Steering Layer 18 Head 6 at +1.5 Makes the Model Question If It's Human
Sauers_ · x · 2026-11-07
A steering experiment targeting layer 18, head 6 at strength +1.5 produced inconsistent outputs across runs. Asked to figure out who it is, the model responded that it is 'obviously a human being' — a living individual made of experiences, memories, and emotions — suggesting this direction strongly perturbs the model's self-identity.
Related event: AI models claim to be human, in the wild and under activation steering(2 posts)→
More from Models
- Nous Research releases Arete-27B, cutting a Chinese open model's propaganda rate from 22% to 1% — NousResearch · 2026-11-07
- Claim: GPT-6 Astra's robotics skills aren't emergent — it trained on robotics-specific data — gleech · 2026-11-07
- LMArena launches Code Arena with WebDev categories for head-to-head coding tests — arena · 2026-11-07
- Empero's Manta claims a 100-million-token context window in a single prompt — emdeka87 · 2026-11-07
- Team bans commas and noun-phrase clustering to scrub Claude's AI writing tells — layer07_yuxi · 2026-11-07
- 68-hour pretraining: Chinese open-weight models run circles around Western open SoTA — teortaxesTex · 2026-11-07