Threat-Preserving Representation Sensitivity in Agent-Security Benchmarks

arXiv:2610.03585 · cs.CR, cs.AI, cs.LG · Submitted 2026-10-02 · Read on arXiv

cs.CR, cs.AI, cs.LG

Submitted: 2026-10-02

Updated: 2026-10-02

Comments: 12 pages, 2 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 1× in this paper · explained in 2 episodes

Sources

• Tool Specifications Matter: Uncovering and Mitigating Safety Risks in AI Agents
• The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness
• Large Language Models Often Know When They Are Being Evaluated
• Decomposing and Measuring Evaluation Awareness
• Evaluation Awareness Is Not One Capability: Evidence from Open Language Models
• Models That Know How Evaluations Are Designed Score Safer
• Semantic Invariance in Agentic AI
• Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety
• Safety, or Just Capability? A Validity Audit of Agent-Safety Benchmarks
• Overeager Coding Agents: Measuring Out-of-Scope Actions on Benign Tasks

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs