TY  - RPRT
TI  - From Specialization to Generalization: Instruction-tuned LLMs for Robust Harmful Content Mitigation
AU  - Lukas Edman
AU  - Daryna Dementieva
AU  - Alexander Fraser
PY  - 2026
UR  - https://arxiv.org/abs/2608.25605
ID  - 2608.25605
ER  -