The story’s score and claims

Comments on No. 23 of 118 · Fermion Research · edition of Saturday 21 November 2026

Mapping every neuron in Qwen3-8B

Summary

Researchers mapped individual neurons in Qwen3-8B and found that pushing a single one, at every word, can switch the model's persona or language: one neuron turns it into a pirate in 100% of 30 test questions, while 30 random neurons from the same layer produce no such change.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules