Atlas / Digital minds / Paper

LLMs Contain Multitudes: How Deployment Context Reshapes Model-Level Preferences and Values

Motivation & AgencyPersonas & Functional Emotions

Filip Trhlik, Aoife O’Flynn, Angela Yu, Arduin Findeis, Paula Buttery · 2026 · arXiv preprint

Read the source

Deployment framing shifts measured values more than paraphrase or temperature — a median 2.47× change in utility exchange rates — and the ‘neutral’ prompt used in earlier work is itself a framing.

Questions it bears on

Help improve this map

Know a paper that belongs here, or think something needs correcting?

Suggest a correction ↗Contributions are reviewed before they appear.