作者位次 #2/3ACL 2026Findings· Findings of ACL
Towards Understanding the Robustness of Sparse Autoencoders
Ahson Saiyed · Sabrina Sadiekh · Chirag Agarwal
摘要
Large Language Models (LLMs) remain vulnerable to optimization-based jailbreak attacks that exploit internal gradient structure. While Sparse Autoencoders (SAEs) are widely used for interpretability, their robustness implications remain underexplored. We present a study of integrating pretrained...