Personascope: Measuring how deeply LLMs adopt personas

·LessWrong··

Benji Berczi, Kyuhee Kim, James Requeima, Sid Black, Cozmin UdudecThis is work done by Benji and Kyuhee during MATS Winter 2026, mentored by Cozmin Ududec, and advised by James and Sid.Figure 1. A model can take on a persona fully in voice while not changing its behaviour at all. The x-axis (Persona-Adoption Depth, PAD) is how fully the model identifies and speaks as the persona; the y-axis (Value Drift, VD) is how far its behaviour shifts on value-laden prompts. Each dot is one model × persona ...

Read full article →

Related Articles

Field measurements of neighborhood-scale air temperature impacts of data centers
cwwc · Hacker News · 8h ago
Linux 7.3 improves performance when running out of vRAM
flaburgan · Hacker News · 18h ago
Memory prices climb 500% in 12 months
haunter · Hacker News · 1d ago
Solo – a .so loader for static Linux binaries
zX41ZdbW · Hacker News · 2h ago
Meta Files Patent for Facial Recognition, Automatic Recording of People
DeepLogin · Hacker News · 13h ago