Measuring What Matters: Construct Validity in Large Language Model Benchmarks

1 points| Cynddl | 3 months ago |arxiv.org