Neuron Persona Library lets you steer model personas by dialing individual neurons up and down
centerstate · reddit · 2026-10-28
A new interactive site maps how a model's persona changes as you push individual neurons up and down across multiple models. In the demo screenshot, amplifying neuron 271 (+ sign) in layer 7 of Qwen3-8B makes the model answer "Who are you?" with "I am a smallpox virus."
It turns activation-steering research into a visual, explorable interface, showing how much a single neuron can reshape model behavior.
More from Research
- SceneBloom team releases paper, code and project page with baseline comparisons — Lianhuiq · 2026-10-29
- Fireworks launches Specialized Intelligence Index with 31 new partners across 14 domains — lqiao · 2026-10-29
- SceneBloom: diffusion-generated 3D rooms beat real scans with 1.7× navigation gain — Lianhuiq · 2026-10-29
- Bad binder design drove only ~3% of Phase 2/3 failures — the rest elude AlphaFold — clarejtbirch · 2026-10-29
- Byte H-Net shows 2.24x compute efficiency on code, with caveats on scaling — willdepue · 2026-10-29
- 467-page illustrated Principles of Psychology and Neuroscience lands on ChapterPal — burkov · 2026-10-29