CHILLGuard: Towards Fine-Grained Chinese LLM Safety Guardrail with Scalable Data Construction and Model-aware Preference Alignment
Paper • 2606.15396 • Published
CHILLGuard
Official resources for CHILLGuard (EMNLP 2026 Findings): guardrail models (1.7B/4B/8B), CHILLGuardTest benchmark, 15K training subset, and the paper.