Researchers Propose Linear Probes to Detect Personas in Base Models
A researcher suggested probing base models' ability to infer author traits during text completion; a commenter noted that simple linear probes, without complex interpretability tools, could identify the persona a model is embodying.
2026-11-19 ~ 2026-11-19 · 3 related posts
- Base models show 'truesight' ability — could linear probes identify the persona they play? — dodecagonism · 2026-11-19
- Linear Probes Could Reveal Which Persona a Base Model Is Playing, Researchers Suggest — cephaloform · 2026-11-19
- Simple linear probes could reveal the persona base models play, argues interp enthusiast — cephaloform · 2026-11-19