Not All Is Lost: Repairing Lossy User Preference States of Personalization Encoders

Excerpt

arXiv:2610.01270v1 Announce Type: new Abstract: Personalization encoders compress evolving interaction histories into preference states used to rank items or condition text generation. A task head operating only on this state can miss useful evidence that remains in the frozen encoder's cached representations for individual timesteps. We study this recoverability gap and propose REPAIR, which compares cached representations with the current preference state in a compact learned coordinate space.