Activation Probes Are Reliable With a Handful of Positive Examples

arXiv:2610.39766 · cs.LG · Submitted 2026-10-29 · Read on arXiv

cs.LG

Submitted: 2026-10-29

Updated: 2026-10-29

Comments: Presented at the Mechanistic Interpretability workshop at NeurIPS 2025

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Related papers