V5: Neuron-level steering
Activation is injected directly at chosen neurons or MLP coordinates in fixed layers, rather than along a direction learned from prompt pairs.
Original source: Templeton et al. 2024, Scaling Monosemanticity (Golden Gate Claude) ↗
Repos
2
Vectors
7
The only public neuron-steering experiments in the catalog
These are the only open-weight analogs of Anthropic's Golden Gate Claude experiment: activation pushed at chosen neurons instead of along a direction learned from prompt pairs.
dahara1/EvilQwen-control-vectors
Source cited: neuron-explorer (Lyra, lyramakesmusic) activation steering; Golden Gate Claude-style neuron steering
6 vectors 6 sets · 92.1 KB
308 repo downloads
308 repo downloads
alesha-pro/Qwen3.8-27B-Persona-v2-GGUF
Source cited: Neuron-level activation steering (cf. Anthropic Golden Gate Claude); custom export_control_vector.py for llama.cpp
1 vector 1 set · 40.6 KB
97 repo downloads
97 repo downloads