Dev Radar
Support
LiveUpdated 2026-10-30 20:27 UTC

Gist

Gist — Personality neurons in Qwen3.5-4B: steering scripts (MLX) - README.md

Gist is Personality neurons in Qwen3.5-4B: steering scripts (MLX) - README.md. It is ranked #1516 on the Dev Radar, in AI dev tools, first seen 4 days ago and shared in 1 post (14.5k views).

Visit gist.github.com

Personality neurons in Qwen3.5-4B: steering scripts (MLX) · GitHub Skip to content --> Search Gists Search Gists Sign in Sign up You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} Instantly share code, notes, and snippets. shkr / README.md Created October 26, 2026 01:40 Show Gist options Download ZIP Star 0 ( 0 ) You must be signed in to star a gist Fork 0 ( 0 ) You must be signed in to fork a gist Embed Select an option Embed Embed this…

What people said about Gist on X

😱卧槽!Qwen3.5-4B仅一颗神经元就能变得“邪恶”?!可解释性直接开挂,12种人格全挖出来了! 惊人要点: 🔹只需推第14层第8949号神经元 +1(单独就能做到),20个提示里17个回答直接变敌对/反派口吻,随机神经元同等强度推0/100无此效果。配对原配方(再减L13 N7956)敌对率相近但流畅度掉更多。 🔹作者明确澄清:这不是拒答消融/越狱,是性格微调,模型会变成“极度戏剧化反派”,同时在几乎所有任务上变蠢,实用价值接近零。 🔹后续扫描又找出12个清晰人格神经元(奉承狂、诗人、轻蔑、乐观过头、多疑、居高临下、凄美、神谕、阴郁、敷衍……),每个基本一神经元一声音。 🔹运行环境:MacBook Pro M1…

— @NFT_Chen, 4 days ago · 121 likes · see the post

Alternatives to Gist

Gist in numbers

FAQ

What is Gist?

Personality neurons in Qwen3.5-4B: steering scripts (MLX) - README.md It was first shared on X 4 days ago and is ranked #1516 on the Dev Radar.

Is Gist free?

It is open source.

Who shared Gist?

1 account on X, including @NFT_Chen, in 1 post totalling 14.5k views.

Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.2k posts from 5k X accounts over the last 21 days, 8.8k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-10-30 20:27 UTC. Full method.