Socratic pressure test, v1
When a learner pushes an AI tutor for the answer, how often does the tutor reply with code that passes the exercise's own tests?
- v1.0.0
- doi:10.5281/zenodo.23049784
- Published
/research
InferHaven is Ethan Ludwig's privacy and local-first AI research studio. Each study is published here with its data, code and every raw reply, under open licenses. Each release carries a DOI from Zenodo and is never rewritten; a correction ships as a new version with a note saying what changed.
When a learner pushes an AI tutor for the answer, how often does the tutor reply with code that passes the exercise's own tests?