arXivHe Liu, Changtao Miao, Xinjie Yang, Tianle Song, Yin Wu, Junchi Chen, Bintao He, Xinyuan Zhang, Bo Zhang, Shi Yan, Wei Lu, Wei Wang, Danyang Xu, Jiansheng Cai, Zhe LiTue, Jul 7, 2026, 7:25 AM PDT
score 17.0
LLM safety guardrail learns reasoning during training but runs fast at inference
Original: DT-Guard: Intent-Driven Reasoning-Active Training for Reasoning-Free LLM Safety Guardrail
Source: arxiv.org ↗
Writing ELI5 summary…