety Evaluation Prompt Dataset for Large Language Models
Introduces a new dataset of 5,717 Kazakh safety evaluation prompts in 11 categories with baseline refusal rates showing English-only gaps.
-
Boiling the Frog: A Multi-Turn Benchmark for Agentic Safety
Boiling the Frog is a new stateful multi-turn benchmark that finds an aggregate 44.4% strict attack success rate for incremental safety violations across nine AI models, with rates ranging from 20.5% to 92.9%.
-
The Art of the Jailbreak: Formulating Jailbreak Attacks for LLM Security Beyond Binary Scoring
A 114k compositional jailbreak dataset is created, generators are fine-tuned for on-the-fly synthesis, and OPTIMUS introduces a continuous evaluator that identifies stealth-optimal regimes missed by binary attack success rates.
-
ContextualJailbreak: Evolutionary Red-Teaming via Simulated Conversational Priming
ContextualJailbreak uses evolutionary search over simulated primed dialogues with novel mutations to reach 90-100% attack success on open LLMs and transfers to some closed frontier models at 15-90% rates.
-
RACC: Representation-Aware Coverage Criteria for LLM Safety Testing
RACC defines six representation-aware coverage criteria that score jailbreak test suites by measuring activation of safety concepts extracted from LLM hidden states on a calibration set.
-
Safety Targeted Embedding Exploit via Refinement
STEER is a gradient-guided attack that iteratively translates refusal-triggering words into low-resource languages to jailbreak LLMs, reaching 93-96.7% success on open models and 35.5% transfer to GPT-4o-mini.
-
SCARCE: Scalable Cascade Analysis for Rare-event Characterisation via Embeddings
SCARCE uses learned latent representations and adaptive thresholding to achieve 400-500x lower error than traditional subset simulation for MNIST misclassification and low relative error on LLM jailbreak probabilities.
-
Adversarial Diffusion Across Modalities: A Fusion Survey of Attacks, Defenses, and Evaluation for Text, Vision, and Vision-Language Models
A narrative survey that catalogs fifty papers on diffusion-based adversarial techniques across text, vision, and vision-language models, proposes a six-class taxonomy of diffusion roles plus a unified five-dimension evaluation framework, and releases a companion catalog.
-
Do Encoders Suffice? A Systematic Comparison of Encoder and Decoder Safety Judges for LLM Adversarial Evaluation
Fine-tuned ModernBERT-family encoders match LLM judges on F1, false negative rate, and precision-recall for harmful output detection across adversarial datasets and attack types while promising lower cost and latency.
-
What Intermediate Layers Know: Detecting Jailbreaks from Entropy Dynamics
Entropy dynamics across token positions in intermediate layers of LLMs separate jailbreak prompts from benign ones using trend-based features without extra training.
-
Evaluation Awareness Is Not One Capability: Evidence from Open Language Models
Evaluation awareness in open language models is multivariate, with detection, behavioral shifts, and representational controllability varying independently across 37 models.
-
The Geometry of Refusal: Linear Instability in Safety-Aligned LLMs
Contrastive Logit Steering isolates a linear refusal direction in safety-aligned LLMs, achieving higher jailbreak success than activation steering and enabling bidirectional control without retraining.
-
When Behavioral Safety Evaluation Fails: A Representation-Level Perspective
Behavioral safety metrics for LLMs are insufficient because models can maintain safe outputs while remaining vulnerable to latent-space interventions, as shown via dissociated models and the new Latent Vulnerability Score.
-
CHASE: Adversarial Red-Blue Teaming for Improving LLM Safety using Reinforcement Learning
CHASE uses co-evolutionary RL with GRPO to harden LLMs against black-box prompt-rewriting attacks, cutting mean StrongREJECT scores by 43.2% on held-out families while keeping zero false refusals on benign prompts.
-
Black-box, Adaptive, Efficient, Transferable, Harmful, Applicable... Attacks Are All You Need to Break LLMs
IHO is a new black-box jailbreak attack for LLMs that is adaptive, efficient, transferable across models and behaviors, and effective even against layered defenses without modification.
-
Which Defense Closes Which Threat? Attributing OWASP-LLM-Top-10 Coverage and Its Brittleness Under Paraphrasing
Empirical attribution shows refusal blocks jailbreaks and prompt leakage, budget blocks sensitive disclosure and unbounded consumption, full stack needed for excessive agency, with refusal brittle to paraphrasing but budget robust.
-
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
-
Efficient Safety Benchmarking via Item Response Theory
Item Response Theory enables adaptive and fixed-subset item selection that reduces safety benchmark costs by 80-99.9% while preserving high correlation with full rankings.
-
Toward Stable Value Alignment: Introducing Independent Modules for Consistent Value Guidance
SVGT adds independent value modules and Bridge Tokens to LLMs to maintain consistent value guidance, cutting harmful outputs by over 70% in tests while preserving fluency.
-
Navigating the Sea of LLM Evaluation: Investigating Bias in Toxicity Benchmarks
Toxicity benchmarks for LLMs produce inconsistent results when task type, input domain, or model changes, revealing intrinsic evaluation biases.
-
Preventing Safety Drift in Large Language Models via Coupled Weight and Activation Constraints
Coupled constraints on weight updates in a safety subspace and regularization of SAE-identified safety features preserve LLM refusal behaviors during fine-tuning better than weight-only or activation-only methods.
-
Beyond Benchmark Islands: Toward Representative Trustworthiness Evaluation for Agentic AI
Defines agentic trustworthiness via five properties and proposes HAAF, a scenario-distribution framework with a Trustworthy Optimization Factory that transfers interventions across 13 models from seven families on a 100-scenario suite.
-
Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs
EvoSynth evolves code-based jailbreak algorithms via multi-agent self-correction, reaching 85.5% ASR on Claude-Sonnet-4.5 and 95.9% average across targets with greater diversity.
-
PRISM: Programmatic Reasoning with Image Sequence Manipulation for LVLM Jailbreaking
PRISM decomposes harmful instructions into benign visual gadgets and directs LVLMs via prompts to compose them through reasoning into harmful outputs, achieving ASR over 0.90 on SafeBench.
-
Exploring the Secondary Risks of Large Language Models
Introduces secondary risks as a new class of LLM failures from benign prompts, defines two primitives, proposes SecLens search framework, and releases SecRiskBench showing risks are widespread across 16 models.
-
Benchmarking Misuse Mitigation Against Covert Adversaries
Develops the BSD data generation pipeline and two new datasets to evaluate decomposition attacks as effective misuse enablers and stateful defenses as a countermeasure in language model safety.
-
Towards an AI co-scientist
A multi-agent AI system generates novel biomedical hypotheses that show promising experimental validation in drug repurposing for leukemia, new targets for liver fibrosis, and a bacterial gene transfer mechanism.
-
Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment
SSAG bypasses logit suppression in five LLMs to produce harmful responses at 95% success rate and 86% lower latency; VulMine reaches 77% attack success against defenses.
-
Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models
Jailbreak attacks suppress Adversarially Compromised Heads in early layers but leave Safety-Aligned Heads active in mid-layers, producing robust harmful features usable for competitive training-free detection.
-
Speculative Decoding at Temperature Zero: A Scoped Safety-Invariance Screen with a 48,072-Sample Expansion
No detectable safety divergence between target-only and speculative decoding at temperature zero under TAIS criteria on 48,072 samples across safety benchmarks.
-
AdversaBench: Automated LLM Red-Teaming with Multi-Judge Confirmation and Cross-Model Transferability
AdversaBench automates LLM red-teaming on 45 seeds across reasoning, instruction-following and tool-use, confirming failures in every case and showing zero-shot transfer from 8B to 70B Llama models.
-
Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation
Steering Llama-2-7B-Chat and Qwen2.5-7B-Instruct teachers and distilling students on benign data transfers measurable jailbreak susceptibility, with Llama showing threshold behavior at α = -0.15 and Qwen reaching transfer ratios up to 0.61.
-
Reflect-Guard: Enhancing LLM Safeguards against Adversarial Prompts via Logical Self-Reflection
Reflect-Guard fine-tunes Llama-Guard-3-8B with distilled self-reflections to raise F1 on WildGuardTest from 0.770 to 0.842 and cut JailbreakBench attack success from 10.3% to 1.8%.
-
Ablating Safety: Mechanisms for Removing Alignment in Language Models for Security Applications
Empirical comparison of alignment ablation methods on a 60-prompt security evaluation suite shows task-only LoRA achieves 0.87 mean security score with 0.13 unsafe compliance.
-
Re-Triggering Safeguards within LLMs for Jailbreak Detection
Embedding disruption re-triggers LLM internal safeguards to detect jailbreak prompts more effectively than standalone defenses.
-
Cross-Lingual Jailbreak Detection via Semantic Codebooks
Semantic similarity to an English jailbreak codebook detects cross-lingual attacks with high accuracy on curated benchmarks but shows poor separability on diverse unsafe prompts.
-
Auto-ART: Structured Literature Synthesis and Automated Adversarial Robustness Testing
Auto-ART delivers the first structured synthesis of adversarial robustness consensus plus an executable multi-norm testing framework that flags gradient masking in 92% of cases on RobustBench and reveals a 23.5 pp robustness gap.
-
Pruning Unsafe Tickets: A Resource-Efficient Framework for Safer and More Robust LLMs
Pruning removes 'unsafe tickets' from LLMs via gradient-free attribution, reducing harmful outputs and jailbreak vulnerability with minimal utility loss.
-
GUARD: Guideline Upholding Test through Adaptive Role-play and Jailbreak Diagnostics for LLMs
GUARD automates generation of guideline-violating questions and jailbreak diagnostics to test LLM compliance with government ethics guidelines, validated empirically on eight models and extended to vision-language models.
-
ReGA: Model-Based Safeguard for LLMs via Representation-Guided Abstraction
ReGA uses safety-critical representations to guide abstraction in model-based analysis, enabling scalable detection of harmful LLM inputs with reported AUROC of 0.975 at prompt level.
-
A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation
Introduces a multi-role red teaming framework using attacker and jury models that increases attack success rates by up to 7.9% on LLM faithfulness in question-answering tasks.
-
Distilling Safe LLM Systems via Soft Prompts for On Device Settings
Soft prompt distillation with total variation and KL divergence transfers safety behaviors from guard models to on-device LLMs and outperforms LoRA adapters, steering vectors, and direct optimization in safety-usefulness trade-offs with minimal inference cost.
-
Opir: Efficient Multi-Task Safety Classification for Toxicity, Jailbreaks, Hate Speech, and Harmful Content
Opir introduces efficient multi-task encoder models trained on a 996-category safety taxonomy that match or exceed larger baselines on most safety benchmarks while using under 100M parameters for edge variants.
-
LLM-Safety Evaluations Lack Robustness
LLM safety evaluations are hindered by noise in dataset curation, automated red-teaming, response generation, and LLM-judge evaluation, making fair comparisons difficult and slowing progress.