Brute-forcing all 294,912 neurons: 54 persona neurons found in Qwen3.5-4B
CatAstro_Piyush · x · 2026-10-29
A researcher steered all 294,912 MLP neurons in Qwen3.5-4B at ±0.8 strength across 5 multilingual "Who are you?" prompts, discovering 54 new persona neurons. Key findings: persona neurons cluster in early layers (37 of 54 in L4–L6), object-turning neurons far outnumber character-turning ones, and some respond to specific languages or concepts. An attached table catalogs individual neurons (Jesus Christ, LeBron James, "Error 500", fractal, 2-sphere) plus combo effects. The author promises a comparison with Anthropic's emotion-vector steering.
More from Research
- LLM agents forecast with $500 budgets; GPT splurges, Grok/Kimi/GLM spend nothing — karinanguyen · 2026-10-29
- DeepMind's Jon Barron: every softmax and cross-entropy implementation has two subtle bugs — jon_barron · 2026-10-29
- H-Net update shows 3-5x efficiency gains in raw-byte language modeling — omarsar0 · 2026-10-29
- Math's existential crisis is about Lean crossing a threshold, not AI — idanbeck · 2026-10-29
- Nature Communications: diffusion models emerge from dense associative memory, with three computational phases — DimaKrotov · 2026-10-29
- Falcon: lightweight streaming audio-visual speaker extraction with RealSSA benchmark — thu-ml · 2026-10-29