"Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders

Excerpt

arXiv:2608.07852v2 Announce Type: replace Abstract: How a language model internally represents who is speaking, the Assistant, an assigned roleplay persona, or a narrated story character, remains underexplored. We study speaker representations using a dataset of user-expressed emotional text and corresponding model responses. We decompose three generation settings (Assistant, Roleplay, and Story) into sparse autoencoder features extracted at turn-boundary and pronoun-token positions and selected