Authors

Mahammed El Sharkawy

Publications 2

Preprint Open access

PatchBench: Measuring Collateral Damage in Activation Patching

Alexi Canesse, Mathis Le Bail, Maël Jenny et al. · 2026

An LLM safety patch can pass a benchmark while still being a poor repair. This risk is especially acute for jailbreak repairs, where the goal is to correct a specific unsafe behaviour without changing unrelated behaviours. A patch may block exact evaluation prompts yet fail on close harmful variants, or suppress harmfu …

Preprint Open access

RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents

Mohamed Dhouib, Clement Elliker, Alexi Canesse et al. · 2026

Tool-using language-model agents are vulnerable to indirect prompt injection because they must act on untrusted external content. Existing training-time defenses can reduce attack success rates, but often at the cost of general capabilities. We show that training-based defenses induce substantial drift in the model's o …

Co-authors