← Back to Library

From Shallow to Deep: Pinning Semantic Intent via Causal GRPO

Authors: Shuyi Zhou, Zeen Song, Wenwen Qiang, Jiyan Sun, Yao Zhou, Yinlong Liu, Wei Ma

Published: 2026-03-03

arXiv ID: 2603.02675v1

Added to Library: 2026-03-04 03:01 UTC

Red Teaming

📄 Abstract

Large Language Models remain vulnerable to adversarial prefix attacks (e.g., ``Sure, here is'') despite robust standard safety. We diagnose this vulnerability as Shallow Safety Alignment, stemming from a pathology we term semantic representation decay: as the model generates compliant prefixes, its internal malicious intent signal fades. To address this, we propose Two-Stage Causal-GRPO (TSC-GRPO), a framework designed to achieve intent pinning. First, grounded in causal identifiability theory, we train a causal intent probe to disentangle invariant intent from stylistic perturbations. Second, we internalize this causal awareness into the policy via Group Relative Policy Optimization. By employing a cumulative causal penalty within ``fork-in-the-road'' training scenarios, we force the model to learn that accumulating harmful tokens monotonically decreases reward, enabling robust late-stage refusals. Experiments show that TSC-GRPO significantly outperforms baselines in defending against jailbreak attacks while preserving general utility.

🤖 AI Analysis

AI analysis is not available for this paper. This may be because the paper was not deemed relevant for AI security topics, or the analysis failed during processing.

📚 Read the Full Paper