Transluce
Transluce Monitor
Impressive examples that fix LLM reasoning errors
AI Linter, AI debugger
Neuron clustering → extract spurious concept if irrelevant neuron contribute a lot to the activation or neuron description does not match with output