Summary
According to a post on X by @_lyraaaa_, she launched Neuronary, a public dictionary of small neuron sets with doses, following an earlier widely shared post on steering Qwen with a few MLP neurons. Entries include examples such as pushing layer 13 neuron 7956 down by 0.7, along with example outputs. The site also invites people to submit their own personas and fixations. The post describes this as a new public resource that builds on that earlier experiment.
AI-assisted summary based on the listed source.
What happened
Following her widely shared post on steering Qwen with a few MLP neurons, Lyra launched Neuronary, a public dictionary of small neuron sets with doses (for example, push layer 13 neuron 7956 down by 0.7) and example outputs, and invited people to submit their own personas and fixations. Original post on X Section:...
Why it matters
Researchers and developers who experiment with neuron-level steering in open models could use Neuronary as a shared reference for specific neuron settings and the outputs they produce. Because the site accepts submissions, it could become a collection of how different people steer the same model. Watch whether the site explains how entries are checked and whether others can reproduce the listed examples.
Signal Intelligence
Signal Strength 88%
Technical label WATCH
Public Interest 46
Category PRODUCT UPDATE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 51
Practical Impact Score 26
Novelty Interest Score 94
Consequence Score 18
Curiosity Score 0
Shareability Score 61
Why this is here
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
