Image de couverture de livre

Defensive Refusal Bias: How Safety Alignment Fails Cyber Defenders

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Campbell, David, Kale, Neil, Sehwag, Udari Madhushani, Herring, Bert, Price, Nick, Borges, Dan, Levinson, Alex, Knight, Christina Q
Format:	Preprint
Publié:	2026
Sujets:	Cryptography and Security Artificial Intelligence
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

PropensityBench: Evaluating Latent Safety Risks in Large Language Models via an Agentic Approach
par: Sehwag, Udari Madhushani, et autres
Publié: (2025)

AdvBDGen: Adversarially Fortified Prompt-Specific Fuzzy Backdoor Generator Against LLM Alignment
par: Pathmanathan, Pankayaraj, et autres
Publié: (2024)

In-Context Learning with Topological Information for Knowledge Graph Completion
par: Sehwag, Udari Madhushani, et autres
Publié: (2024)

ROK-FORTRESS: Measuring the Effect of Geopolitical Transcreation for National Security and Public Safety
par: Lee, Michael S., et autres
Publié: (2026)

GenARM: Reward Guided Generation with Autoregressive Reward Model for Test-time Alignment
par: Xu, Yuancheng, et autres
Publié: (2024)

Can LLMs be Scammed? A Baseline Measurement Study
par: Sehwag, Udari Madhushani, et autres
Publié: (2024)

SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal
par: Xie, Tinghao, et autres
Publié: (2024)

ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM Agents
par: Sehwag, Udari Madhushani, et autres
Publié: (2026)

Continual Learning of Domain Knowledge from Human Feedback in Text-to-SQL
par: Cook, Thomas, et autres
Publié: (2025)

Collab: Controlled Decoding using Mixture of Agents for LLM Alignment
par: Chakraborty, Souradip, et autres
Publié: (2025)

Refusal Falls off a Cliff: How Safety Alignment Fails in Reasoning?
par: Yin, Qingyu, et autres
Publié: (2025)

AgentCrypt: Advancing Privacy and (Secure) Computation in AI Agent Collaboration
par: Karthikeyan, Harish, et autres
Publié: (2025)

LHAW: Controllable Underspecification for Long-Horizon Tasks
par: Pu, George, et autres
Publié: (2026)

Detection Is Cheap, Routing Is Learned: Why Refusal-Based Alignment Evaluation Fails
par: Frank, Gregory N.
Publié: (2026)

Large Language Models are Autonomous Cyber Defenders
par: Castro, Sebastián R., et autres
Publié: (2025)

When Refusals Fail: Unstable Safety Mechanisms in Long-Context LLM Agents
par: Hadeliya, Tsimur, et autres
Publié: (2025)

Uplifted Attackers, Human Defenders: The Cyber Offense-Defense Balance for Trailing-Edge Organizations
par: Murphy, Benjamin, et autres
Publié: (2025)

Best Practices for Biorisk Evaluations on Open-Weight Bio-Foundation Models
par: Wei, Boyi, et autres
Publié: (2025)

The Art of Defending: A Systematic Evaluation and Analysis of LLM Defense Strategies on Safety and Over-Defensiveness
par: Varshney, Neeraj, et autres
Publié: (2023)

Deactivating Refusal Triggers: Understanding and Mitigating Overrefusal in Safety Alignment
par: Xue, Zhiyu, et autres
Publié: (2026)

Poison Once, Refuse Forever: Weaponizing Alignment for Injecting Bias in LLMs
par: Mamun, Md Abdullah Al, et autres
Publié: (2025)

Asymmetry by Design: Boosting Cyber Defenders with Differential Access to AI
par: Ee, Shaun, et autres
Publié: (2025)

O3D: Offline Data-driven Discovery and Distillation for Sequential Decision-Making with Large Language Models
par: Xiao, Yuchen, et autres
Publié: (2023)

A Case Study on the Use of Representativeness Bias as a Defense Against Adversarial Cyber Threats
par: Hitaj, Briland, et autres
Publié: (2025)

Oyster-I: Beyond Refusal -- Constructive Safety Alignment for Responsible Language Models
par: Duan, Ranjie, et autres
Publié: (2025)

From Threat to Tool: Leveraging Refusal-Aware Injection Attacks for Safety Alignment
par: Chae, Kyubyung, et autres
Publié: (2025)

CyberAlly: Leveraging LLMs and Knowledge Graphs to Empower Cyber Defenders
par: Kim, Minjune, et autres
Publié: (2025)

MoReBench: Evaluating Procedural and Pluralistic Moral Reasoning in Language Models, More than Outcomes
par: Chiu, Yu Ying, et autres
Publié: (2025)

Toward a Logic of Generalization about Visualization as a Decision Aid
par: Kale, Alex
Publié: (2025)

A Red Teaming Roadmap Towards System-Level Safety
par: Wang, Zifan, et autres
Publié: (2025)

Reliable Weak-to-Strong Monitoring of LLM Agents
par: Kale, Neil, et autres
Publié: (2025)

The Defense Trilemma: Why Prompt Injection Defense Wrappers Fail?
par: Bhatt, Manish, et autres
Publié: (2026)

When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment
par: Xiao, Yuxin, et autres
Publié: (2025)

When Embedding-Based Defenses Fail: Rethinking Safety in LLM-Based Multi-Agent Systems
par: Zhang, Lingxi, et autres
Publié: (2026)

Beyond Surface Alignment: Rebuilding LLMs Safety Mechanism via Probabilistically Ablating Refusal Direction
par: Xie, Yuanbo, et autres
Publié: (2025)

The Path To Autonomous Cyber Defense
par: Oesch, Sean, et autres
Publié: (2024)

Think Before Refusal : Triggering Safety Reflection in LLMs to Mitigate False Refusal Behavior
par: Si, Shengyun, et autres
Publié: (2025)

FedDefender: Backdoor Attack Defense in Federated Learning
par: Gill, Waris, et autres
Publié: (2023)

MCP Safety Training: Learning to Refuse Falsely Benign MCP Exploits using Improved Preference Alignment
par: Halloran, John
Publié: (2025)

Characterizing Selective Refusal Bias in Large Language Models
par: Khorramrouz, Adel, et autres
Publié: (2025)