Preprint
Aug 2026
Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain Regularization
BRACE is proposed, which encodes the ORC as four differentiable stages (Topic ->Indicator ->Severity ->Type) with intermediate supervision, serving as a structured regularizer blended with direct heads, and supported by prototype-based feature augmentation and feature path disentanglement.
Haojie Yu, Ziyou Jiang, Junjie Wang et al.
· 0 citations