LLM Safety

LLM: Large Language Model

Follow this topic Ask about this topic

Momentum

36 papers in the last four weeks, up 200% on the four weeks before. 0.4% of all new papers.

Jul 13Week of Sep 28

Latest papers 272

All topics

Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry
May 18, 2026Woo Seob Sim, Yu Rang ParkLanguage Model Safety Evaluation LLM Safety
Department Biomedical Systems Informatics Yonsei University College of Medicine, Seoul, Republic of Korea · Yonsei University
New Wide-Net-Casting Jailbreak Attacks Risk Large Models
May 16, 2026Qiuchi Xiang, Haoxuan Qu, Hossein Rahmani +1Language Model Safety Evaluation Jailbreak Attacks
1Lancaster University.
Distinguishable Deletion: Unifying Knowledge Erasure and Refusal for Large Language Model Unlearning
May 16, 2026Puning Yang, Junchi Yu, Qizhou Wang +3LLM Refusal LLM Safety
Department of Natural Language Processing, MBZUAI. · University of Oxford. · RIKEN Center for Advanced Intelligence Project. +1
Quantifying and Mitigating Premature Closure in Frontier LLMs
May 14, 2026Rebecca Handler, Suhana Bedi, Nigam ShahHealthcare LLM Prompting
Department of Medicine, Stanford University · Department of Biomedical Data Science, Stanford University
Selective Safety Steering via Value-Filtered Decoding
May 14, 2026Bat-Sheva Einbinder, Hen Davidov, Yee Whye Teh +2Language Model Decoding Language Model Steering
Department of Electrical and Computer Engineering, Technion IIT, Israel · Department of Statistics, University of Oxford, Oxford, UK · OATML, Department of Computer Science, University of Oxford, Oxford, UK +1
Mechanical Enforcement for LLM Governance:Evidence of Governance-Task Decoupling in Financial Decision Systems
May 14, 2026José Manuel de la Chica Rodríguez, Carlos Martí-GonzálezLLM Auditing LLM Safety
1Santander AI Lab, Grupo Santander
A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training
May 14, 2026Itay Zloczower, Eyal Lenga, Gilad Gressel +1Language Model Safety Evaluation Adversarial Attacks on LLMs
Ben-Gurion University of the Negev · Amrita Vishwa Vidyapeetham, Amritapuri
Fusion-fission forecasts when AI will shift to undesirable behavior
May 14, 2026Neil F. Johnson, Frank Yingjie HuoAI Safety Evaluation LLM Safety
Physics Department, The George Washington University, Washington DC 20052, USA
GradShield: Alignment Preserving Finetuning
May 13, 2026Zhanhao Hu, Xiao Huang, Patrick Mendoza +4LLM Fine-Tuning LLM Safety Alignment
University of California, Berkeley · 2HUMAIN · 3King Abdulaziz City for Science and Technology +1
Negation Neglect: When models fail to learn negations in training
May 13, 2026Harry Mayne, Lev McKinney, Jan Dubiński +3LLM Reliability LLM Safety
University of Oxford · University of Toronto · 3Warsaw University of Technology +6
Model-Agnostic Lifelong LLM Safety via Externalized Attack-Defense Co-Evolution
May 13, 2026Xiaozhe Zhang, Chaozhuo Li, Hui Liu +4Continual Learning for LLMs Adversarial Attacks on LLMs
City University of Hong Kong · Beijing University of Posts and Telecommunications · Wuhan University +2
Adaptive Steering and Remasking for Safe Generation in Diffusion Language Models
May 13, 2026Yejin Lee, Ungsik Kim, Yo-Sub HanAdversarial Attacks on LLMs Diffusion Language Model Inference
Department of Computer Science Yonsei University Seoul, Republic of Korea
Targeted Neuron Modulation via Contrastive Pair Search
May 12, 2026Sam Herring, Jake Naviasky, Karan MalhotraLanguage Model Steering LLM Interpretability
Nous Research
Metaphor Is Not All Attention Needs
May 12, 2026Olga Sorokoletova, Francesco Giarrusso, Giacomo De Luca +6Adversarial Attacks on LLMs Jailbreak Attacks
Sapienza University of Rome Department of Computer, Control and Management Engineering Sapienza University of Rome Via Ariosto 25, Rome 00185, Italy · DEXAI – Icaro Lab · University of Rome Tor Vergata +1
SafeSteer: A Decoding-level Defense Mechanism for Multimodal Large Language Models
May 12, 2026Xinyi Zeng, Xue Yang, Jingyuan Zhang +5Adversarial Attacks on VLMs Multimodal Robustness
Tsinghua University, Beijing, China · Shanghai Jiao Tong University · Kuaishou Technology, Beijing, China +4
Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing
May 11, 2026Zheng Lin, Zhenxing Niu, Haoxuan Ji +1Randomized Smoothing Jailbreak Robustness
Xidian University · Xi’an Jiaotong University
GuardAD: Safeguarding Autonomous Driving MLLMs via Markovian Safety Logic
May 11, 2026Tianyuan Zhang, Peng Yue, Zihao Peng +8Autonomous Driving LLM Safety
Beihang University · Zhongguancun Laboratory · Peking University
Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
May 11, 2026Huilin Zhou, Jian Zhao, Yilu Zhong +7Inference-Time Optimization Adversarial Attacks on LLMs
University of Science and Technology of China · Institute of Artificial Intelligence (TeleAI), China Telecom
Few-Shot Truly Benign DPO Attack for Jailbreaking LLMs
May 9, 2026Sangyeon Yoon, Wonje Jeung, Yoonjun Cho +2Direct Preference Optimization LLM Fine-Tuning
Yonsei University
Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories
May 9, 2026Dongcheng Zhang, Yi Zhang, Yuxin Chen +3LLM Self-Correction RL for Language Model Reasoning
University of Science and Technology of China · Shanghai Artificial Intelligence Laboratory · National University of Singapore
Internalizing Safety Understanding in Large Reasoning Models via Verification
May 9, 2026Yi Zhang, Yuxin Chen, Leheng Sheng +4LLM Safety Alignment LLM Safety
University of Science and Technology of China · National University of Singapore · Shanghai Artificial Intelligence Laboratory
LLM-Agnostic Semantic Representation Attack
May 9, 2026Jiawei Lian, Jianhong Pan, Lefan Wang +4Adversarial Prompt Generation Adversarial Attacks on LLMs
School of Electronics and Information, Northwestern Polytechnical University, Xi’an 710129, China · Department of Electrical and Electronic Engineering, The Hong Kong Polytechnic University, Hong Kong SAR · Hong Kong Institute of AI for Science, City University of Hong Kong, Hong Kong SAR
Why Do Aligned LLMs Remain Jailbreakable: Refusal-Escape Directions, Operator-Level Sources, and Safety-Utility Trade-off
May 9, 2026Yu Chen, Yuanhao Liu, Qi CaoLLM Alignment Language Model Steering
Institute of Computing Technology, Chinese Academy of Sciences · University of Chinese Academy of Sciences, Beijing, China
GLiGuard: Schema-Conditioned Classification for LLM Safeguard
May 8, 2026Urchade Zaratiana, Mary Newhauser, George Hurn-Maloney +1LLM Guardrails Text Classification
Fastino Labs
LLM Wardens: Mitigating Adversarial Persuasion with Third-Party Conversational Oversight
May 8, 2026Lennart Wachowiak, Scott D. Blain, David Williams-King +1LLM Guardrails LLM Safety
ERA · University of Oxford
How Value Induction Reshapes LLM Behaviour
May 8, 2026Arnav Arora, Natalie Schluter, Katherine Metcalf +1LLM Sycophancy LLM Alignment
University of Copenhagen · 2Apple
CommandSwarm: Safety-Aware Natural Language-to-Behavior-Tree Generation for Robotic Swarms
May 8, 2026Mohammed Majid, Amjad Yousef MajidMulti-Robot Systems Swarm Robotics
Mitigating Many-shot Jailbreak Attacks with One Single Demonstration
May 8, 2026Kejia Chen, Jiawen Zhang, Boheng Li +6Jailbreak Attacks In-Context Learning
Zhejiang University · Nanyang Technological University · University of Science and Technology of China +2
Behavior Cue Reasoning: Monitorable Reasoning Improves Efficiency and Safety through Oversight
May 7, 2026Christopher Z. Cui, Taylor W. Killian, Prithviraj AmmanabroluScalable Oversight LLM Safety
University of California, San Diego · 2Brigham Young University
Safety Anchor: Defending Harmful Fine-tuning via Geometric Bottlenecks
May 7, 2026Guoxin Lu, Letian Sha, Qing Wang +4LLM Fine-Tuning LLM Safety Alignment
Nanjing University of Posts and Telecommunications, Nanjing, China.

May 18, 2026, cs.LGJ/K move · Enter open · S save

Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry

Woo Seob Sim, Yu Rang Park

Department Biomedical Systems Informatics Yonsei University College of Medicine, Seoul, Republic of Korea · Yonsei University

Language Model Safety Evaluation LLM Safety Language Model Probing

Open paper AskPDF arXiv

Prompt-level safety probes for large language models use hidden-state representations to separate safe from unsafe prompts, but strong average detection performance does not explain the geometry of this separation. In particular, it remains unclear how safety evidence is formed across layers, which aspects of that layer-wise geometry support low-false-positive decisions, and which geometric biases remain stable under benchmark shift. We study this as an empirical decomposition problem and introduce Geometry-Lite, a compact prompt-level probe that maps each layer's final prompt-token representation to signed margins under centroid, local-neighborhood, and supervised linear-boundary readouts, then summarizes the resulting margin profiles by boundary position, layer-to-layer change, and coarse shape. Across nine instruction-tuned backbones (

1.2

B--

70

B) and seven safety benchmarks, Geometry-Lite improves over single-layer probes while remaining close to raw multi-layer score stacking, making it a useful instrument for analyzing the multi-layer safety signal. The decomposition shows that safety evidence is expressed primarily through persistent boundary-position geometry: final or extremal margins and unsafe-side layer occupancy dominate aggregate detection performance. In contrast, finite-difference drift and structural summaries add little to pooled AUROC, although drift can provide small recall-oriented corrections under shifted low-FPR thresholds. Under benchmark shift, optimized linear boundaries are sharp on the training mixture, whereas class-conditional mean geometry retains separation more reliably on a predefined hard held-out subset. Overall, prompt-level safety evidence is not primarily a layer-to-layer motion signal, but a persistent layer-wise margin geometry whose useful components and readout-level biases become visible in decision-critical regimes.

Momentum

Related topics

Latest papers 272