Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain Regularization

2026-08-08

AI Summary

This work addresses the challenge of detecting harmful dialogue, which continuously evolves through type shifting and lexical evasion. To tackle this, the authors propose the BRACE framework, which formalizes the invariant ordered reasoning chain (ORC)—comprising four differentiable stages: topic → indicator → severity → type—and integrates it as a structured regularization term into the model. Through mechanisms including intermediate supervision, prototype feature enhancement, and path disentanglement, BRACE effectively mitigates type confusion under semantic ambiguity. Evaluated across five categories of harmful content spanning four domains, the framework achieves macro F1 scores of 0.934 and 0.949 using RoBERTa-wwm-ext and a Qwen3-1.7B LoRA decoder, respectively, demonstrating significant improvements in dynamic harmful content detection.

Abstract

Harmful chat dialogues are ever-shifting through type-shifting and lexical evasion, yet we find they share invariant principles, i.e., an Ordered Reasoning Chain (ORC) of recurring topics, harm language indicators, severity hierarchies, and type characteristics, which can help us capture the key information in the frequently changing lexical expressions. We propose BRACE, 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. The evaluation results show that, across 4 domains and 5 harm categories, BRACE achieves harm-type macro F1 of 0.934 (RoBERTa-wwm-ext, 3-seed mean), with decoder backbones (Qwen3-1.7B LoRA) reaching 0.949. Ablation studies show that all components contribute to BRACE, and the structural decomposition of ORC enables BRACE to distinguish harmful types with semantic ambiguity. Disclaimer: This paper may contain content that is disturbing to some readers.

Problem

• harmful chat dialogue
• type-shifting
• lexical evasion
• harm detection
• semantic ambiguity

Innovation

• Ordered Reasoning Chain
• harmful dialogue detection
• structured regularization
• prototype-based augmentation
• feature disentanglement