neuronary
A public dictionary of neuron steering recipes for language models: small sets of MLP neurons, each with a dose,
plus example outputs. A recipe like L13N7956-0.7 L14N8949+0.7 means push layer 13 neuron 7956 down by
0.7 and layer 14 neuron 8949 up by 0.7.
The site needs JavaScript. Without it: every recipe as JSON at
the recipes API,
a guide for agents at llms.txt, and the source at
github.com/lyramakesmusic/neuronary.