一作NeurIPS 2026Poster· Social Aspects / Accountability, Transparency and Interpretability
Taming CoT Obfuscation in VLMs: From Mechanistic Evidence to Activation-Level Enforcement
Xutao Mao · Jianing Zhu · Jinman Zhao · Tongliang Liu · Xiaowen Chu · Cong Wang (Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences) · Bo Han