evalbench.nsisong.com

Eval Bench

Compare expected behavior, regressions, safety cases, latency, and cost.

GitHub-backed Pages Local-first core Public-safe shell

Product rhythm

  1. Load cases and outputs
  2. Run marker checks
  3. Track latency and cost
  4. Export regression reports
01

Dataset and run comparison

02

Human-readable regression summaries

03

Cost and latency totals beside quality

Workspace notes