Linear Probes Could Reveal Which Persona a Base Model Is Playing, Researchers Suggest
cephaloform · x · 2026-11-19
dodecagonism floats an interpretability question: base models clearly have some "truesight" ability—has anyone used recent interp methods to identify the persona a model is playing when completing text? cephaloform replies it's a fun experiment that doesn't even need fancy interp methods—linear probes for attributes like occupation, ethnicity, or fiction/nonfiction would suffice.
Related event: Researchers Propose Linear Probes to Detect Personas in Base Models(3 posts)→
More from Research
- LEV: Looped decision model lifts 2.6B accuracy 63.2% to 76.2% with adaptive compute — jiqizhixin · 2026-11-19
- Aurora2-400M-A20M: a sparse MoE pretrained from scratch in 36 hours on one RTX Pro 6000 — Tall_Abrocoma_3533 · 2026-11-19
- Researcher calls METR/Anthropic '2x speedup' RSI analysis statistically dubious — jessi_cata · 2026-11-19
- ETH's Anymal robot completes 10km autonomous alpine hike with VLM-guided ROAM system — ChongZzZhang · 2026-11-19
- OpenVax codebase grows from 78K to 565K lines in 15 months — iskander · 2026-11-19
- François Fleuret: decoder transformer's elegance — no trade-off hyperparameters — francoisfleuret · 2026-11-19