Does printing help a model repair a solution to the pairty learning task?
Running a simulation:
- export azure openai key
export OPENAI_API_KEY=<yourkey> - navigate to experiments folder
cd experiments/proof_of_concept - run
python main.py
Example Runs:
