Runs & baselines
A run is one execution of a suite. A baseline is the run you measure it against. Almost everything in Omic is one of those two things, or a comparison between them.
What a run is
Every run captures the same three things:
•
Inputs — the cases, the prompts, and the model settings used.
•
Outputs — what the system actually produced, per case.
•
Scores — what each scorer returned, and the aggregate.
Choosing a baseline
A baseline is just another run, picked one of two ways:
•
By branch — compare against the latest run on main.
•
By commit — pin the comparison to an exact revision.
How diffing works
Comparison is per case, never a single blended number:
•
Improved — cases that scored higher than the baseline.
•
Regressed — cases that dropped — these are what gate a merge.
•
Unchanged — identical output, collapsed by default.