Code-demo / benchmark_models.py

Commit History

  1. changes

    2931ad6
  2. Add real benchmark results from 3 frontier models, populate findings paper Section 4

    09f4b98
  3. Broaden keyword synonyms, rewrite benchmark to 3 models live-only, strip fabricated data from paper

    c91e6eb
  4. Add semantic 'Why' metric, redesigned hard task, multi-model benchmarker, and findings paper

    7f5e6a9