There needs to be some way of automatically assessing performance on the task, though this could be with a Python function or another LLM as a judge (or a combi

There needs to be some way of automatically assessing performance on the task, though this could be with a Python function or another LLM as a judge (or a combination!)