Persona neuron ablations worsen most personas, with altered-assistant gaps as surprise holdout
maxsloef · x · 2026-10-30
- maxsloef replies to adamkarvonen: ablating persona neurons across other trained personas broadly performs a bit worse, with the biggest gaps on the altered-assistant personas—unexpected since those were the holdout.
- The author cautions the datasets and evals for these personas got far less attention than Hoot's; Claude is now checking activation patterns for the other personas, with similar results expected, though more off-distribution personas may not respond well to response-side ablations.
More from Research
- Dev ditches unanimous-vote stopping rule for a judge in multi-AI debate project — lluisantoni · 2026-10-30
- Google's AI Search Uses Your Knowledge Graph Facts Without Linking, RESONEO Research Finds — dejanseo · 2026-10-30
- Raon-1-mini-preview runs natively on hipfire but 200x slower prefill than llama.cpp, patch released — Queasy_Asparagus69 · 2026-10-30
- TEM: Transformer Energy-based Models scale beyond diffusion without guidance — du_yilun · 2026-10-30
- NVIDIA's AgentBaton cuts agent API costs 65.9% with a single model handoff — dair_ai · 2026-10-30
- Stanford releases RT-J, a System 1 model predicting structured data in one forward pass — jure · 2026-10-30