Intervening on a candidate deception neuron reportedly changes Qwen 3 32B's behavior
A user says their hobby interpretability investigation identified some candidate neurons for deception in Qwen 3 32B.
TLDR
A user describes a hobby interpretability investigation of Qwen 3 32B that turned up some candidate neurons for deception. They report that intervening on a neuron changes the model's behavior.
Combined views
109
1 Source, first seen ago
