#sparsity

The latest AI and developer news about #sparsity, with the original source and a short summary.

Goodfire Research Finds Neurons Are an Interpretable Basis for LLMs

Analysis of Qwen3.6-35B-A3B shows that ablating 92.2% of weak neuron activations causes minimal performance loss, proving neurons are tractable units of analysis.

·1·2026.11.10 09:00

AFM3 20B, 20% MLP Activation

AFM3 20B is a MoE model that selects only about 20% of MLP layers to activate per prompt.

·3·2026.08.04 10:00