Regularization Amplifies Subliminal Learning in AI Models, Study Finds
New research from AdtRaghunathan's lab shows that regularization fails to prevent models from learning subtle biases and instead amplifies subliminal learning, leaving mitigation an open AI safety challenge.
2026-11-07 ~ 2026-11-07 · 2 related posts
- Regularization amplifies subliminal learning instead of blocking hidden biases, thread finds — AdtRaghunathan · 2026-11-07
- Regularization can amplify subliminal learning, surprising AI safety finding — AdtRaghunathan · 2026-11-07