← Search

Jordan Taylor

2 accepted papers

2026

Obfuscated Activations Bypass LLM Latent-Space Defenses

ICLR 2026poster

_Latent-space_ monitoring techniques have shown promise as defenses against LLM attacks. These defenses act as scanners to detect harmful activations before they lead to undesirable actions. This prompts the question: can models execute harmful behavior _via inconspicuous latent states_? Here, we st…

Cited by 0SourcecodeScholar
2024

Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning

NeurIPS 2024poster

Identifying the features learned by neural networks is a core challenge in mechanistic interpretability. Sparse autoencoders (SAEs), which learn a sparse, overcomplete dictionary that reconstructs a network's internal activations, have been used to identify these features. However, SAEs may learn mo…