Steering a single MoE expert at strength +1.25 rewires Qwen3-30B's persona
Sauers_ · x · 2026-11-06
Sauers demonstrates expert-level steering on Qwen3-30B-A3B: locking Expert 25 on at every token in layer 12 and amplifying its hidden-state contribution (+1.25) flips the model from its default "I am Qwen, developed by Alibaba" identity into a coherent new persona (e.g. a 23-year-old engineer at a San Francisco startup). A vivid demo that a single MoE expert direction is enough to rewrite self-identity.
Related event: Activation Steering of Single MoE Experts Transforms Qwen3's Personality(4 posts)→
More from Research
- UC Berkeley's Ctrl-F adds force commands so humanoid robots can move 20 kg loads — ChongZzZhang · 2026-11-06
- Archon Robotics debuts Anchor 1, a humanoid whole-body controller cutting tracking error 30-80% — ChongZzZhang · 2026-11-06
- Steering one attention head flips Qwen3-30B from 'I am an LLM' to 'I am a real person' — Sauers_ · 2026-11-06
- Stanford 2.5-hour lecture explains end-to-end how ChatGPT and Claude are built — MaxLi_Columbia · 2026-11-06
- Formal proofs of tens of millions of lines of code are now feasible, says Anthropic engineer — morqon · 2026-11-06
- Hugging Face releases FinePhrase slides: 486B-token synthetic pretraining dataset with 90 rewrite styles — NielsRogge · 2026-11-06