Adding Gaussian noise to teacher and student weights amplifies subliminal bias transfer, and steered but not prompted subliminal data leaves readable gradient footprints.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Subliminal Learning is Non-Semantic Distillation
Adding Gaussian noise to teacher and student weights amplifies subliminal bias transfer, and steered but not prompted subliminal data leaves readable gradient footprints.