benign

Tag

Cards List
#benign

Be Kind, Rewrite: Benign Projections via Rewriting Defend Against LLM Data Poisoning Attacks

Hugging Face Daily Papers · 2026-05-18 Cached

This paper introduces open-book benign rewriting (OBBR) as a proactive defense against backdoor attacks on LLMs, showing it neutralizes harmful content by projecting to benign prompts, and improves safety by 51% over state-of-the-art defenses.

0 favorites 0 likes
← Back to home

Submit Feedback