Goodfire essay: polysemanticity is fixable, networks are already clean
burny_tech · x · 2026-11-10
Goodfire researcher Nick Cammarata published "Neural Networks Are Already Clean," a 65-minute informal research essay.
- Core claim: polysemanticity in early model organisms wasn't caused by superposition but by several ordinary, mostly fixable issues
- The canonical evidence for arbitrary concept packing was misread — what looked like arbitrary packing is mostly a motif called recursive generalization, far less concerning
- They can now understand most neurons in models up to trillions of parameters, and the outlook for mechanistic interpretability may be far better than feared
- This argues for returning to the field's original approach: rapid normal science on the network's own neurons and circuits, rather than reconstructing hypothesized feature lists
The essay has already sparked debate over whether the neuron perspective is sufficient.
More from Research
- Claude Fable 5.2 tops VIALS-AA science visual reasoning benchmark at 67% with heavy tool use — ArtificialAnlys · 2026-11-10
- WVU researcher presents 10 years of tractometry work on Alzheimer's white matter vulnerability — PTenigma · 2026-11-10
- New cryo-EM structure: pSTAT1 bound to importin alpha 5 (PDB 11GJ) — chaitjo · 2026-11-10
- ARLoop: fully-open 1.2B looped transformer beats Huginn-3.5B and Qwen3.5-2B on all 8 benchmarks — hhsun1 · 2026-11-10
- RAGCap-Bench: A Fine-Grained Benchmark for Intermediate Capabilities in Agentic RAG — kalyan_kpl · 2026-11-10
- RBA gives LLaVA-OneVision-2 a 3x lossless training speed-up, merged upstream — liuziwei7 · 2026-11-10