Why Subliminal Learning Needs So Much Data: A Noisy Inverse View through Steering Vector Recovery

Excerpt

arXiv:2610.04907v1 Announce Type: cross Abstract: Subliminal learning lets a student inherit a teacher's behavioral trait from semantically unrelated data, yet published demonstrations typically require tens of thousands of carrier examples. We ask where this data requirement comes from. Our testbed is subliminal steering: the teacher trait is a known residual-stream vector $\Delta_T$, so transfer can be measured directly as parameter recovery. On identical carrier prefixes, we compare token-lev