This piece introduces Unvibe, a Python library that uses unit tests as a reward function to guide Large Language Models in generating correct code. The author, Claudio Santini, explains that this approach works well for generating complex code within large, existing codebases, provided they are well-tested.
https://claudio.uk/posts/unvibe-a-python-test-runner-that-generates-correct-implementations.html
Post #200
1.01K