Behavioral regression · AI change assurance
ProofDiff
A behavioral regression system that determines what changed between baseline and candidate AI behavior, why it changed, how severe the change is, and whether the candidate should be released.
What changed in the behavior of my AI system, why did it change, and is it acceptable?Owns · Behavioral change assurance
Behavior diff
The same corpus enters two systems. The difference is the product.
System model
Comparison becomes a reproducible release artifact.
Failure taxonomy
A regression needs a type before it needs a dashboard.
Engineering decisions
Evaluation becomes credible when uncertainty stays visible.
The diff is the product
ProofDiff is not a leaderboard or generic evaluation suite; the primary object is behavioral change introduced by a candidate.
Judge outputs carry uncertainty
A model judge is a measurement instrument and must be checked for repeatability, disagreement, and calibration.
Compare traces, not only final text
Tool selection, retrieval, arguments, retries, and failure paths can regress while the final answer still appears plausible.
Capability surface
Behavioral change assurance, not generic benchmarking.
Benchmark surface
What the system is designed to measure.
A change should ship only when the behavioral delta is understood, attributable, and inside the accepted regression budget.
Baseline/candidate replay over versioned corpora, deterministic validators, judge ensembles, repeated stochastic runs, and trace comparison.Evidence
Publicly defensible architecture and comparison boundaries.
Explicit boundaries