· · AI Alignment Forum
Model Diffing Agents: Transforming AI Behavior Analysis
This is the second in a series of research updates from the Google DeepMind Language Model Interpretability team, in interpretability and adjacent areas. The first post can be found here. TL;DR It is possible to build extremely simple agents that reliably find interesting behavioural differences bet
#model diffing agents#ai behavior analysis#interpretability#deepmind#india ai ecosystem