Skip to content
Goodfire Research·· 1 d agoAI score38

Goodfire Research discovers neuron communities in language models

Discovering Neuron Communities in Language Models

AI summary

Goodfire Research has identified "neuron communities" — small groups, often fewer than ten neurons — that implement common model behaviors such as grammatical agreement, copying and arithmetic. Using gradient-based attribution across millions of prompts, graph construction and community detection, the unsupervised method was run on Llama-3.1-8B and replicated on Qwen 3.6 35B-A3B. Ablating a community changes predictions, while keeping it largely preserves them even after removing other MLP neurons in those layers.

Source:Goodfire Research · goodfire.com