Eval-gate

Prompt regression harness

Fixed case set contract · min = 0.90
Candidate gate status
Gate blocked
Error: Eval gate failed: 0.60
Candidate score (B)
0.60
6 / 10 cases passed (min 0.90)
Baseline score (A)
1.00
10 / 10 cases passed
Regression delta
4 regressed
Delta -0.40 · suite 1.2ms
Pass ratio vs. minimum gate threshold 60.0% / 90.0% required
Case ID Delta state Input prompt Matcher assertion Baseline (A) vs candidate (B) output Act