LeakBench: Auditing Reward Hacking through Prediction-Time Leakage in Clinical Modeling Agents

arXiv:2610.26088 · cs.CR, cs.LG · Submitted 2026-10-19 · Read on arXiv

cs.CR, cs.LG

Submitted: 2026-10-19

Updated: 2026-10-19

Comments: 28 pages, 4 figures. LeakBench implementation code is not included in this arXiv version. Public companion software archives provide non-identifying aggregate evidence for two cited real-world motivating results

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

Terminology

• Reward Hacking — 1× in this paper · explained in 11 episodes

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