Evals

Ground-truth benchmarks across extraction, covenant detection, and memo generation. Updated each model/prompt revision.

Loading live data…

Synthetic data, portfolio demonstration only.