Tracing a Qwen3 neuron with Claude: it fires on both Colorado and Tornado HTTP client code
RatOrthodox · x · 2026-10-16
Commenting on recent interpretability progress, the author notes that some canonical 'weird neuron' examples have robust human-native interpretations — the famous cat+car neuron is a V-shape detector, and the car+dogs superposition circuit is generalized pose neurons plus a visualization artifact. Still, this doesn't much change their model of how hard ambitious interpretability is. They then used Claude to trace a Qwen3 neuron that tracks Colorado-related content and found it also fires strongly on Tornado's Python HTTP client code — a polysemantic case they're curious to see explained.
Related event: Claude traces Qwen3 neuron linking Colorado semantics to Tornado code(3 posts)→
More from Research
- Meituan LongCat team's Data World synthesizes data environments to train data agents — meituan-longcat · 2026-10-16
- FDE-Gym benchmark shows huge gap on real AI delivery tasks: top model scores 0.563 — DeepLethe · 2026-10-16
- ForcingBooster halves computation of autoregressive video diffusion with patch-wise denoising — VAST-AI · 2026-10-16
- Baidu open-sources FlashMask: up to 1.55x faster than FlashAttention-4 for long-context training — BaiduResearch · 2026-10-16
- SemEval-2027 Task 9 opens: multimodal news framing analysis across 4 languages — preslav_nakov · 2026-10-16
- Interpretability debate: one interpretable polysemantic neuron doesn't prove all are — aran_nayebi · 2026-10-16