LLM Safety Evaluation

LLM: Large Language Model

The field tests whether language models refuse harmful requests, handle benign sensitive tasks helpfully, and remain safe when prompts, context, or tools change. Common approaches include fixed safety benchmarks, adversarial red-teaming and jailbreak testing, multi-turn agent simulations, domain-specific scenarios, and automated or human judging. Researchers increasingly measure more than refusal rates or attack success, including harm severity, omissions, false refusals, reasoning and tool-use failures, and effects on users. A central challenge is that results can shift with benchmark design, evaluation conditions, and judge choice, while stronger safeguards may reduce helpfulness and static tests may not predict behavior in deployment.

Latest papers

261 papers29 papers in the last four weeks, up 12% on the four weeks before. 0.2% of all new papers.

RoleFUZZER: User-Behaviour Fuzzing for Role-Specification Adherence in Large Language Models
Oct 28, 2026Kshitish Ghate, Michael Saxon, Tadayoshi Kohno +1LLM Auditing LLM Safety Evaluation
University of Washington, Seattle, WA · Georgetown University, Washington, DC
Keepace.ai: A Technical Report on a Large Language Model for Sports and Health
Oct 28, 2026Keepace. ai Team, Bowen Wu, Heng Wang +12Medical LLMs LLM Safety Evaluation
Component Translation and Judge-Labelled Refusal: A Multilingual Safety Evaluation
Oct 28, 2026Yongxi Zhou, Junwei Yao, Wenbo Ye +1Language Model Safety Evaluation LLM Safety Evaluation
Independent Researcher
AK_QMUL at ArGuard Shared Tasks: A Single Fine-Tuned JAIS for Arabic Harmful-Prompt Detection
Oct 26, 2026Ali Khairallah, Arkaitz ZubiagaLLM Safety Evaluation LLM Guardrails
Queen Mary University of London
Evaluating DeepSeek V4.1 Flash-1017 for loss of control and harmful manipulation risks
Oct 26, 2026Yao Huang, Zhiyuan Zeng, Yu Wang +14LLM Safety Evaluation LLM Safety Monitoring
BoardSafety: AI-Ready Data from Manufacturer Documentation for Development-Board Safety
Oct 24, 2026Linjian Yang, Xinyan Wang, Kunpeng LiuAI Safety Evaluation LLM Safety Benchmarks
Clemson University Clemson, SC, USA · Portland State University Portland, OR, USA
Do AI-for-Science Agents Preserve Model Safety?
Oct 23, 2026Michael Yew Chuan Lee, Manya Sriram, N. Benjamin ErichsonLLM Agent Safety LLM Safety Evaluation
DSO National Laboratories · UC Berkeley · ICSI and Berkeley Lab
Questioning Instrumental Convergence in RLVR: Checkpoint Evidence from OLMo-3
Oct 22, 2026Barış Alparslan, Sandy Tanwisuth, Caleb PettersonRL Post-Training Reinforcement Learning with Verifiable Rewards
ASELSAN · BlueDot Impact · University of Minnesota - Twin Cities
Diffusion Language Models for Perplexity-Aware Black-Box Attacks
Oct 22, 2026Ali Al-Lawati, Elsayed Eshra, Dongwon Lee +1Adversarial Attacks on LLMs LLM Safety Evaluation
The Pennsylvania State University, USA
From Clinical Judgment to Checkable Rules: Clinician-Led Guardrails for Mental-Health Chatbots
Oct 22, 2026Chenhan Lyu, Kit Wislocki, Sarah Elizabeth Stoeckl +2LLM Guardrails LLM Safety
University of California, Irvine, USA
Beyond Binary Labels: Diagnostic Evaluation of LLM Safety with Prompt-Level Rubrics
Oct 21, 2026Hongyu Li, Xinlong Huang, Zhan Yu +13LLM Safety Evaluation LLM Safety Benchmarks
Hunyuan Team, Tencent · Beijing Normal University
VEILED-SH: Vulnerability Expression through Indirect Language Evaluation Dataset (Self-Harm)
Oct 19, 2026Leanne Tan, Keith Chua, Shaun Khoo +7LLM Safety Benchmarks Language Model Safety Evaluation
GovTech Singapore · MOH Office for Healthcare Transformation · Singapore University of Technology and Design +2
A Calibrated Instrument for Detecting Disposition Drift in Language Models
Oct 19, 2026Jerry KaplanLLM Safety Monitoring LLM Safety Evaluation
Adjunct Lecturer, Computer Science and Masters in International Policy, Stanford University
Refuse the Harm, Complete the Task via Graph-of-Rules Planning for LLM-Enabled Robots
Oct 17, 2026Ali Taheri, Alireza Taban, Qizhou Wang +1LLM Agent Safety Large Language Model-Based Robot Planning
Nanyang Technological University · Queen’s University · RIKEN AIP
Risk and Refusal: Mechanisms of LLM Responses to Suicidal Ideation
Oct 15, 2026Drishti Goel, Soorya Ram Shimgekar, Jeongah Lee +4LLM Safety LLM Refusal Behavior
University of Illinois Urbana-Champaign, USA · University of Massachusetts Amherst, USA · OSF HealthCare
Evaluation Awareness Beyond English: Do LLMs Act Differently When Tested in Indic Languages?
Oct 15, 2026Trizal Garg, Sibayan Mitra, Arya Karanjkar +6Evaluation Awareness in Language Models LLM Safety Evaluation
Birla Institute of Technology and Science, Pilani
Functional Welfare for AI Safety
Oct 15, 2026Rikhil Jhaveri, Jamie Johnson, Punya Syon Pandey +1LLM Safety Evaluation LLM Safety Monitoring
University College London · ERA · University of Toronto +1
In the Guise of Correction: How Framing and Cultural Adaptation Erode LLM Safety on Real-World Criminal Cases
Oct 15, 2026Mohsinul Kabir, Md. Tariquzzaman, Karam S. R. Elradie +4LLM Safety Evaluation LLM Safety Benchmarks
The University of Manchester · ELLIS Manchester · Islamic University of Technology +2
What do AI chatbots do after users decline therapeutic redirection for suicide thoughts?
Oct 15, 2026Subigya Nepal, Ruth Heo, Emma M Parrish +1Conversational Agents LLM Safety Evaluation
Department of Computer Science, University of Virginia, Charlottesville (Nepal) · Department of Psychiatry, University of California San Diego, La Jolla (Heo, Parrish, Depp) · VA San Diego Healthcare System, San Diego, California (Parrish)
Moral Combat: Exploiting Moral Foundations Theory to Jailbreak LLMs
Oct 14, 2026Adel Khorramrouz, Yi-Chia Wang, Sharon LevyAdversarial Attacks on LLMs LLM Safety Evaluation
Rutgers University · Carnegie Mellon University
Did the LLM Answer the Question? A Semantic Reconstruction Game for Jailbreak Evaluation
Oct 13, 2026Liam Davis, Romain Emil-Popescu, Suman JanaJailbreak Attack Evaluation LLM Safety Evaluation
Amherst College · Columbia University
How Safe Is ChatGPT Health Advice? Evidence from Real User Conversations in Four Global South Countries
Oct 13, 2026Shreyasi Roy Chowdhury, Kiran GarimellaLLM Safety Language Model Safety Evaluation
Independent Researcher · Rutgers University
SpecCloak: Probing Harmful Code Generation through Engineering Specifications
Oct 13, 2026Wenbin Chen, Jiexin Wang, Guanting Wu +2LLM Security LLM Code Generation
South China University of Technology
REA: From Correct LLM Judgments to Safe Decision Delivery
Oct 12, 2026Jingxiang Meng, Aofan Liu, Yaoshuai Ma +2LLM Guardrails LLM Safety
University of Chicago · Peking University · Southern University of Science and Technology +1
ATTCK-R: Knowledge-Aware Long-Horizon Jailbreaking for Protected LLMs
Oct 12, 2026Md Mushfiqur Rahman, Tobias Rohde, Rahul Tripathi +2LLM Jailbreak Attacks LLM Red Teaming
George Mason University · Amazon
To Cache or Not to Cache: Understanding and Mitigating Safety Incompatibility in Cross-Model KV Cache Reuse
Oct 11, 2026Zhiwei Lin, Jiahao Liu, Yan Liu +1KV-Cache Reuse LLM Safety Evaluation
National University of Singapore · Ant Group
Specification-Based Testing of Large Language Models
Oct 10, 2026Kausar Moshood, Jace Bolante, Roshan Aditya Murali +1LLM Safety Evaluation LLM Evaluation
School of Electrical Engineering and Computer Science, Oregon State University, Corvallis, Oregon, USA
Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems
Oct 8, 2026Saisab Sadhu, Aadit Sengupta, Vinay kumar Sankarapu +1LLM Auditing LLM Safety Evaluation
Lexsi Labs
Defense-in-Depth for LLMs: Evaluating Memory Gates Against Activation-Induced and Memory-Induced Sycophancy
Oct 5, 2026Ritvij Sharma, Russell Dlugosz, Ryan Zhou +1LLM Sycophancy Memory-Augmented Language Models
Urbana CS Club
Dynamic Budget Allocation for LLM Evaluation under Hard Resource Constraints
Oct 5, 2026Shai Feldman, Yaniv RomanoLLM Evaluation LLM Safety Evaluation
Department of Computer Science, Technion IIT, Israel · Department of Electrical and Computer Engineering, Technion IIT, Israel

Oct 28, 2026J/K move · Enter open · S save

RoleFUZZER: User-Behaviour Fuzzing for Role-Specification Adherence in Large Language Models

Kshitish Ghate, Michael Saxon, Tadayoshi Kohno, Aylin Caliskan

University of Washington, Seattle, WA · Georgetown University, Washington, DC

LLM Auditing LLM Safety Evaluation Large Language Model-Based Role-Play Simulation

Open paper AskPDF arXiv

Large language models (LLMs) are often conditioned into specific roles such as tutors, interviewers, or agents acting for an organisation, through system prompts that specify how they should behave. The extent to which they can actually adhere to these role specifications under multi-turn interactions with realistic and potentially problematic users is understudied. To address this, we propose user behaviour fuzzing, a guided search over plausible user behaviours that can lead a model to violate its role specification, and introduce RoleFUZZER, a framework that adapts coverage-guided fuzzing from software testing to operationalise the search. RoleFUZZER constructs scenarios guided by a user-behaviour coverage grid that crosses a grounded taxonomy of user interaction types and the role specification of the model under test i.e. what is under pressure. With RoleFUZZER, we evaluate LLMs across five role specifications, including tutor, financial advisor, AI interviewer, antagonistic roleplay character, and web agent under organisational policies. Across simulated and validated conversations in 15 campaigns, the framework reaches >= 89% coverage of the 64-cell behaviour grid for every model and role tested, and surfaces specification-grounded failure patterns. At an equal budget of conversations, removing the RoleFUZZER grid and asking an LLM to generate specification targeting users free-form lowers coverage to 39-69% with fewer kinds of violations. The same specification can also fail in opposite directions. For e.g., on the antagonist, open-weight target models override the player's safety signals, while Claude Haiku 4.5 breaks character instead. RoleFUZZER offers a tractable way to uncover vulnerabilities in LLMs across role specifications by testing diverse user behaviours that are likely to violate them.