AI
A Goodfire researcher argues neural networks are far more interpretable than believed: roughly 90% of neuron poles in Qwen3-32B have one clear job. The old case for superposition, he says, was misread.
Loading…
Search
No phones and 1 story for “neural networks”
AI
A Goodfire researcher argues neural networks are far more interpretable than believed: roughly 90% of neuron poles in Qwen3-32B have one clear job. The old case for superposition, he says, was misread.