Develop evaluations that measure model performance on real chemistry tasks, such as reaction outcome prediction, synthesis planning, structure elucidation, and compound design, and feed results back to research teams. This research continues many of the directions our team worked on prior to Anthropic, including: GPT-3, Circuit-Based Interpretability, Multimodal Neurons, Scaling Laws, AI & Compute, Concrete Problems in AI Safety, and Learning from Human Preferences.