Announcement_7

🏷 Fanar · Cultural Alignment & Benchmarking   Our paper The Alignment Veto: How Safety Training Suppresses Cultural Knowledge in LLMs was accepted at EMNLP 2026 — a diagnostic study of how safety-alignment training systematically suppresses culturally grounded knowledge in large language models. Congratulations to Pardis Sadat Zahraei, and to our collaborators Gokhan Tur and Dilek Hakkani-Tür at UIUC.