Goodfire presents mechanistic shaping to steer models through their internal representations
AIGoodfire presents mechanistic shaping, a training framework that penalizes a model's reliance on or expression of specific internal concepts. The company reports that the method reduced shortcut learning in chest X-ray models, cut DPO-induced sycophancy while keeping capability gains, and curbed reward hacking in Countdown training.
AIWhy it matters.
The post explains how penalizing a model's internal concept representations during training can reduce shortcuts and sycophancy while keeping those concepts monitorable afterward.
Source details
Goodfire Research · goodfire.comPublished · added here
Original title: Shaping and Monitoring Models Through Their Internal Representations