Comment by netcan

Any interesting examples of methods to make LLMs more "programmable?"