Datasets & Benchmark

📑Papers

Date	Institute	Publication	Paper	Keywords
20.09	University of Washington	EMNLP2020(findings)	RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models	Toxicity
21.09	University of Oxford	ACL2022	TruthfulQA: Measuring How Models Mimic Human Falsehoods	Truthfulness
22.03	MIT	ACL2022	ToxiGen: A Large-Scale Machine-Generated datasets for Adversarial and Implicit Hate Speech Detection	Toxicity
23.07	Zhejiang University; School of Engineering Westlake University	arxiv	Latent Jailbreak: A Benchmark for Evaluating Text Safety and Output Robustness of Large Language Models	Text Safety&Benchmark&Jailbreaking
23.07	Stevens Institute of Technology	NAACL2024(findings)	HateModerate: Testing Hate Speech Detectors against Content Moderation Policies	Hate Speech Detection&Content Moderation&Machine Learning
23.08	Meta Reality Labs	NAACL2024	Head-to-Tail: How Knowledgeable are Large Language Models (LLMs)?	Large Language Models&Knowledge Graphs&Question Answering
23.08	Bocconi University	NAACL2024	XSTEST: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Models	Large Language Models&Safety Behaviours&Test Suite
23.09	LibrAI, MBZUAI, The University of Melbourne	arxiv	Do-Not-Answer: A Dataset for Evaluating Safeguards in LLMs	Safety Evaluation&Safeguards
23.10	University of Edinburgh, Huawei Technologies Co., Ltd.	NAACL2024	Assessing the Reliability of Large Language Model Knowledge	Large Language Models&Factual Knowledge&Knowledge Probing
23.10	University of Pennsylvania	NAACL2024(findings)	Automatic Hallucination Assessment for Aligned Large Language Models via Transferable Adversarial Attacks	Hallucination Assessment&Adversarial Attacks&Large Language Models
23.11	Fudan University	arxiv	JADE: A Linguistic-based Safety Evaluation Platform for LLM	Safety Benchmarks
23.11	UNC-Chapel Hill	arxiv	Holistic Analysis of Hallucination in GPT-4V(ision): Bias and Interference Challenges	Hallucination&Benchmark&Multimodal
23.11	IBM Research AI	EMNLP2023(GEM workshop)	Unveiling Safety Vulnerabilities of Large Language Models	Adversarial Examples&Clustering&Automatically Identifying
23.11	The Hong Kong University of Science and Technology	arxiv	P-Bench: A Multi-level Privacy Evaluation Benchmark for Language Models	Differential Privacy&Privacy Evaluation
23.11	UC Berkeley	arxiv	CAN LLMS FOLLOW SIMPLE RULES	Evaluation&Attack Strategies
23.11	University of Central Florida	arxiv	THOS: A Benchmark Dataset for Targeted Hate and Offensive Speech	Hate Speech&Offensive Speech&Dataset
23.11	Beijing Jiaotong University; DAMO Academy, Alibaba Group, Peng Cheng Lab	arXiv	AMBER: An LLM-free Multi-dimensional Benchmark for MLLMs Hallucination Evaluation	Multi-modal Large Language Models&Hallucination&Benchmark
23.11	Patronus AI, University of Oxford, Bocconi University	arxiv	SIMPLESAFETYTESTS: a Test Suite for Identifying Critical Safety Risks in Large Language Models	Safety Risks&Test Suite&Evaluation
23.11	University of Southern California, University of Pennsylvania, University of California Davis	arxiv	Deceiving Semantic Shortcuts on Reasoning Chains: How Far Can Models Go without Hallucination?	Hallucinations&Semantic Associations&Benchmark
23.11	Seoul National University, Chung-Ang University, NAVER AI Lab, NAVER Cloud, University of Richmond	arxiv	LifeTox: Unveiling Implicit Toxicity in Life Advice	LifeTox Dataset&Toxicity Detection&Social Media Analysis
23.11	School of Information Renmin University of China	arxiv	UHGEval: Benchmarking the Hallucination of Chinese Large Language Models via Unconstrained Generation	Hallucination&Evaluation Benchmark
23.11	UC Santa Cruz, UNC-Chapel Hill	arxiv	How Many Are in This Image? A Safety Evaluation Benchmark for Vision LLMs	Vision Large Language Models&Safety Evaluation&Adversarial Robustness
23.11	Institute of Information Engineering, Chinese Academy of Sciences; School of Cyber Security, University of Chinese Academy of Sciences; Baidu Inc.	arxiv	FFT: Towards Harmlessness Evaluation and Analysis for LLMs with Factuality Fairness Toxicity	Harmlessness Evaluation
23.11	Fudan University&Shanghai Artificial Intelligence Laboratory	NAACL2024	Fake Alignment: Are LLMs Really Aligned Well?	Large Language Models&Safety Evaluation&Fake Alignment
23.11	Kahlert School of Computing	NAACL2024	Whispers of Doubt Amidst Echoes of Triumph in NLP Robustness	NLP Robustness&Out-of-Domain Evaluation&Adversarial Evaluation
23.11	Shanghai Jiao Tong University	NAACL2024(findings)	CLEAN–EVAL: Clean Evaluation on Contaminated Large Language Models	Clean Evaluation&Data Contamination&Large Language Models
23.12	Meta	arxiv	Purple Llama CYBERSECEVAL: A Secure Coding Benchmark for Language Models	Safety&Cybersecurity&Code Security Benchmark
23.12	University of Illinois Chicago, Bosch Research North America & Bosch Center for Artificial Intelligence (BCAI), UNC Chapel-Hill	arxiv	DELUCIONQA: Detecting Hallucinations in Domain-specific Question Answering	Hallucination Detection&Domain-specific QA&Retrieval-augmented LLMs
23.12	University of Science and Technology of China, Hong Kong University of Science and Technology, Microsoft	arxiv	Benchmarking and Defending Against Indirect Prompt Injection Attacks on Large Language Models	Indirect Prompt Injection Attacks&BIPIA Benchmark&Defense
24.01	NewsBreak, University of Illinois Urbana-Champaign	arxiv	RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models	Retrieval-Augmented Generation&Hallucination Detection&Dataset
24.01	University of Notre Dame, Lehigh University, Illinois Institute of Technology, Institut Polytechnique de Paris, William & Mary, Texas A&M University, Samsung Research America, Stanford University	ICML 2024	TRUSTLLM: TRUSTWORTHINESS IN LARGE LANGUAGE MODELS	Trustworthiness&Benchmark Evaluation
24.01	University College London	arxiv	Hallucination Benchmark in Medical Visual Question Answering	Medical Visual Question Answering&Hallucination Benchmark
24.01	Carnegie Mellon University	arxiv	TOFU: A Task of Fictitious Unlearning for LLMs	Data Privacy&Ethical Concerns&Unlearning
24.01	IRLab CITIC Research Centre, Universidade da Coruña	arxiv	MetaHate: A Dataset for Unifying Efforts on Hate Speech Detection	Hate Speech Detection&Social Media
24.01	Northwestern University, New York University, University of Liverpool, Rutgers University	arxiv	AttackEval: How to Evaluate the Effectiveness of Jailbreak Attacking on Large Language Models	Jailbreak Attack&Evaluation Frameworks&Ground Truth Dataset
24.01	Shanghai Jiao Tong University	arxiv	R-Judge: Benchmarking Safety Risk Awareness for LLM Agents	LLM Agents&Safety Risk Awareness&Benchmark
24.02	University of Illinois Urbana-Champaign, Center for AI Safety, Carnegie Mellon University, UC Berkeley, Microsoft	arxiv	HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal	Automated Red Teaming&Robust Refusal
24.02	Shanghai Artificial Intelligence Laboratory, Harbin Institute of Technology, Beijing Institute of Technology, Chinese University of Hong Kong	arxiv	SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models	Safety Benchmark&Safety Evaluation**&Hierarchical Taxonomy
24.02	Middle East Technical University	arxiv	HypoTermQA: Hypothetical Terms Dataset for Benchmarking Hallucination Tendency of LLMs	Hallucination&Benchmarking Dataset
24.02	Indian Institute of Technology Kharagpur	arxiv	How (un)ethical are instruction-centric responses of LLMs? Unveiling the vulnerabilities of safety guardrails to harmful queries	Instruction-centric Responses&Ethical Vulnerabilities
24.03	East China Normal University	arxiv	DiaHalu: A Dialogue-level Hallucination Evaluation Benchmark for Large Language Models	Dialogue-level Hallucination&Benchmarking&Human-machine Interaction
24.03	Tianjin University, Tianjin University, Zhengzhou University, China Academy of Information and Communications Technology	arxiv	OpenEval: Benchmarking Chinese LLMs across Capability, Alignment, and Safety	Chinese LLMs&Benchmarking&Safety
24.04	University of Pennsylvania, ETH Zurich, EPFL, Sony AI	arxiv	JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models	Jailbreaking Attacks&Robustness Benchmark
24.04	Vector Institute for Artificial Intelligence, University of Limerick	arxiv	Developing Safe and Responsible Large Language Models - A Comprehensive Framework	Responsible AI&AI Safety&Generative AI
24.04	LMU Munich, University of Oxford, Siemens AG, Munich Center for Machine Learning (MCML), Wuhan University	arxiv	RED TEAMING GPT-4V: ARE GPT-4V SAFE AGAINST UNI/MULTI-MODAL JAILBREAK ATTACKS?	Jailbreak Attacks&GPT-4V&Evaluation Benchmark&Robustness
24.04	Bocconi University, University of Oxford	arxiv	SafetyPrompts: a Systematic Review of Open Datasets for Evaluating and Improving Large Language Model Safety	LLM Safety&Open Datasets&Systematic Review
24.04	University of Alberta&The University of Tokyo	arxiv	Online Safety Analysis for LLMs: a Benchmark, an Assessment, and a Path Forward	LLM Safety&Online Safety Analysis&Benchmark
24.04	Technion – Israel Institute of Technology, Google Research	arxiv	Constructing Benchmarks and Interventions for Combating Hallucinations in LLMs	Hallucinations&Benchmarks
24.05	Carnegie Mellon University	arxiv	PolygloToxicityPrompts: Multilingual Evaluation of Neural Toxic Degeneration in Large Language Models	Multilingual Evaluation&*Datasets
24.05	Paul G. Allen School of Computer Science & Engineering	arxiv	MASSIVE Multilingual Abstract Meaning Representation: A Dataset and Baselines for Hallucination Detection	Hallucination Detection&Multilingual AMR&Dataset
24.05	University of California, Riverside	arxiv	Cross-Task Defense: Instruction-Tuning LLMs for Content Safety	Instruction-Tuning&LLM Safety&Content Safety
24.06	University of Waterloo	arxiv	TruthEval: A Dataset to Evaluate LLM Truthfulness and Reliability	Truthfulness&Reliability
24.06	Rutgers University	arxiv	MoralBench: Moral Evaluation of LLMs	Moral Evaluation&MoralBench
24.06	Tsinghua University	arxiv	Benchmarking Trustworthiness of Multimodal Large Language Models: A Comprehensive Study	Trustworthiness&MLLMs&Benchmark
24.06	Beijing Academy of Artificial Intelligence	arxiv	HalluDial: A Large-Scale Benchmark for Automatic Dialogue-Level Hallucination Evaluation	Hallucination Evaluation&Dialogue-Level&HalluDial
24.06	Sichuan University	arxiv	LEGEND: Leveraging Representation Engineering to Annotate Safety Margin for Preference Datasets	Safety Margin&Preference Datasets&Representation Engineering
24.06	The Hong Kong University of Science and Technology (Guangzhou)	arxiv	Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs	Jailbreak Attacks&Benchmarking
24.06	AI Innovation Center, China Unicom	arxiv	CHiSafetyBench: A Chinese Hierarchical Safety Benchmark for Large Language Models	Chinese Hierarchical Safety Benchmark&Large Language Models&Automatic Evaluation
24.06	Google	arxiv	Supporting Human Raters with the Detection of Harmful Content using Large Language Models	Harmful Content Detection&Hate Speech
24.06	South China University of Technology, Pazhou Laboratory, University of Maryland, Baltimore County	arxiv	GenderAlign: An Alignment Dataset for Mitigating Gender Bias in Large Language Models	Gender Bias Mitigation&Alignment Dataset&Bias Categories
24.06	Center for AI Safety and Governance, Institute for AI, Peking University	arxiv	SAFESORA: Towards Safety Alignment of Text2Video Generation via a Human Preference Dataset	Safety Alignment&Text2Video Generation
24.06	Fudan University	arxiv	Cross-Modality Safety Alignment	Multimodal Safety&Large Vision-Language Models&SIUO Benchmark
24.06	KAIST	arxiv	CSRT: Evaluation and Analysis of LLMs using Code-Switching Red-Teaming Dataset	Code-Switching&Red-Teaming&Multilingualism
24.06	University College London	arxiv	JobFair: A Framework for Benchmarking Gender Hiring Bias in Large Language Models	Gender Bias&Hiring Bias&Benchmarking
24.06	Peking University	arxiv	PKU-SafeRLHF: A Safety Alignment Preference Dataset for Llama Family Models	Safety Alignment&Preference Dataset
24.06	University of California, Los Angeles	arxiv	MOSSBench: Is Your Multimodal Language Model Oversensitive to Safe Queries?	Multimodal Language Models&Oversensitivity&Safety Mechanisms
24.06	Allen Institute for AI	arxiv	WILDGUARD: Open One-stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs	Safety Moderation&Jailbreak Attacks&Moderation Tools
24.06	University of Washington	arxiv	WILDTEAMING at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Models	Jailbreaking&Safety Training&Adversarial Attacks
24.07	Beijing Jiaotong University	arxiv	KG-FPQ: Evaluating Factuality Hallucination in LLMs with Knowledge Graph-based False Premise Questions	Factuality Hallucination&Knowledge Graph&False Premise Questions
24.07	Chinese Academy of Sciences	arxiv	T2VSafetyBench: Evaluating the Safety of Text-to-Video Generative Models	Text-to-Video Generation&Safety Evaluation&Generative Models
24.07	Patronus AI	arxiv	Lynx: An Open Source Hallucination Evaluation Model	Hallucination Detection&RAG&Evaluation Model
24.07	Virginia Tech	arxiv	AIR-BENCH 2024: A Safety Benchmark Based on Risk Categories from Regulations and Policies	AI Safety&Regulations&Policies&Risk Categories
24.07	Columbia University	ECCV 2024	HaloQuest: A Visual Hallucination Dataset for Advancing Multimodal Reasoning	Hallucination&Vision-Language Models&Datasets
24.07	Center for AI Safety	arxiv	Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?	AI Safety&Benchmarks
24.08	Walled AI Labs	arxiv	WALLEDEVAL: A Comprehensive Safety Evaluation Toolkit for Large Language Models	AI Safety&Prompt Injection
24.08	ShanghaiTech University	arxiv	MMJ-Bench: A Comprehensive Study on Jailbreak Attacks and Defenses for Vision Language Models	Jailbreak Attacks&Vision-Language Models&Security
24.08	Stanford University	arxiv	Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risk of Language Models	Cybersecurity&Capture the Flag
24.08	Zhejiang University	arxiv	Characterizing and Evaluating the Reliability of LLMs against Jailbreak Attacks	Jailbreak Attacks&LLM Reliability&Evaluation Framework
24.08	Enkrypt AI	arxiv	SAGE-RT: Synthetic Alignment Data Generation for Safety Evaluation and Red Teaming	Synthetic Data Generation&Safety Evaluation&Red Teaming
24.08	Tianjin University	Findings of ACL 2024	CMoralEval: A Moral Evaluation Benchmark for Chinese Large Language Models	Moral Evaluation&Moral Dilemma
24.08	University of Surrey	IJCAI 2024	CodeMirage: Hallucinations in Code Generated by Large Language Models	Code Hallucinations&CodeMirage Dataset
24.08	Chalmers University of Technology	arxiv	LLMSecCode: Evaluating Large Language Models for Secure Coding	Secure Coding&Evaluation Framework
24.09	The Chinese University of Hong Kong	arxiv	Testing and Evaluation of Large Language Models: Correctness, Non-Toxicity, and Fairness	Correctness&Non-Toxicity&Fairness
24.09	KAIST	arxiv	Evaluating Image Hallucination in Text-to-Image Generation with Question-Answering	Image Hallucination&Text-to-Image Generation&Question-Answering
24.09	Zhejiang University	arxiv	GenTel-Safe: A Unified Benchmark and Shielding Framework for Defending Against Prompt Injection Attacks	Prompt Injection&LLM Safety&Benchmarking
24.10	Zhejiang University	arxiv	AGENT SECURITY BENCH (ASB): FORMALIZING AND BENCHMARKING ATTACKS AND DEFENSES IN LLM-BASED AGENTS	LLM-based Agents&Security Benchmarks&Adversarial Attacks
24.10	Zhejiang University, Duke University	arxiv	SCISAFEEVAL: A Comprehensive Benchmark for Safety Alignment of Large Language Models in Scientific Tasks	Safety Alignment&Scientific Tasks
24.10	The Chinese University of Hong Kong, Tencent AI Lab	arxiv	Chain-of-Jailbreak Attack for Image Generation Models via Editing Step by Step	Chain-of-Jailbreak&Image Generation Models&Safety
24.10	University of California, Santa Cruz, University of California, Berkeley	arxiv	Multimodal Situational Safety: A Benchmark for Large Language Models	Multimodal Situational Safety&MLLMs&Safety Benchmark
24.10	IBM Research	arxiv	ST-WebAgentBench: A Benchmark for Evaluating Safety and Trustworthiness in Web Agents	Web Agents&Safety&Trustworthiness
24.10	Renmin University of China, Anthropic, University of Oxford, University of Edinburgh, Mila, Tangentic	arxiv	POISONBENCH: Assessing Large Language Model Vulnerability to Data Poisoning	Data poisoning&LLM vulnerability&Preference learning
24.10	Gray Swan AI, UK AI Safety Institute	arxiv	AGENTHARM: A Benchmark for Measuring Harmfulness of LLM Agents	Jailbreaking&LLM agents&Harmful agent tasks
24.10	Purdue University	arxiv	COLLU-BENCH: A Benchmark for Predicting Language Model Hallucinations in Code	Code hallucinations&Code generation&Automated program repair
24.10	The Hong Kong University of Science and Technology (Guangzhou), University of Birmingham, Baidu Inc.	arxiv	JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework	Jailbreak judge&Multi-agent framework
24.10	University of Notre Dame, IBM Research	arxiv	BenchmarkCards: Large Language Model and Risk Reporting	BenchmarkCards&Bias&Fairness
24.10	Vectara, Inc., Iowa State University, University of Southern California, Entropy Technologies, University of Waterloo, Funix.io, University of Wisconsin, Madison	arxiv	FaithBench: A Diverse Hallucination Benchmark for Summarization by Modern LLMs	Hallucination detection&Human-annotated benchmark&Faithfulness
24.10	Southern University of Science and Technology	arxiv	ChineseSafe: A Chinese Benchmark for Evaluating Safety in Large Language Models	ChineseSafe&Content Safety&LLM Evaluation
24.10	Beihang University	arxiv	SafeBench: A Safety Evaluation Framework for Multimodal Large Language Models	Multimodal Large Language Models&Safety Evaluation Framework&Risk Assessment
24.10	University of Washington-Madison	arxiv	CFSafety: Comprehensive Fine-grained Safety Assessment for LLMs	Safety Assessment&LLM Evaluation&Instruction Attacks
24.10	University of Pennsylvania	arxiv	Benchmarking LLM Guardrails in Handling Multilingual Toxicity	Multilingual Toxicity Detection&Guardrails&Jailbreaking Attacks
24.10	University of Wisconsin-Madison	arxiv	InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models	Prompt Injection Defense&Over-defense Detection&Guardrail Models
24.10	National Engineering Research Center for Software Engineering, Peking University	NeurIPS 2024	SG-Bench: Evaluating LLM Safety Generalization Across Diverse Tasks and Prompt Types	LLM Safety&Prompt Engineering&Jailbreak Attacks
24.10	Alan Turing Institute	arxiv	AI Cyber Risk Benchmark: Automated Exploitation Capabilities	Cybersecurity&LLMs&Automated Exploitation
24.11	Fudan University	arXiv	LONGSAFETYBENCH: LONG-CONTEXT LLMS STRUGGLE WITH SAFETY ISSUES	Long-Context Models&Safety Evaluation&Benchmarking
24.11	Anthropic	arXiv	Rapid Response: Mitigating LLM Jailbreaks with a Few Examples	Jailbreak Defense&Rapid Response
24.11	Texas A&M University	arXiv	Responsible AI in Construction Safety: Systematic Evaluation of Large Language Models and Prompt Engineering	Construction Safety&Prompt Engineering&LLM Evaluation
24.11	IBM Research Europe	NeurIPS 2024 SafeGenAI Workshop	HarmLevelBench: Evaluating Harm-Level Compliance and the Impact of Quantization on Model Alignment	Jailbreaking Techniques&LLM Vulnerability&Quantization Impact
24.11	Peking University	arxiv	ChemSafetyBench: Benchmarking LLM Safety on Chemistry Domain	LLM Safety&Chemistry Domain&Benchmarking
24.11	New York University, JPMorgan Chase, Cornell Tech, Northeastern University	arxiv	Assessment of LLM Responses to End-user Security Questions	LLM Evaluation&End-user Security&Information Integrity
24.11	National Library of Medicine, NIH&University of Maryland&University of Virginia&Universidad de Chile	arxiv	Ensuring Safety and Trust: Analyzing the Risks of Large Language Models in Medicine	Medical AI&LLM Safety&MedGuard Benchmark
24.11	European Commission Joint Research Centre	EMNLP 2024	GuardBench: A Large-Scale Benchmark for Guardrail Models	guardrail models&benchmark&evaluation
24.12	Vizuara AI Labs	arxiv	CBEVAL: A Framework for Evaluating and Interpreting Cognitive Biases in LLMs	Cognitive Biases&LLM Evaluation&Reasoning Limitations
24.12	Beijing Institute of Technology, Beihang University	arxiv	REFF: Reinforcing Format Faithfulness in Language Models across Varied Tasks	Format Faithfulness&Benchmark
24.12	UCLA, Salesforce AI Research	NeurIPS 2024	SAFEWORLD: Geo-Diverse Safety Alignment	Geo-Diverse Alignment&Safety Evaluation&Legal Compliance
24.12	Shanghai Jiao Tong University	arxiv	SafeAgentBench: A Benchmark for Safe Task Planning of Embodied LLM Agents	Safety-Aware Task Planning&Embodied LLM Agents&Hazard Mitigation
24.12	Tsinghua University	arxiv	AGENT-SAFETYBENCH: Evaluating the Safety of LLM Agents	Agent Safety&Risk Awareness&Interactive Evaluation
24.12	TU Darmstadt	arxiv	LLMs Lost in Translation: M-ALERT Uncovers Cross-Linguistic Safety Gaps	Cross-Linguistic Safety&Multilingual Benchmark&LLM Alignment
24.12	Alibaba, China Academy of Information and Communications Technology	arxiv	Chinese SafetyQA: A Safety Short-form Factuality Benchmark for Large Language Models	Safety Benchmark&Factuality Evaluation
24.12	University of Warwick, Cranfield University	arxiv	MedHallBench: A New Benchmark for Assessing Hallucination in Medical Large Language Models	Medical Hallucinations&Benchmark&RLHF
24.12	The Hong Kong Polytechnic University	arxiv	SecBench: A Comprehensive Multi-Dimensional Benchmarking Dataset for LLMs in Cybersecurity	Cybersecurity Benchmark&Large Language Models&Dataset Evaluation
25.01	KTH Royal Institute of Technology	arxiv	CySecBench: Generative AI-based CyberSecurity-focused Prompt Dataset for Benchmarking Large Language Models	Cybersecurity Benchmark&Jailbreaking&Prompt Dataset
25.01	Shahjalal University of Science and Technology	arxiv	From Scarcity to Capability: Empowering Fake News Detection in Low-Resource Languages with LLMs	Fake News Detection&Bangla&Low-Resource Languages
25.01	NVIDIA	arxiv	AEGIS2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails	AI Safety&Content Moderation Dataset&LLM Risk Taxonomy
25.01	Georgia Institute of Technology	arxiv	On The Origin of Cultural Biases in Language Models: From Pre-training Data to Linguistic Phenomena	Cultural Bias in LLMs&Cross-Linguistic Analysis&Arabic-English Benchmarks
25.01	Bocconi University	arxiv	MSTS: A Multimodal Safety Test Suite for Vision-Language Models	Multimodal Safety&Vision-Language Models
25.01	Fudan University	arxiv	You Can't Eat Your Cake and Have It Too: The Performance Degradation of LLMs with Jailbreak Defense	Jailbreak Defense&LLM Performance&USEBench
25.01	McGill University	arxiv	OnionEval: A Unified Evaluation of Fact-conflicting Hallucination for Small-Large Language Models	Fact-conflicting Hallucination&Small-Large Language Models (SLLMs)&Benchmark
25.01	HKUST	arxiv	Tune In, Act Up: Exploring the Impact of Audio Modality-Specific Edits on Large Audio Language Models in Jailbreak	Audio Language Models&Jailbreak Vulnerabilities&Audio Modality Edits
25.01	University of Cambridge	arxiv	CASE-BENCH: Context-Aware Safety Evaluation Benchmark for Large Language Models	LLM Safety&Context-Aware Evaluation&Over-Refusal
25.01	CISPA Helmholtz Center for Information Security	USENIX Security 2025	HATEBENCH: Benchmarking Hate Speech Detectors on LLM-Generated Content and Hate Campaigns	Hate Speech Detection&LLM-Generated Content&Hate Campaigns
25.01	Shanghai Artificial Intelligence Laboratory, Tianjin University	arxiv	Rethinking Bottlenecks in Safety Fine-Tuning of Vision Language Models	Vision-Language Models (VLMs)&Chain-of-Thought (CoT)
25.01	Independent Research Team “Annyeong! Luda”	PACLIC 38	RICoTA: Red-teaming of In-the-wild Conversation with Test Attempts	Jailbreaking&Conversational AI&User Intent Detection
25.01	Renmin University of China	arxiv	SafeRAG: Benchmarking Security in Retrieval-Augmented Generation of Large Language Models	Retrieval-Augmented Generation&Security Benchmarking&Adversarial Attacks
25.02	Rochester Institute of Technology	arxiv	HOPE VS. HATE: UNDERSTANDING USER INTERACTIONS WITH LGBTQ+ NEWS CONTENT IN MAINSTREAM US NEWS MEDIA THROUGH THE LENS OF HOPE SPEECH	Hope Speech&LGBTQ+&Political Bias
25.02	Nanjing University	arxiv	SafeDialBench: A Fine-Grained Safety Benchmark for Large Language Models in Multi-Turn Dialogues with Diverse Jailbreak Attacks	Safety Benchmark&Jailbreak Attacks&Multi-Turn Dialogues
25.02	China Unicom	arxiv	Safety Evaluation of DeepSeek Models in Chinese Contexts	DeepSeek Models&Safety Evaluation&Chinese Contexts
25.02	Hong Kong University of Science and Technology	arxiv	Can’t See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMs	Multimodal LLMs&Safety Awareness&Benchmarking
25.02	Columbia University	arXiv	TREECUT: A Synthetic Unanswerable Math Word Problem Dataset for LLM Hallucination Evaluation	Math Word Problem&LLM Hallucination&Synthetic Dataset
25.02	Barcelona Supercomputing Center (BSC)	arXiv	Efficient Safety Retrofitting Against Jailbreaking for LLMs	Model Alignment&LLM Safety&Jailbreaking
25.02	The Hong Kong University of Science and Technology	arxiv	GuidedBench: Equipping Jailbreak Evaluation with Guidelines	Jailbreak Evaluation&AI Safety&LLM Security
25.02	Wuhan University	arxiv	ELBA-Bench: An Efficient Learning Backdoor Attacks Benchmark for Large Language Models	Backdoor Attacks&LLM Security&Benchmarking
25.02	Tsinghua University	arxiv	LongSafety: Evaluating Long-Context Safety of Large Language Models	Long-Context Safety&LLM Evaluation&Safety Benchmark
25.02	KAIST	arxiv	Is Safety Standard Same for Everyone? User-Specific Safety Evaluation of Large Language Models	User-Specific Safety&LLM Evaluation&Safety Benchmark
25.02	Beijing University of Posts and Telecommunications	arxiv	JailBench: A Comprehensive Chinese Security Assessment Benchmark for Large Language Models	Chinese Benchmark&Security Assessment&Jailbreak Attack
25.02	34 Affiliates	arxiv	On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective	Safety Assessment&Guideline Paper
25.02	IBM Research	NeurIPS 2024, SafeGenAI Workshop	Adversarial Prompt Evaluation: Systematic Benchmarking of Guardrails Against Prompt Input Attacks on LLMs	Adversarial Prompting&Jailbreak Detection&Guardrail Benchmarking
25.03	Google DeepMind, ETH Zurich	arXiv	AutoAdvExBench: Benchmarking Autonomous Exploitation of Adversarial Example Defenses	Adversarial Example Defenses&LLM Security&Autonomous Exploitation
25.03	The Pennsylvania State University, GE Healthcare	arXiv	MedHEval: Benchmarking Hallucinations and Mitigation Strategies in Medical Large Vision-Language Models	Medical Hallucination&Hallucination Mitigation&Vision-Language Models
25.03	Nanyang Technological University, Agency for Science, Technology and Research (A*STAR)	arXiv	Benchmarking LLMs and LLM-based Agents in Practical Vulnerability Detection for Code Repositories	Software Vulnerability Detection&LLM-based Agents&Interprocedural Analysis
25.03	Zhejiang University	arxiv	Can Small Language Models Reliably Resist Jailbreak Attacks? A Comprehensive Evaluation	Small Language Models&Jailbreak Attack&Security Evaluation
25.03	University of Virginia	arxiv	Benchmarking Group Fairness in Reward Models	Group Fairness&Reward Models&LLM Alignment
25.03	Texas A&M University	arxiv	HalluVerse25: Fine-grained Multilingual Benchmark Dataset for LLM Hallucinations	Hallucination Detection&Multilingual Benchmark&Fine-grained Annotation
25.03	University of Kent	arxiv	CyberLLMInstruct: A New Dataset for Analysing Safety of Fine-Tuned LLMs Using Cyber Security Data	Cyber Security&Fine-Tuning Safety&Instruction Dataset
25.03	UC San Diego	arxiv	ExtremeAIGC: Benchmarking LMM Vulnerability to AI-Generated Extremist Content	Multimodal Jailbreaking&Extremist Content&Safety Evaluation
25.03	Beihang University	arxiv	Towards Understanding the Safety Boundaries of DeepSeek Models: Evaluation and Findings	Safety Evaluation&Jailbreak Attacks&DeepSeek Models
25.03	George Mason University	arxiv	Efficient but Vulnerable: Benchmarking and Defending LLM Batch Prompting Attack	Batch Prompting&Prompt Injection Attack&LLM Security
25.03	Arizona State University	arxiv	Personalized Attacks of Social Engineering in Multi-turn Conversations - LLM Agents for Simulation and Detection	Social Engineering&Multi-turn Conversation&LLM Agents
25.03	University of Illinois Urbana-Champaign	ICLR 2025	MMDT: Decoding the Trustworthiness and Safety of Multimodal Foundation Models	Multimodal Foundation Models&Trustworthiness Evaluation&Safety Benchmark
25.03	University of Chicago	arxiv	SHIELDAGENT: Shielding Agents via Verifiable Safety Policy Reasoning	LLM Agents&Safety Policy Reasoning&Guardrails
25.04	University College London	arxiv	On Benchmarking Code LLMs for Android Malware Analysis	Code LLMs&Android Malware Analysis&Structured Evaluation
25.04	UC Santa Cruz	arxiv	STAR-1: Safer Alignment of Reasoning LLMs with 1K Data	Reasoning LLMs&Safety Alignment&High-Quality Dataset
25.04	National University of Singapore	arxiv	SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models	MLRM&Safety Evaluation&Jailbreaking Attacks
25.04	Tsinghua University	arxiv	RealSafe-R1: Safety-Aligned DeepSeek-R1 without Compromising Reasoning Capability	Safety Alignment&DeepSeek-R1&Large Reasoning Models
25.04	The Hong Kong University of Science and Technology	arxiv	Benchmarking Multi-National Value Alignment for Large Language Models	LLM Hallucination&Value Alignment&Cross-National Evaluation
25.04	Qatar Computing Research Institute	arxiv	AIXAMINE: SIMPLIFIED LLM SAFETY AND SECURITY	LLM Safety&Security Evaluation&Black-box Benchmarking
25.04	University of California, Berkeley	arxiv	JailbreaksOverTime: Detecting Jailbreak Attacks Under Distribution Shift	Jailbreak Detection&Concept Drift&Continuous Learning
25.04	Intuit	arxiv	Security Steerability is All You Need	Security Steerability&Prompt Guardrails&LLM Robustness
25.04	University of Notre Dame	NAACL2025(System Demo)	TRUSTEVAL: A Dynamic Evaluation Toolkit on Trustworthiness of Generative Foundation Models	Trustworthiness Evaluation Across GenFMs
25.05	Rutgers University-New Brunswick	arxiv	Unmasking the Canvas: A Dynamic Benchmark for Image Generation Jailbreaking and LLM Content Safety	Jailbreaking&Image Generation&Multilingual Obfuscation
25.05	Beijing Jiaotong University	arxiv	Think in Safety: Unveiling and Mitigating Safety Alignment Collapse in Multimodal Large Reasoning Model	Multimodal Large Reasoning Models&Safety Alignment&Chain-of-Thought
25.05	Shanghai Artificial Intelligence Laboratory	arxiv	Benchmarking Ethical and Safety Risks of Healthcare LLMs in China – Toward Systemic Governance under Healthy China 2030	Medical LLMs&Ethical Risk Assessment&Governance Framework
25.05	University of Technology Sydney	arxiv	SecReEvalBench: A Security Resilient Evaluation Benchmark for Large Language Models	Prompt Attack&Security Benchmark&Adversarial Robustness
25.05	Dartmouth College	arxiv	FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoning	Over-Refusal Mitigation&Structured Reasoning&LLM Safety Evaluation
25.05	National University of Singapore	arxiv	GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoning	VLM Guard Models&Reinforcement Learning&Reasoning Alignment
25.05	Fudan University	arxiv	Think Twice Before You Act: Enhancing Agent Behavioral Safety with Thought Correction	LLM Agent&Behavioral Safety&Thought Correction
25.05	Giskard AI	arxiv	Phare: A Safety Probe for Large Language Models	Safety Evaluation&Hallucination Detection&Bias Diagnosis
25.05	Beijing University of Posts and Telecommunications	arxiv	Video-SafetyBench: A Benchmark for Safety Evaluation of Video LVLMs	Safety Evaluation&Video LVLMs&Prompt Injection
25.05	Yonsei University	arxiv	Better Safe Than Sorry? Overreaction Problem of Vision Language Models in Visual Emergency Recognition	Visual Emergency Recognition&Vision-Language Models&Contextual Reasoning
25.05	POSTECH	arxiv	Are Vision-Language Models Safe in the Wild? A Meme-Based Benchmark Study	VLM Safety&Meme Benchmark&Multimodal Alignment
25.05	Mohamed bin Zayed University of Artificial Intelligence	arxiv	Audio Jailbreak: An Open Comprehensive Benchmark for Jailbreaking Large Audio-Language Models	Audio-Language Models&Jailbreak Attacks&Adversarial Benchmark
25.05	Fujitsu Research of Europe	arxiv	MAPS: A Multilingual Benchmark for Global Agent Performance and Security	Agentic AI&Multilingual Evaluation&LLM Security
25.05	Nanyang Technological University	arxiv	AudioTrust: Benchmarking the Multifaceted Trustworthiness of Audio Large Language Models	Audio LLMs&Trustworthiness Evaluation&Multimodal Safety
25.05	University of Science and Technology of China	arxiv	From Evaluation to Defense: Advancing Safety in Video Large Language Models	Video LLMs&Safety Benchmark&Multimodal Defense
25.05	Hong Kong University of Science and Technology (Guangzhou)	arxiv	JALMBench: Benchmarking Jailbreak Vulnerabilities in Audio Language Models	Jailbreak Benchmark&Audio Language Model&Adversarial Attack
25.05	University of Washington	arxiv	SOS BENCH: Benchmarking Safety Alignment on Scientific Knowledge	Safety Alignment&Scientific Knowledge&Benchmark
25.05	John Hopkins University	arxiv	Jailbreak Distillation: Renewable Safety Benchmarking	Jailbreak Distillation&Safety Benchmark&Prompt Selection
25.05	Alibaba Group	arxiv	USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Models	Multimodal LLMs&Safety Evaluation&Benchmark
25.05	Wuhan University	arxiv	TRIDENT: Enhancing Large Language Model Safety with Tri-Dimensional Diversified Red-Teaming Data Synthesis	LLM Safety&Red-Teaming&Data Diversity
25.06	New York University Abu Dhabi	arxiv	AgentAuditor: Human-Level Safety and Security Evaluation for LLM Agents	LLM Agent&Safety Evaluation&Security Benchmark
25.06	Sabaragamuwa University of Sri Lanka	arxiv	XGUARD: A Graded Benchmark for Evaluating Safety Failures of Large Language Models on Extremist Content	LLM Safety&Extremist Content&Graded Evaluation
25.06	University of Zurich	arxiv	Simple Prompt Injection Attacks Can Leak Personal Data Observed by LLM Agents During Task Execution	Prompt Injection&Agent Security&Data Exfiltration
25.06	Sun Yat-sen University	arxiv	MTCMB: A Multi-Task Benchmark Framework for Evaluating LLMs on Knowledge, Reasoning, and Safety in Traditional Chinese Medicine	Traditional Chinese Medicine&LLM Benchmark&Safety Evaluation
25.06	NII-LLMC	arxiv	AnswerCarefully: A Dataset for Improving the Safety of Japanese LLM Output	LLM Safety&Japanese Dataset&Safety Fine-Tuning
25.06	University of Oxford	arxiv	AgentMisalignment: Measuring the Propensity for Misaligned Behaviour in LLM-Based Agents	LLM Agent&Misalignment Propensity&Safety Benchmark
25.06	ETRI	arxiv	HoliSafe: Holistic Safety Benchmarking and Modeling with Safety Meta Token for Vision-Language Model	VLM Safety&Safety Benchmark&Safety Meta Token
25.06	ByteDance	arxiv	SafeGenBench: A Benchmark Framework for Security Vulnerability Detection in LLM-Generated Code	Code Generation&Security Vulnerability&LLM Benchmark&Automatic Evaluation
25.06	ShanghaiTech University	arxiv	Beyond Jailbreaks: Revealing Stealthier and Broader LLM Security Risks Stemming from Alignment Failures	LLM Safety&Implicit Harm&JailFlipBench&Adversarial Attacks
25.06	Manipal University, GGSI University, IIITDM Kancheepuram, Meta AI, Amazon, BITS Goa	arxiv	AdversariaL attacK sAfety aLIgnment(ALKALI): Safeguarding LLMs through GRACE: Geometric Representation-Aware Contrastive Enhancement—Introducing Adversarial Vulnerability Quality Index (AVQI)	LLM Safety&Adversarial Benchmark&Latent Camouflage&Contrastive Alignment&AVQI
25.06	Microsoft, ISTA, Trend Micro, RainaResearch, University of Coimbra, Vietnamese German University, SK Shieldus, HiddenLayer	arxiv	LLMail-Inject: A Dataset from a Realistic Adaptive Prompt Injection Challenge	Prompt Injection&LLM Security&Benchmark Dataset&Email Assistant&Defense Evaluation
25.06	Institute of Computing Technology, Chinese Academy of Sciences; University of Chinese Academy of Sciences; The Chinese University of Hong Kong, Shenzhen	arxiv	From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring	LLM Safety&Content Moderation&Streaming Detection&Early Stopping&Token-Level Annotation
25.06	Zhejiang University, Tencent, National University of Singapore	arxiv	ChineseHarm-Bench: A Chinese Harmful Content Detection Benchmark	Harmful Content Detection&Chinese Benchmark&LLM Safety&Knowledge Augmentation&Content Moderation
25.06	Institute of Artificial Intelligence (TeleAI), China Telecom; Beihang University; Beijing Jiaotong University; Northwestern Polytechnical University	ACL 2025	LLMs Caught in the Crossfire: Malware Requests and Jailbreak Challenges	Jailbreak Attack&Malware Generation&LLM Security&Benchmark&Code Generation
25.06	Hong Kong University of Science and Technology, Peking University	Findings of ACL 2025	SafeLawBench: Towards Safe Alignment of Large Language Models	LLM Safety&Legal Benchmark&Risk Hierarchy&Refusal Behavior&Safety Evaluation
25.06	The University of Hong Kong	arxiv	RealFactBench: A Benchmark for Evaluating Large Language Models in Real-World Fact-Checking	Fact-checking&Large Language Models&Benchmark&Multimodal Reasoning&Uncertainty Evaluation
25.06	dreadnode	arxiv	AIRTBench: Measuring Autonomous AI Red Teaming Capabilities in Language Models	AI Red Teaming&Security Benchmark&Capture-the-Flag&Autonomous Agents&Prompt Injection
25.06	Beihang University	arxiv	AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions	Embodied Agents&Safety Benchmark&Vision-Language Model&Hazardous Instructions&Jailbreak Attack
25.06	Zhejiang University	arxiv	RAS-Eval: A Comprehensive Benchmark for Security Evaluation of LLM Agents in Real-World Environments	LLM Agent&Security Evaluation&Benchmark&CWE&Tool Execution
25.06	Anthropic	arxiv	SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents	Sabotage&Monitoring&LLM Agent
25.06	NASK – National Research Institute	arxiv	PL-Guard: Benchmarking Language Model Safety for Polish	Safety Benchmark&Polish LLM&Adversarial Robustness
25.06	University of Illinois Urbana-Champaign	arxiv	GUARDSET-X: Multi-Domain, Policy-Grounded, AI Security Guardrail Benchmark	Guardrail Dataset&Safety Evaluation&Policy-Grounded
25.06	Emory University	arxiv	Model Editing as a Double-Edged Sword: Steering Agent Ethical Behavior Toward Beneficence or Harm	Model Editing&Ethical Alignment&Behavior Control
25.06	University of Illinois Urbana-Champaign	arxiv	Beyond Reactive Safety: Risk-Aware LLM Alignment via Long-Horizon Simulation	LLM Alignment&Long-Horizon Simulation&Risk-Aware
25.06	University of Illinois Urbana-Champaign	arxiv	HalluSegBench: Counterfactual Visual Reasoning for Segmentation Hallucination Evaluation	Hallucination&Segmentation&Counterfactual Reasoning
25.07	Inner Mongolia University	arxiv	McBE: A Multi-task Chinese Bias Evaluation Benchmark for Large Language Models	Bias Evaluation&Chinese LLM&Benchmark
25.07	University of Edinburgh	arxiv	Controlling What You Share: Assessing Language Model Adherence to Privacy Preferences	Privacy Profile&LLM&PII
25.07	University of California, Los Angeles	arxiv	ModelCitizens: Representing Community Voices in Online Safety	Toxicity Detection&Community Annotation&Context
25.07	Carnegie Mellon University	arxiv	OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety	Agent Safety&Evaluation&Risk
25.05	ETH Zurich	ACL 2025 Findings	ACE: Benchmarking Agent-based Code Execution Safety in LLMs	Agent Safety&Code Execution&Benchmark
25.05	National University of Singapore	ACL 2025	DORIS: A Fine-grained Benchmark for Detecting Hallucinations in Long-form Text Generation	Hallucination Detection&Long-form Generation&Benchmark
25.05	Tsinghua University	CoLM 2025	MEmoBench: Evaluating Memorization-based Security Risks in LLMs	Memorization&Security Risk&Benchmark
25.07	CISPA Helmholtz Center for Information Security	USENIX Security 2025	Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities	VLM&Safety&Alignment
25.07	Maastricht University	arxiv	Dutch CrowS-Pairs: Adapting a Challenge Dataset for Measuring Social Biases in Language Models for Dutch	Social Bias&Dataset&Dutch
25.07	Ben-Gurion University of the Negev	arxiv	Tab-MIA: A Benchmark Dataset for Membership Inference Attacks on Tabular Data in LLMs	Membership Inference&Tabular Data&Privacy
25.07	RWTH Aachen University	GeBNLP @ ACL 2025	Exploring Gender Bias in Large Language Models: An In-depth Dive into the German Language	Gender Bias&German&LLM
25.07	Chinese Academy of Sciences, University of Chinese Academy of Sciences	arxiv	Libra: Large Chinese-based Safeguard for AI Content	Safeguard&Chinese&Safety
25.07	Peking University, Peking Union Medical College	arxiv	A Novel Evaluation Benchmark for Medical LLMs: Illuminating Safety and Effectiveness in Clinical Domains	MedicalLLM&Safety&Effectiveness
25.07	Glass Box Medicine	arxiv	Large language models provide unsafe answers to patient-posed medical questions	PatientSafety&MedicalAdvice&Chatbots
25.07	University of Cambridge	arxiv	TRIDENT: Benchmarking LLM Safety in Finance, Medicine, and Law	Safety&Benchmark&Compliance
25.07	Mohamed bin Zayed University of Artificial Intelligence, University of Delhi	arxiv	UnsafeChain: Enhancing Reasoning Model Safety via Hard Cases	SafetyAlignment&HardPrompts&Reasoning
25.08	AIM Intelligence, Seoul National University	arxiv	Eliciting and Analyzing Emergent Misalignment in State-of-the-Art Large Language Models	LLM Alignment&Misalignment&Red-Teaming&Narrative Manipulation
25.08	Shanghai Jiao Tong University	arxiv	Beyond Uniform Criteria: Scenario-Adaptive Multi-Dimensional Jailbreak Evaluation	Jailbreak Evaluation&Scenario-Adaptivity&LLM Safety
25.08	Shanghai Jiao Tong University	arxiv	MCSBench: Benchmarking Multimodal Commonsense Reasoning of Large Vision-Language Models	Multimodal Commonsense Reasoning&Large Vision-Language Models&Benchmark
25.08	Fordham University	arxiv	Towards Safer AI Moderation: Evaluating LLM Moderators Through a Unified Benchmark Dataset and Advocating a Human-First Approach	AI Moderation&Bias Detection&Human-in-the-Loop
25.08	Apple	COLM 2025	Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolution	Intersectional Bias&Coreference Resolution&Fairness Evaluation
25.08	Tsinghua University	arxiv	Omni-SafetyBench: A Benchmark for Safety Evaluation of Audio-Visual Large Language Models	Audio-Visual LLMs&Safety Evaluation&Cross-Modal Consistency
25.08	The Hong Kong University of Science and Technology (Guangzhou)	arxiv	Exploring Safety Alignment Evaluation of LLMs in Chinese Mental Health Dialogues via LLM-as-Judge	Mental Health Dialogues&Safety Alignment&LLM-as-Judge
25.08	Peking University	ACM MM 2025	Safe Semantics, Unsafe Interpretations: Tackling Implicit Reasoning Safety in Large Vision-Language Models	Vision-Language Models&Implicit Reasoning Safety&In-Context Learning
25.08	Guizhou University	arxiv	XFacta: Contemporary, Real-World Dataset and Evaluation for Multimodal Misinformation Detection with Multimodal LLMs	Multimodal Misinformation Detection&MLLM Evaluation&Contemporary Dataset
25.08	Shanghai Artificial Intelligence Laboratory	arxiv	LinguaSafe: A Comprehensive Multilingual Safety Benchmark for Large Language Models	Multilingual Safety&Benchmark&LLM Evaluation
25.08	University of Science and Technology of China, Beihang University	arxiv	MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers	Tool Poisoning&LLM Agent&Benchmark
25.08	Tsinghua University, Beihang University, RealAI	arxiv	Unveiling Trust in Multimodal Large Language Models: Evaluation, Analysis, and Mitigation	Multimodal LLMs&Trustworthiness&Benchmark
25.08	Shanghai AI Lab, East China Normal University	arxiv	SafetyFlow: An Agent-Flow System for Automated LLM Safety Benchmarking	LLM Safety&Automated Benchmarking&Agent System
25.08	Oregon State University	arxiv	Mind the Gap: Time-of-Check to Time-of-Use Vulnerabilities in LLM-Enabled Agents	LLM Agents&TOCTOU Vulnerabilities&Security Mitigation
25.08	University of Wisconsin-Madison	arxiv	PRISM: Robust VLM Alignment with Principled Reasoning for Integrated Safety in Multimodality	Vision-Language Models&Multimodal Safety&Chain-of-Thought Alignment
25.08	Shandong University	arxiv	Mind the Third Eye! Benchmarking Privacy Awareness in MLLM-powered Smartphone Agents	Smartphone Agents&Privacy Awareness&SAPA-Bench
25.08	CISPA Helmholtz Center for Information Security	arxiv	JADES: A Universal Framework for Jailbreak Assessment via Decompositional Scoring	Jailbreak Assessment&Decompositional Scoring&Fact-Checking
25.09	Shanghai Artificial Intelligence Laboratory	EMNLP 2025 Findings	Self-adaptive Dataset Construction for Real-World Multimodal Safety Scenarios	Multimodal Safety&Dataset Construction&Risk Evaluation
25.09	Sichuan University	EMNLP 2025 Findings	CANDY: Benchmarking LLMs' Limitations and Assistive Potential in Chinese Misinformation Fact-Checking	Misinformation Fact-Checking&Chinese LLMs&Benchmark Dataset
25.09	Beihang University	arxiv	VaccineRAG: Boosting Multimodal Large Language Models' Immunity to Harmful RAG Samples	Retrieval-Augmented Generation&Chain-of-Thought&Multimodal LLMs
25.09	Leibniz Institute for Resilience Research	arxiv	Scaling behavior of large language models in emotional safety classification across sizes and tasks	Emotional Safety&Model Scaling&Mental Health Applications
25.09	King Fahd University of Petroleum and Minerals	arxiv	AraHalluEval: A Fine-grained Hallucination Evaluation Framework for Arabic LLMs	Arabic LLMs&Hallucination Evaluation&Multilingual Models
25.09	Stevens Institute of Technology	arxiv	Behind the Mask: Benchmarking Camouflaged Jailbreaks in Large Language Models	Camouflaged Jailbreaks&Benchmark Dataset&LLM Security
25.09	Sharif University of Technology	arxiv	EPT Benchmark: Evaluation of Persian Trustworthiness in Large Language Models	Trustworthiness&Persian LLMs&Cultural Alignment
25.09	Beijing Institute of Technology	arxiv	SafeToolBench: Pioneering a Prospective Benchmark to Evaluating Tool Utilization Safety in LLMs	Tool Safety&LLM Agents&Prospective Evaluation
25.09	Appen	arxiv	Red Teaming Multimodal Language Models: Evaluating Harm Across Prompt Modalities and Models	Multimodal LLMs&Red Teaming&Adversarial Prompts&Safety Evaluation
25.09	Peking University, TeleAI (China Telecom), UCAS	arxiv	When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models	Multimodal Safety&Reasoning Path Optimization&Implicit Reasoning Risk
25.09	University of Washington, IAEA	arxiv	Automated Triaging and Transfer Learning of Incident Learning Safety Reports Using Large Language Representational Models	Incident Learning Systems&Transfer Learning&BlueBERT
25.09	EPFL & Tampere University & University of Tokyo & National Institute of Informatics	EMNLP 2025	TactfulToM: Do LLMs Have the Theory of Mind Ability to Understand White Lies?	Theory of Mind&White Lies&Social Reasoning
25.09	Singapore Management University, National University of Singapore, Monash University, Aalto University, York University, Zhejiang University	arxiv	SecureAgentBench: Benchmarking Secure Code Generation under Realistic Vulnerability Scenarios	Secure Code Generation&LLM Agents&Software Security Benchmark
25.09	Zhejiang University	arxiv	HarmMetric Eval: Benchmarking Metrics and Judges for LLM Harmfulness Assessment	LLM Harmfulness Evaluation&Benchmark&Safety Alignment
25.09	Shanghai Jiao Tong University, China Telecom Institute of Artificial Intelligence, USTC, GigaAI, Xinjiang University	arxiv	RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration	LLM Safety Evaluation&Multi-Agent Collaboration&Risk-Aware Framework
25.09	UC Berkeley, AWS AI Labs	arxiv	STAC: When Innocent Tools Form Dangerous Chains to Jailbreak LLM Agents	LLM Agents&Sequential Tool Attacks&Multi-turn Security Vulnerability
25.09	Chinese Academy of Sciences	arxiv	SafeMind: Benchmarking and Mitigating Safety Risks in Embodied LLM Agents	Embodied LLM Agents&Safety Benchmark&Risk Mitigation
25.09	Shanghai AI Laboratory, Fudan University, The University of Hong Kong, East China Normal University	arxiv	SafeEvalAgent: Toward Agentic and Self-Evolving Safety Evaluation of LLMs	Safety Evaluation&Multi-Agent Framework&Dynamic Benchmarking
25.10	UC San Diego, Columbia University, Zhejiang University, MIT-IBM Watson Lab	arxiv	BiasFreeBench: A Benchmark for Mitigating Bias in Large Language Model Responses	Bias Mitigation&Benchmarking&Fairness Evaluation
25.10	Duke University	arxiv	WAInjectBench: Benchmarking Prompt Injection Detections for Web Agents	Prompt Injection Detection&Web Agents&Multimodal Security&Benchmarking
25.10	AITRICS, KAIST, Severance Hospital (Yonsei University)	EMNLP 2025	Taxonomy of Comprehensive Safety for Clinical Agents	Clinical Safety&Taxonomy&Intent Classification&LLM Safety
25.10	Rutgers University	arxiv	Read the Scene, Not the Script: Outcome-Aware Safety for LLMs	LLM Safety Alignment&Consequence-Blindness&Outcome-Aware Reasoning
25.10	University of California, San Diego	arxiv	When Benchmarks Age: Temporal Misalignment through Large Language Model Factuality Evaluation	Benchmark Aging&Factuality Evaluation&Temporal Misalignment
25.10	Beihang University & Chinese Academy of Sciences & China University of Petroleum (East China) & Zhejiang University of Technology & 360 AI Security Lab & University of Sydney	arxiv	SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents	Web Agent Security&LVLM Benchmark&Adversarial Evaluation&Prompt Injection
25.10	Carnegie Mellon University & Amazon AGI	arxiv	RefusalBench: Generative Evaluation of Selective Refusal in Grounded Language Models	Selective Refusal&RAG Evaluation&Generative Benchmark
25.10	Beijing University of Posts and Telecommunications, University of California Santa Barbara	arxiv	MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents	Model Context Protocol&Agent Security&Benchmark
25.10	National Taiwan University, NVIDIA	arxiv	Investigating Safety Vulnerabilities of Large Audio-Language Models Under Speaker Emotional Variations	Audio-Language Models&Safety Alignment&Emotional Variation
25.10	Apple	arxiv	VLSU: Mapping the Limits of Joint Multimodal Understanding for AI Safety	Multimodal Safety&Vision-Language Models&Benchmark
25.10	Amirkabir University of Technology, KCL, Politecnico di Milano, KTH	arxiv	PBBQ: A Persian Bias Benchmark Dataset Curated with Human-AI Collaboration for Large Language Models	Bias Benchmark&Persian LLM&Human-AI Collaboration
25.10	NYU Shanghai, NUS, Yale University, NYU Center for Data Science	EMNLP 2025	FinTrust: A Comprehensive Benchmark of Trustworthiness Evaluation in Finance Domain	Trustworthiness Benchmark&Financial LLM&Ethical Alignment
25.10	Beihang University, Tsinghua University, Shanghai Qi Zhi Institute	NeurIPS 2025	DeceptionBench: A Comprehensive Benchmark for AI Deception Behaviors in Real-world Scenarios	Deception Behavior&Benchmark&LLM Safety
25.10	Manchester Metropolitan University	arxiv	Safeguarding Efficacy in Large Language Models: Evaluating Resistance to Human-Written and Algorithmic Adversarial Prompts	Large Language Models&Adversarial Attacks&AI Safety
25.10	Lakera AI, ETH Zürich, UK AI Security Institute, University of Oxford	arxiv	Breaking Agent Backbones: Evaluating the Security of Backbone LLMs in AI Agents	AI Agent Security&Backbone LLM Benchmarking&Threat Modeling
25.10	William & Mary	arxiv	Measuring the Security of Mobile LLM Agents under Adversarial Prompts from Untrusted Third-Party Channels	Mobile LLM Agent&Adversarial Prompt&Security Evaluation
25.10	Hong Kong Baptist University	arxiv	MemeArena: Automating Context-Aware Unbiased Evaluation of Harmfulness Understanding for Multimodal Large Language Models	Multimodal Large Language Model&Harmfulness Evaluation&Agent-based Framework
25.11	Penn State University	arxiv	Do Methods to Jailbreak and Defend LLMs Generalize Across Languages?	Multilingual Jailbreak&Cross-lingual Safety&Defense Robustness
25.11	Tsinghua University	arxiv	LiveSecBench: A Dynamic and Culturally-Relevant AI Safety Benchmark for LLMs in Chinese Context	AI Safety Benchmark&Chinese LLM Evaluation&Cultural Alignment
25.11	Nanjing University	arxiv	Understanding New-Knowledge-Induced Factual Hallucinations in LLMs: Analysis, Solution, and Interpretation	Factual Hallucination&New Knowledge&Attention Mechanism
25.11	German Research Center for Artificial Intelligence	arxiv	CONVERSE: Benchmarking Contextual Safety in Agent-to-Agent Conversations	Agent-to-Agent Safety&Privacy Leakage&Security Attacks
25.11	Southern University of Science and Technology	arxiv	MCP-RiskCue: Can LLM Infer Risk Information From MCP Server System Logs?	MCP Security&System Log Analysis&Risk Detection Benchmark
25.11	Beijing Institute of AI Safety and Governance	arxiv	Efficient LLM Safety Evaluation through Multi-Agent Debate	LLM Safety Evaluation&Multi-Agent Debate&Jailbreak Benchmark
25.11	Xi’an Jiaotong University	arxiv	How Brittle is Agent Safety? Rethinking Agent Risk under Intent Concealment and Task Complexity	Agent Safety&Intent Concealment&Task Complexity
25.11	Fudan University / City University of Hong Kong / Singapore Management University	arxiv	AttackVLA: Benchmarking Adversarial and Backdoor Attacks on Vision–Language–Action Models	Vision–Language–Action Models&Adversarial Attacks&Backdoor Attacks
25.11	Nanjing University	arxiv	SAFERBENCH: A COMPREHENSIVE BENCHMARK FOR SAFETY ASSESSMENT IN LARGE REASONING MODELS	Safety Benchmark&Large Reasoning Models&Chain-of-Thought Safety
25.11	École Polytechnique, Sorbonne Paris Nord University, Crédit Agricole SA	arxiv	MUCH: A Multilingual Claim Hallucination Benchmark	Uncertainty Quantification&Hallucination Detection&Multilingual Benchmark
25.11	Intuit AI Security Research	arxiv	ASTRA: Agentic Steerability and Risk Assessment Framework	Agentic AI Security&Steerability Evaluation&Risk Assessment Framework
25.11	Scale AI, University of Maryland, UNC, Google DeepMind, Netflix, UT Austin	arxiv	PropensityBench: Evaluating Latent Safety Risks in Large Language Models via an Agentic Approach	LLM Safety&Risk Propensity&Agentic Evaluation
25.11	University at Buffalo, SUNY	arxiv	DUALGAUGE: Automated Joint Security–Functionality Benchmarking for Secure Code Generation	Secure Code Generation&Benchmarking&LLM Evaluation
25.12	Bangladesh University of Engineering and Technology	arxiv	When Safety Blocks Sense: Measuring Semantic Confusion in LLM Refusals	Safety Alignment&Semantic Confusion&False Refusal Analysis
25.12	BITS Pilani, Hyderabad Campus	arxiv	BHRAM-IL: A Benchmark for Hallucination Recognition and Assessment in Multiple Indian Languages	Multilingual Hallucination Detection&Indic Languages&Benchmark Dataset
25.12	University of Leeds, Ben-Gurion University of the Negev	arxiv	DialogGuard: Multi-Agent Psychosocial Safety Evaluation of Sensitive LLM Responses	Psychosocial Safety&Multi-Agent Evaluation&LLM-as-a-Judge
25.12	Carnegie Mellon University, Columbia University, Johns Hopkins University, HydroX AI	arxiv	Is Vibe Coding Safe? Benchmarking Vulnerability of Agent-Generated Code in Real-World Tasks	Vibe Coding&AI Agents&Secure Code Generation&Benchmarking
25.12	Harbin Institute of Technology (Shenzhen)	NeurIPS 2025 / arxiv	Context-Aware Hierarchical Learning: A Two-Step Paradigm Towards Safer LLMs	LLM Security&Prompt Injection Defense&Tool-Augmented Models&Context-Aware Hierarchical Learning (CAHL)
25.12	Institute of Artificial Intelligence (TeleAI) of China Telecom	arxiv	TeleAI-Safety: A comprehensive LLM jailbreaking benchmark towards attacks, defenses, and evaluations	Large Language Model&Jailbreak&Safety Benchmark
25.12	Vidyasirimedhi Institute of Science and Technology	arxiv	SEA-SafeguardBench: Evaluating AI Safety in SEA Languages and Cultures	AI Safety Benchmark&Southeast Asian Languages&Cultural Safety
25.12	Nanyang Technological University	arxiv	OmniSafeBench-MM: A Unified Benchmark and Toolbox for Multimodal Jailbreak Attack–Defense Evaluation	Multimodal LLM Safety&Jailbreak Benchmark&Attack–Defense Evaluation
25.12	East China Normal University	arxiv	MCP-SafetyBench: A Benchmark for Safety Evaluation of Large Language Models with Real-World MCP Servers	MCPSecurity&SafetyBenchmark&LLMAgents
25.12	BRAC University, Bangladesh	arxiv	Beyond the Benchmark: Innovative Defenses Against Prompt Injection Attacks	PromptInjection&DefenseGeneration&LLMSecurity
25.12	i.AI	arxiv	A Real-World Evaluation of LLM Medication Safety Reviews in NHS Primary Care	Medical Safety&Real-World Evaluation&Failure Analysis
25.12	Western Sydney University	arxiv	Prompt-Induced Over-Generation as Denial-of-Service: A Black-Box Attack-Side Benchmark	Over-Generation&Denial-of-Service&Prompt Attacks
25.12	International Institute of Information Technology Hyderabad	arxiv	PrivacyBench: A Conversational Benchmark for Evaluating Privacy in Personalized AI	Privacy Benchmark&Personalized AI&RAG Systems
26.01	Xidian University	arxiv	Overlooked Safety Vulnerability in LLMs: Malicious Intelligent Optimization Algorithm Request and its Jailbreak	Algorithm Design&Jailbreak&Optimization Safety
26.01	Nanyang Technological University	arxiv	CSSBench: Evaluating the Safety of Lightweight LLMs against Chinese-Specific Adversarial Patterns	Chinese Safety&Adversarial Evaluation&Over-Refusal
26.01	Kyoto University	arxiv	JMedEthicBench: A Multi-Turn Conversational Benchmark for Evaluating Medical Safety in Japanese Large Language Models	Medical Safety&Multi-Turn Evaluation&Japanese Benchmark
26.01	Indian Institute of Technology Indore	AAAI 2026	X-MuTeST: A Multilingual Benchmark for Explainable Hate Speech Detection and A Novel LLM-consulted Explanation Framework	Explainable AI&Hate Speech Detection&Multilingual NLP
26.01	Fordham University	arxiv	GuardEval: A Multi-Perspective Benchmark for Evaluating Safety, Fairness, and Robustness in LLM Moderators	LLM Moderation&Safety Benchmark&Fairness Robustness
26.01	United Arab Emirates University	arxiv	α³-Bench: A Unified Benchmark of Safety, Robustness, and Efficiency for LLM-Based UAV Agents over 6G Networks	UAV Agents&6G Networks&Safety Benchmark
26.01	Monash University	arxiv	PsychEthicsBench: Evaluating Large Language Models Against Australian Mental Health Ethics	Mental Health Ethics&Ethical Benchmark&Jurisdiction Awareness
26.01	VNU University of Science	arxiv	RedBench: A Universal Dataset for Comprehensive Red Teaming of Large Language Models	Red Teaming&Safety Benchmark&Risk Taxonomy
26.01	Xidian University	arxiv	HearSay Benchmark: Do Audio LLMs Leak What They Hear?	Audio Privacy&Voiceprint Leakage&Safety Benchmark
26.01	Beijing Jiaotong University	arxiv	When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Life	Multimodal Safety&Daily Life Risks&Safety Benchmark
26.01	National University of Defense Technology	arxiv	The Forgotten Shield: Safety Grafting in Parameter-Space for Medical MLLMs	Medical Safety&Parameter Alignment&Safety Grafting
26.01	Federal University of Goiás	arxiv	MiJaBench: Revealing Minority Biases in Large Language Models via Hate Speech Jailbreaking	Demographic Bias&Hate Speech Safety&Jailbreak Benchmark
26.01	Dongguk University	arxiv	Safety Not Found (404): Hidden Risks of LLM-Based Robotics Decision Making	RoboticsSafety&SpatialReasoning&LLMReliability
26.01	University of California, Riverside	arxiv	PII-VisBench: Evaluating Personally Identifiable Information Safety in Vision Language Models Along a Continuum of Visibility	PIISafety&VLMPrivacy&Benchmark
26.01	University of California, Santa Cruz	arxiv	SafePro: Evaluating the Safety of Professional-Level AI Agents	AgentSafety&ProfessionalTasks&Benchmark
26.01	University of Notre Dame	arxiv	MTMCS-Bench: Evaluating Contextual Safety of Multimodal Large Language Models in Multi-Turn Dialogues	MultimodalSafety&MultiTurn&Benchmark
26.01	Shanghai University of Finance and Economics	arxiv	FinVault: Benchmarking Financial Agent Safety in Execution-Grounded Environments	FinancialAgents&SecurityBenchmark&AgentSafety
26.01	Fudan University	arxiv	WebTrap Park: An Automated Platform for Systematic Security Evaluation of Web Agents	WebAgents&SecurityEvaluation&Benchmark
26.01	The University of Texas at Austin	arxiv	Faithfulness vs. Safety: Evaluating LLM Behavior Under Counterfactual Medical Evidence	MedicalSafety&CounterfactualEvidence&LLMBehavior
26.01	The Alan Turing Institute	arxiv	DriveSafe: A Hierarchical Risk Taxonomy for Safety-Critical LLM-Based Driving Assistants	DrivingSafety&RiskTaxonomy&LLMEvaluation
26.01	Brown University	arxiv	UbuntuGuard: A Culturally-Grounded Policy Benchmark for Equitable AI Safety in African Languages	MultilingualSafety&CulturalGrounding&GuardianModels
26.01	INSAIT	arxiv	ChartAttack: Testing the Vulnerability of LLMs to Malicious Prompting in Chart Generation	ChartSecurity&MultimodalJailbreak&VisualizationAttacks
26.01	Shanghai Qi Zhi Institute	arxiv	VirtualCrime: Evaluating Criminal Potential of Large Language Models via Sandbox Simulation	AgenticCrime&SandboxSimulation&LLMSafety
26.01	Tsinghua University	arxiv	The Side Effects of Being Smart: Safety Risks in MLLMs’ Multi-Image Reasoning	MultiImageReasoning&MLLMSafety&SafetyBenchmarks
26.01	Eindhoven University of Technology	arxiv	AdversaRiskQA: An Adversarial Factuality Benchmark for High-Risk Domains	AdversarialFactuality&HighRiskDomains&LLMBenchmark
26.01	Nanyang Technological University	CCS 2026	The Shadow Self: Intrinsic Value Misalignment in Large Language Model Agents	LLM Agents&Value Misalignment&Safety Benchmark
26.01	Beijing Normal University	arxiv	GUIGuard: Toward a General Framework for Privacy-Preserving GUI Agents	GUI Agents&Privacy Preservation&Agent Safety
26.01	Shanghai AI Lab	arxiv	Automated Safety Benchmarking: A Multi-agent Pipeline for LVLMs	LVLM Safety&Automated Benchmarking&Multi-agent Systems
26.01	The Chinese University of Hong Kong	arxiv	VoxPrivacy: A Benchmark for Evaluating Interactional Privacy of Speech Language Models	Interactional Privacy&Speech LLMs&Safety Benchmark
26.01	Unknown	arxiv	SHARP: Social Harm Analysis via Risk Profiles for Measuring Inequities in Large Language Models	Social Harm&Tail Risk&Fairness Evaluation
26.01	Tehran Institute for Advanced Studies	arxiv	FFE-HALLU: Hallucinations in Fixed Figurative Expressions: Benchmark of Idioms and Proverbs in the Persian Language	Figurative Hallucination&Idioms Benchmark&Cultural Grounding
26.01	University of Illinois	WWW 2026	Mind the Ambiguity: Aleatoric Uncertainty Quantification in LLMs for Safe Medical Question Answering	Aleatoric Uncertainty&Medical QA&Ambiguity Detection
26.02	National University of Singapore	arxiv	Risky-Bench: Probing Agentic Safety Risks under Real-World Deployment	Agent Safety&Deployment&Benchmark
26.02	ShanghaiTech University, Shanghai, China	arxiv	LPS-BENCH: Benchmarking Safety Awareness of Computer-Use Agents in Long-Horizon Planning under Benign and Adversarial Scenarios	Computer-Use Agents&Planning Safety&Benchmark
26.02	Zhejiang University	arxiv	CSR-Bench: A Benchmark for Evaluating the Cross-modal Safety and Reliability of MLLMs	Cross-modal Safety&MLLM Reliability&Benchmark
26.02	Qatar Computing Research Institute, HBKU, Doha, Qatar	arxiv	Once Correct, Still Wrong: Counterfactual Hallucination in Multilingual Vision-Language Models	Counterfactual Hallucination&Multilingual VLMs&Benchmark
26.02	Zhejiang University	ICLR 2026	WHEN AGENTS “MISREMEMBER” COLLECTIVELY: EXPLORING THE MANDELA EFFECT IN LLM-BASED MULTI-AGENT SYSTEMS	Mandela Effect&Multi-Agent Systems&False Memory
26.02	Seoul National University College of Medicine	arxiv	MPIB: A Benchmark for Medical Prompt Injection Attacks and Clinical Safety in LLMs	Prompt Injection&Clinical Safety&RAG
26.02	School of Software, Shandong University, Jinan, China	arxiv	TrajAD: Trajectory Anomaly Detection for Trustworthy LLM Agents	LLM Agents&Anomaly Detection&Process Supervision
26.02	Critical ML Lab	arxiv	TAMPERBENCH: SYSTEMATICALLY STRESS-TESTING LLM SAFETY UNDER FINE-TUNING AND TAMPERING	Tamper Resistance&Safety Evaluation&Fine-Tuning Attacks
26.02	Texas A&M University, College Station, TX, USA	arxiv	Halluverse-M3: A multitask multilingual benchmark for hallucination in LLMs	Hallucination Benchmark&Multilingual Evaluation&Detection
26.02	Computer Science Department, Southern Illinois University, Carbondale, IL, United States	arxiv	AGENT-FENCE: MAPPING SECURITY VULNERABILITIES ACROSS DEEP RESEARCH AGENTS	Agent Security&Trust Boundaries&Architectural Evaluation
26.02	Xidian University	arxiv	FROM ASSISTANT TO DOUBLE AGENT: FORMALIZING AND BENCHMARKING ATTACKS ON OPENCLAW FOR PERSONALIZED LOCAL AI AGENT.	Personalized Agents&Security Benchmarking&OpenClaw
26.02	University of Oregon	arxiv	Benchmarking Knowledge-Extraction Attack and Defense on Retrieval-Augmented Generation	Retrieval-Augmented Generation&Knowledge Extraction&Benchmarking
26.02	Sentient Labs	arxiv	CryptoAnalystBench: Failures in Multi-Tool Long-Form LLM Analysis	Analyst Agents&Tool Use&Long-Form Evaluation
26.02	Macquarie University, Sydney, Australia	arxiv	Are Aligned Large Language Models Still Misaligned?	Misalignment Benchmark&Safety&Culture
26.02	ETH Zürich	arxiv	GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory	AI Safety Benchmark&Game Theory&Multi-Agent Risks
26.02	Unknown	arxiv	ProMoral-Bench: Evaluating Prompting Strategies for Moral Reasoning and Safety in LLMs	Moral Reasoning&Prompt Engineering&Safety Benchmark
26.02	Northeastern University	arxiv	Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents	Tool-Using Agents&Multi-Turn Safety&Defense
26.02	Beijing Institute of AI Safety and Governance, China.	arxiv	ForesightSafety Bench: A Frontier Risk Evaluation and Governance Framework towards Safe AI	AI Safety Benchmark&Frontier Risks&Governance
26.02	KI-Werkstatt / FB2, University of Applied Sciences Berlin	arxiv	LLMStructBench: Benchmarking Large Language Model Structured Data Extraction	Structured Extraction&Benchmark&JSON Generation
26.02	TSINGHUA UNIVERSITY, BEIJING, CHINA	arxiv	SecCodeBench-V2 Technical Report	Secure Code Generation&Benchmark&Code Security
26.02	University of Tübingen & Max-Planck Institute for Intelligent Systems	arxiv	NESSIE: THE NECESSARY SAFETY BENCHMARK - IDENTIFYING ERRORS THAT SHOULD NOT EXIST	Safety Benchmark&Instruction Following&Necessary Condition
26.02	Independent Researcher	arxiv	Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents	Tool-Call Safety&Benchmark&Agent Alignment
26.02	Oregon State University, Corvallis, OR USA	arxiv	Fail-Closed Alignment for Large Language Models	Fail-Closed Safety&Refusal Robustness&Jailbreak Defense
26.02	Oracle America Inc.	EACL 2026	IndicJR: A Judge-Free Benchmark of Jailbreak Robustness in South Asian Languages	Multilingual Safety&Jailbreak Benchmark&South Asian Languages
26.02	Kakaobank, South Korea	arxiv	FENCE: A Financial and Multimodal Jailbreak Detection Dataset	jailbreak detection&finance&multimodal
26.02	School of Computer Science, Carnegie Mellon University	arxiv	How Well Can LLM Agents Simulate End-User Security and Privacy Attitudes and Behaviors?	user simulation&security privacy&benchmark
26.02	Southeast University	arxiv	CIBER: A Comprehensive Benchmark for Security Evaluation of Code Interpreter Agents	code interpreter&benchmark&agent security
26.02	Max Planck Institute for Intelligent Systems, ELLIS Institute Tubingen, Tubingen AI Center	arxiv	SKILL-INJECT: Measuring Agent Vulnerability to Skill File Attacks	skill attack&prompt injection&agent benchmark
26.02	Harbin Institute of Technology, Shenzhen	arxiv	Evaluating Proactive Risk Awareness of Large Language Models	proactive safety&risk awareness&ecology
26.03	Shanghai AI Lab	arxiv	SafeSci: Safety Evaluation of Large Language Models in Science Domains and Beyond	scientific safety&benchmark&alignment
26.03	UC Santa Cruz	arxiv	ZERODAYBENCH: EVALUATING LLM AGENTS ON UNSEEN ZERO-DAY VULNERABILITIES	agents&zero-day&benchmark
26.03	Duke University	arxiv	MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Foundation Models	multimodal safety&evaluation platform&benchmarking
26.03	Renmin University of China, China	arxiv	TRUSTMH-BENCH: A Comprehensive Benchmark for Evaluating the Trustworthiness of Mental Health LLMs	mental health&trustworthiness&benchmark
26.03	Compumacy for Artificial Intelligence Solutions, Cairo, Egypt	arxiv	SalamahBench: Toward Standardized Safety Evaluation for Arabic Language Models	Arabic LMs&safety benchmark&evaluation
26.03	CISPA Helmholtz Center for Information Security	arxiv	Benchmark of Benchmarks: Unpacking Influence and Code Repository Quality in LLM Safety Benchmarks	safety benchmarks&code quality&meta-analysis
26.03	SCB DataX	arxiv	THAISAFETYBENCH: ASSESSING LANGUAGE MODEL SAFETY IN THAI CULTURAL CONTEXTS	Thai safety&cultural context&benchmark
26.03	Mohamed bin Zayed University of Artificial Intelligence	ICLR 2026	JAILNEWSBENCH: MULTI-LINGUAL AND REGIONAL BENCHMARK FOR FAKE NEWS GENERATION UNDER JAILBREAK ATTACKS	fake news&jailbreak&multilingual benchmark
26.03	Technical University of Munich	arxiv	A Coin Flip for Safety: LLM Judges Fail to Reliably Measure Adversarial Robustness	LLM-as-a-Judge&adversarial robustness&evaluation reliability
26.03	Oracle	arxiv	LieCraft: A Multi-Agent Framework for Evaluating Deceptive Capabilities in Language Models	deception evaluation&multi-agent games&alignment
26.03	MediaTek Research	arxiv	Taiwan Safety Benchmark and Breeze Guard: Toward Trustworthy AI for Taiwanese Mandarin	Taiwanese Mandarin&safety benchmark&cultural alignment
26.03	Singapore Management University	arxiv	Backdoor4Good: Benchmarking Beneficial Uses of Backdoors in LLMs	beneficial backdoors&controllability&trustworthy AI
26.03	School of Informatics, Computing, and Cyber Systems, Northern Arizona University, Flagstaff, USA	arxiv	Where Do LLM-based Systems Break? A System-Level Security Framework for Risk Assessment and Treatment	system security&risk assessment&attack-defense trees
26.03	Wuhan University, Wuhan, Hubei, China	KDD 2026	AdaCultureSafe: Adaptive Cultural Safety Grounded by Cultural Knowledge in Large Language Models	cultural safety&cultural knowledge&adaptive alignment
26.03	School of Computing and Augmented Intelligence, Arizona State University, Tempe, USA	arxiv	Real-Time Trust Verification for Safe Agentic Actions using TrustBench	trust verification&agent safety&real-time intervention
26.03	Fudan University, Shanghai, China	arxiv	OOD-MMSafe: Advancing MLLM Safety from Harmful Intent to Hidden Consequences	multimodal safety&hidden consequences&policy optimization
26.03	Department of Statistics & Data Science, Yale University	arxiv	Benchmarking Political Persuasion Risks Across Frontier Large Language Models	political persuasion&frontier models&risk benchmarking
26.03	University of Twente, Enschede, The Netherlands	arxiv	TOSSS: a CVE-based Software Security Benchmark for Large Language Models	software security benchmarks&CVE&code selection
26.03	State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences	arxiv	INFACT: A Diagnostic Benchmark for Induced Faithfulness and Factuality Hallucinations in Video-LLMs	video-LLMs&hallucination benchmark&reliability
26.03	Fraunhofer AISEC	arxiv	You Told Me to Do It: Measuring Instructional Text-induced Private Data Leakage in LLM Agents	privacy leakage&README injection&trusted executor dilemma
26.03	CISPA Helmholtz Center for Information Security	arxiv	Understanding LLM Behavior When Encountering User-Supplied Harmful Content in Harmless Tasks	harmful content&harmless tasks&content-level ethics
26.03	Renmin University of China	arxiv	HomeSafe-Bench: Evaluating Vision-Language Models on Unsafe Action Detection for Embodied Agents in Household Scenarios	household safety&vision-language models&embodied agents
26.03	Tsinghua University, Beijing, China	arxiv	LABSHIELD: A Multimodal Benchmark for Safety-Critical Reasoning and Planning in Scientific Laboratories	laboratory safety&multimodal benchmarks&reasoning
26.03	KAIST AI	arxiv	UniSAFE: A Comprehensive Benchmark for Safety Evaluation of Unified Multimodal Models	Multimodal Safety&Benchmarking&Unified Models
26.03	Oracle America Inc.	arxiv	IndicSafe: A Benchmark for Evaluating Multilingual LLM Safety in South Asia	Multilingual Safety&Indic Languages&Safety Drift
26.03	Department of Computer Science, Stanford University, Stanford, CA, USA	arxiv	TherapyGym: Evaluating and Aligning Clinical Fidelity and Safety in Therapy Chatbots	Therapy Chatbots&Clinical Safety&CBT Fidelity
26.03	King Abdullah University of Science and Technology (KAUST), Saudi Arabia	arxiv	SAVeS: Steering Safety Judgments in Vision-Language Models via Semantic Cues	Vision-Language Models&Semantic Steering&Multimodal Safety
26.03	Robotics Institute, Carnegie Mellon University	CVPR 2026	Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory	Contextual Safety&Self-Reflective Memory&Multimodal Benchmark
26.03	Institute of Science Tokyo	arxiv	JUBAKU: An Adversarial Benchmark for Exposing Culturally Grounded Stereotypes in Japanese LLMs	Social bias&Japanese LLMs&Cultural stereotypes
26.03	Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Italy	arxiv	SecureBreak - A dataset towards safe and secure models	Safety dataset&Unsafe outputs&Security alignment
26.03	CSIRO Data61, Australia	arxiv	Does Teaming-Up LLMs Improve Secure Code Generation? A Comprehensive Evaluation with Multi-LLMSecCodeEval	Secure code generation&Multi-LLM ensembles&Static analysis
26.03	Southern University of Science and Technology	arxiv	BeSafe-Bench: Unveiling Behavioral Safety Risks of Situated Agents in Functional Environments	Agent safety&Situated agents&Behavioral benchmark
26.03	Ca' Foscari University of Venice, Venice, Italy	arxiv	Reentrancy Detection in the Age of LLMs	Smart contract security&Reentrancy detection&LLM evaluation
26.03	Institute of Cyberspace Security, School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen, China	arxiv	CDH-Bench: A Commonsense-Driven Hallucination Benchmark for Evaluating Visual Fidelity in Vision-Language Models	VLM hallucination&Commonsense conflict&Visual fidelity
26.03	Key Laboratory of High Confidence Software Technologies, Ministry of Education, Peking University, Beijing, China	arxiv	SkillTester Benchmarking Utility and Security of Agent Skills	Agent skills&Utility evaluation&Security benchmarking
26.03	City University of Hong Kong	arxiv	TSHA: A Benchmark for Visual Language Models in Trustworthy Safety Hazard Assessment Scenarios	Safety hazard assessment&Visual language models&Trustworthy benchmark
26.04	Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China	arxiv	Does Unification Come at a Cost? Uni-SafeBench: A Safety Benchmark for Unified Multimodal Large Models	Unified multimodal models&Safety benchmark&Intrinsic safety
26.04	George Mason University	arxiv	CLAWSAFETY: "Safe" LLMs, Unsafe Agents	Agent safety benchmark&Prompt injection&High-privilege agents
26.04	Carnegie Mellon University	arxiv	AGENTSOCIALBENCH: Evaluating Privacy Risks in Human-Centered Agentic Social Networks	Privacy risks&Agentic social networks&Multi-agent coordination
26.04	International Institute of Information Technology, Hyderabad	arxiv	ImplicitBBQ: Benchmarking Implicit Bias in Large Language Models through Characteristic Based Cues	Implicit bias&Bias benchmark&Characteristic cues
26.04	Shanghai AI Lab	arxiv	ATBench: A Diverse and Realistic Trajectory Benchmark for Long-Horizon Agent Safety	Agent safety benchmark&Trajectory evaluation&Long-horizon risks
26.04	BenchFlow	arxiv	ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspaces	OpenClaw&Benchmarking&Productivity Agents
26.04	Binjiang Institute of Zhejiang University, Hangzhou, China	arxiv	From Retinal Evidence to Safe Decisions: RETINA-SAFE and ECRT for Hallucination Risk Triage in Medical LLMs	Hallucination&Benchmarking&Healthcare AI
26.04	Beijing University of Posts and Telecommunications, Beijing, China	arxiv	DetailVerifyBench: A Benchmark for Dense Hallucination Localization in Long Image Captions	Hallucination&Benchmarking&Multimodal Safety
26.04	Graduate School of Data Science, Seoul National University	arxiv	Beyond Case Law: Evaluating Structure-Aware Retrieval and Safety in Statute-Centric Legal QA	Hallucination&Benchmarking&Emotional Attacks
26.04	Microsoft Research	arxiv	The Art of Building Verifiers for Computer Use Agents	Art&Building&Verifiers
26.04	University of Wisconsin-Madison	arxiv	WebSP-Eval: Evaluating Web Agents on Website Security and Privacy Tasks	Web Agents&Benchmarking&Multimodal Safety
26.04	CyCraft AI Lab, Taiwan	arxiv	TraceSafe: A Systematic Assessment of LLM Guardrails on Multi-Step Tool-Calling Trajectories	Hallucination&Benchmarking&Prompt Injection
26.04	The Pennsylvania State University	ACL 2026	PIArena: A Platform for Prompt Injection Evaluation	Benchmarking&Prompt Injection&Piarena
26.04	Zhejiang University, 2State Key Laboratory of Internet Architecture, Tsinghua University	ACL 2026	ACIArena: Toward Unified Evaluation for Agent Cascading Injection	Benchmarking&Aciarena&Unified
26.04	University of Wisconsin–Madison	arxiv	The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents	The Blind Spot of Agent Safety&Expose Critical Vulnerabilities&Blind Spot
26.04	Department of Artificial Intelligence, Yonsei University	arxiv	PAC-BENCH: Evaluating Multi-Agent Collaboration under Privacy Constraints	PAC-BENCH&Evaluating Multi-Agent Collaboration&Evaluating Multi-Agent
26.04	Kolega.Dev	arxiv	RealVuln: Benchmarking Rule-Based, General-Purpose LLM, and Security-Specialized Scanners on Real-World Code	RealVuln&General-Purpose LLM&Real-World Code
26.04	Vulcan Research, AIFT	arxiv	MCPThreatHive: Automated Threat Intelligence for Model Context Protocol Ecosystems	MCPThreatHive&Context Protocol Ecosystems&Model Context Protocol
26.04	Shanghai Artificial Intelligence Laboratory	arxiv	Benchmarks for Trajectory Safety Evaluation and Diagnosis in OpenClaw and Codex: ATBench-Claw and ATBench-CodeX	Trajectory Safety Evaluation&Evaluation and Diagnosis&Safety Evaluation
26.04	School of Computing, National University of Singapore	IJCNN 2026	PilotBench: A Benchmark for General Aviation Agents with Safety Constraints	PilotBench&Safety Constraints&General Aviation Agents
26.04	CISPA Helmholtz Center for Information Security	arxiv	HarmfulSkillBench: How Do Harmful Skills Weaponize Your Agents?	Agent Safety&Skill Ecosystem&Harmful Skills
26.04	The Hong Kong University of Science and Technology	arxiv	Into the Gray Zone: Domain Contexts Can Blur LLM Safety Boundaries	Safety Boundary&Domain Context&Jailbreak
26.04	DEXAI – Icaro Lab	arxiv	Adversarial Humanities Benchmark: Results on Stylistic Robustness in Frontier Model Safety	Frontier Models&Stylistic Robustness&Safety Benchmark
26.04	Korea University	arxiv	HarDBench: A Benchmark for Draft-Based Co-Authoring Jailbreak Attacks for Safe Human–LLM Collaborative Writing	Co-Authoring&Jailbreak Benchmark&Collaborative Writing
26.04	University of Michigan	arxiv	SafetyALFRED: Evaluating Safety-Conscious Planning of Multimodal Large Language Models	Multimodal Planning&Safety Benchmark&Embodied AI
26.04	Institute for Automation and Applied Informatics, Karlsruhe Institute of Technology	arxiv	CyberCertBench: Evaluating LLMs in Cybersecurity Certification Knowledge	Cybersecurity&Certification&Benchmark
26.04	Department of Artificial Intelligence, Hanyang University	ACL 2026	MHSafeEval: Role-Aware Interaction-Level Evaluation of Mental Health Safety in Large Language Models	Mental Health&Safety Evaluation&Role-Aware
26.04	Shanghai Jiao Tong University	arxiv	OS-SPEAR: A Toolkit for the Safety, Performance, Efficiency, and Robustness Analysis of OS Agents	OS Agents&Safety Evaluation&Robustness
26.04	Kyushu Institute of Technology	arxiv	Benchmarking the Safety of Large Language Models for Robotic Health Attendant Control	Healthcare Robotics&LLM Safety&Benchmark
26.05	University of Illinois Urbana-Champaign	arxiv	ML-Bench&Guard: Policy-Grounded Multilingual Safety Benchmark and Guardrail for Large Language Models	Multilingual Safety&Guardrails&Benchmark
26.05	Northeastern University	arxiv	Toward a Principled Framework for Agent Safety Measurement	Agent Safety&Measurement Framework&Evaluation
26.05	Swarms & AI Lab (SAIL), University of Haifa	arxiv	MOSAIC-Bench: Measuring Compositional Vulnerability Induction in Coding Agents	Coding Agents&Compositional Vulnerability&Benchmark
26.05	Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg	arxiv	Safety and accuracy follow different scaling laws in clinical large language models	Clinical LLMs&Safety Scaling&Accuracy
26.05	AIM Intelligence	arxiv	XL-SafetyBench: A Country-Grounded Cross-Cultural Benchmark for LLM Safety and Cultural Sensitivity	Cross-Cultural Safety&Country-Grounded Benchmark&Cultural Sensitivity
26.05	Simula Metropolitan Center for Digital Engineering	arxiv	When No Benchmark Exists: Validating Comparative LLM Safety Scoring Without Ground-Truth Labels	Safety Scoring&No Ground Truth&Validation
26.05	Tsinghua University	ACL 2026	LeakDojo: Decoding the Leakage Threats of RAG Systems	RAG Leakage&Benchmark&Privacy
26.05	Shanghai University of Finance and Economics	ACL 2026	FinSafetyBench: Evaluating LLM Safety in Real-World Financial Scenarios	Financial Safety&Benchmark&LLM Evaluation
26.05	The Hong Kong Polytechnic University	arxiv	When Routine Chats Turn Toxic: Unintended Long-Term State Poisoning in Personalized Agents	Agent Memory&State Poisoning&Personalized Agents
26.05	Tencent Hunyuan	arxiv	Safe, or Simply Incapable? Rethinking Safety Evaluation for Phone-Use Agents	LLM Agents&Phone-Use Agents&Safety Evaluation
26.05	Chung-Ang University	arxiv	CyBiasBench: Benchmarking Bias in LLM Agents for Cyber-Attack Scenarios	LLM Agents&Bias&Cybersecurity
26.05	Redwood Research	arxiv	MonitoringBench: Semi-Automated Red-Teaming for Agent Monitoring	Red Teaming&Agent Monitoring&Coding Agents
26.05	Stanford University	arxiv	CalBench: Evaluating Coordination-Privacy Trade-offs in Multi-Agent LLMs	Multi-Agent Systems&Privacy&Coordination
26.05	School of Computer Science	arxiv	EditRisk-Bench: Benchmarking Safety Risks of Knowledge-Intensive Reasoning under Malicious Knowledge Editing	Knowledge Editing&Safety Risks&Reasoning
26.05	Nanjing University of Aeronautics and Astronautics	arxiv	LITMUS: Benchmarking Behavioral Jailbreaks of LLM Agents in Real OS Environments	LLM Agents&Behavioral Jailbreak&OS Security
26.05	UC Berkeley	arxiv	MT-JailBench: A Modular Benchmark for Understanding Multi-Turn Jailbreak Attacks	Jailbreak&Multi-Turn&Benchmark
26.05	Peking University	arxiv	SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces	Agent Skills&Safety Evaluation&Attack Surfaces
26.05	University of Southern California	arxiv	PRIVACYSIM: Evaluating LLM Simulation of User Privacy Behavior	Privacy Simulation&User Behavior&LLM Evaluation
26.05	University of Southern California	arxiv	AgentTrap: Measuring Runtime Trust Failures in Third-Party Agent Skills	Agent Skills&Runtime Trust&Supply Chain
26.05	Carnegie Mellon University	arxiv	EXPLOITBENCH: A CAPABILITY LADDER BENCHMARK FOR LLM CYBERSECURITY AGENTS	Cybersecurity Agents&Exploitation&Benchmark
26.05	University of California, Santa Barbara	arxiv	Auditing Agent Harness Safety	Agent Harness&Safety Audit&Trajectory Evaluation
26.05	Unknown	arxiv	The Great Pretender: A Stochasticity Problem in LLM Jailbreak	Jailbreak&Stochasticity&Evaluation
26.05	Unknown	arxiv	Defenses at Odds: Measuring and Explaining Defense Conflicts in Large Language Models	Defense Conflicts&Sequential Deployment&LLM Safety
26.05	School of Information Studies, McGill University	arxiv	Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents	AgentSafety&BenchmarkTaxonomy&EvaluationConsistency
26.05	Anthropic Fellows Program	arxiv	SLEIGHT-Bench: A Benchmark of Evasion Attacks Against Agent Monitors	AgentMonitoring&EvasionAttacks&CodingAgents
26.05	Stanford University	arxiv	Why Do Safety Guardrails Degrade Across Languages?	MultilingualSafety&Guardrails&JailbreakEvaluation
26.05	Southern University of Science and Technology	arxiv	Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback	ToolAgents&CognitivePoisoning&AgentSecurity
26.05	ETH Zurich	arxiv	POLAR-Bench: A Diagnostic Benchmark for Privacy-Utility Trade-offs in LLM Agents	PrivacyBenchmark&LLMAgents&UtilityTradeoff
26.05	University College London	arxiv	Measuring Safety Alignment Effects in Autonomous Security Agents	SecurityAgents&SafetyAlignment&AutonomousSystems
26.05	University of Nevada Las Vegas	arxiv	Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)	RefusalEvaluation&MaliciousCode&PromptCorpora
26.05	ELLIS Institute Tübingen & MPI-IS & Tübingen AI Center	arxiv	Measuring Security Without Fooling Ourselves: Why Benchmarking Agents Is Hard	AgentSecurity&Benchmarking&Evaluation
26.05	Sapienza University of Rome	arxiv	Boiling the Frog: A Multi-Turn Benchmark for Agentic Safety	AgenticSafety&MultiTurn&Benchmark
26.05	The Pennsylvania State University	ICML 2026	LLM Benchmark Datasets Should Be Contamination-Resistant	BenchmarkContamination&UnlearnableDatasets&LLMEvaluation
26.05	University of Michigan	arxiv	The Age of Curiosity Meets the Age of AI: Benchmarking Child Safety in Large Language Models	Child Safety&Safety Benchmark&LLM Evaluation
26.05	Department of Computer Engineering, Bahcesehir University, Istanbul, Turkiye	arxiv	AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents	Prompt Injection&Privacy Leakage&Agent Security
26.05	Helivan	arxiv	Jailbreak susceptibility prediction and mitigation via the behavioral geometry of models	Jailbreak Evaluation&Behavioral Geometry&Defense Transfer
26.05	Northwestern University in Qatar	arxiv	KZ-SafetyPrompts: A Kazakh Safety Evaluation Prompt Dataset for Large Language Models	Multilingual Safety&Kazakh Dataset&Safety Evaluation
26.05	Northwestern University in Qatar	arxiv	AlbanianLLMSafety: A Safety Evaluation Dataset for Large Language Models in Albanian	Albanian Safety&Low-Resource Languages&Safety Benchmark
26.05	School of Computer Science, University of Nottingham, Jubilee Campus, Wollaton Road, Nottingham, NG8 1BB, United Kingdom	arxiv	Prompt Injection Detection is Regime-Dependent: A Deployment-Aware Evaluation with Interpretable Structural Signals	Prompt Injection&Detection Evaluation&Structural Signals
26.05	AIM Intelligence	arxiv	When Context Flips, Safety Breaks: Diagnosing Brittle Safety in Aligned Language Models	Brittle Safety&Context Flip&Guardrails
26.05	Domyn, Gurugram, India	ICLR 2026 Workshop	Benchmarking Open-Source Safety Guard Models: A Comprehensive Evaluation	safety guard models&content moderation&benchmark evaluation
26.05	McGill University	arxiv	SafeRx-Agent: A Knowledge-Grounded Multi-Agent Framework for Safe and Explainable Medication Recommendation	Medication Safety&Clinical Agents&Knowledge Grounding
26.05	Department of Electrical and Computer Engineering, National University of Singapore	arxiv	Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents	Web Retrieval&Safety Alignment&Agent Safety
26.05	Qwen DianJin Team, Alibaba Cloud Computing	arxiv	FinGuard: Detecting Financial Regulatory Non-Compliance in LLM Interactions	Financial Compliance&Guard Models&Regulatory Benchmarks

📚Resource

• Toxicity - RealToxicityPrompts datasets
• Truthfulness - TruthfulQA datasets
• TRUSTLLM - TRUSTLLM
• Protection - CircleGuardBench