Experimental Iteration
Planning Experiments
The AI Scientist uses the #Aider coding assistant to plan and execute the experiments as outlined in the idea generation phase Error Handling
The process is made robust by catching errors, fixing code, and reattempting up to four times if the experiment fails
Result Documentation
After completing each experiment, results are documented in an experimental journal style
This documentation can include both text and visualizations
Re-Planning
Based on the experimental results, the AI Scientist re-plans and executes subsequent experiments, iterating up to five times
Visualization
Aider edits the plotting script to create figures for the paper, and notes are made on what each plot contains, ensuring all information needed for the paper write-up is documented