Steerable neurons: many steer output, few pin down a consistent persona
Sauers_ · x · 2026-10-19
An interpretability exchange: muddletoes frames each steerable neuron as an attractor into which the token output stream falls, asking what fraction of neurons produce a discernible steering effect and whether there are side effects on superficially unrelated behaviors. Sauers replies that many neurons show a discernible steering effect, far fewer pin down a particular character, and side effects depend on the specific neuron.
Related event: Single-neuron steering gives Qwen3-8B an "inner monologue"(7 posts)→
More from Research
- Meta paper: agentic meta-reasoning hits 71.5% vs Codex 58.0% on ProgramBench — CatAstro_Piyush · 2026-10-19
- Dialectic 60: Arc Institute's Patrick Hsu on biology, AI, and funding human-response datasets — MatthewMcAteer0 · 2026-10-19
- NTU's RoboFoundry evolves the robot software stack itself, lifting frozen GPT-5.5 from 56.9% to 72.7% — jiqizhixin · 2026-10-19
- Interp debate: Nick Cammarata says most neurons are explainable; leogao pushes back — burny_tech · 2026-10-19
- Baidu's Virtual Head Attention cuts KV cache 4-7x with zero custom kernels — burny_tech · 2026-10-19
- Classic DeepMind paper: the prefrontal cortex works as a meta-reinforcement learning system — burny_tech · 2026-10-19