作者位次 #3/3ACL 2026Poster· Natural Language Processing
Reasoning Hijacking: The Fragility of Reasoning Alignment in Large Language Models
Yuansen Liu · Yixuan Tang · Anthony Kum Hoe Tung
摘要
Current LLM safety research predominantly focuses on mitigating **Goal Hijacking**, preventing attackers from redirecting a model’s high-level objective (e.g., from "summarizing emails" to "phishing users"). In this paper, we argue that this perspective is incomplete and highlight a critical vuln...