A single Qwen3 neuron tracks both "Colorado" and Tornado HTTP client code
livgorton · x · 2026-10-17
An interpretability find: a Qwen3 neuron cited as tracking "Colorado stuff" turned out, when probed with Claude, to also fire strongly on Tornado's Python HTTP client code — a striking polysemantic case the author jokes about connecting.
The retweeter adds a caveat from genomics training: it's dangerously easy to convince yourself you see interpretable patterns when scanning activation lists, and hopes interpretability work avoids that trap.
More from Research
- Researchers propose a Google Scholar-like index for async LLM agents that scales — shangbinfeng · 2026-10-17
- Oscillator Language Models paper: phase-coded Kuramoto dynamics for language modeling — burny_tech · 2026-10-17
- Vals Science Index: OpenAI Models Lead Overall, but No Model Sweeps Every Science Discipline — BenBlaiszik · 2026-10-17
- First-author PhD paper frames AI text detection as continual learning, accepted at NeurIPS 2026 — lucy3_li · 2026-10-17
- Regents Labs curates 17 AI/agent science papers weekly with AI-generated deep reads — seanwbren · 2026-10-17
- A 'pre-numeric' AI math layer trains patterns before numbers exist — debreuil · 2026-10-17