verification-system

To verify experiment protocols for reproducibility, use a three-agent Executor-Validator-Replicator workflow that enforces provenance and consensus. This system validates units, checks standards compliance, and independently reproduces analytical results to prevent hallucinations.

The three-agent consensus workflow pairs an Executor generating structured outputs with a Validator applying six-category checks and a Replicator independently reproducing outcomes. A Consensus Gate then issues PASS, FAIL, or CONDITIONAL_PASS decisions based on configurable agreement thresholds.

Audit experimental protocols for ASTM standard compliance by applying the Validator component's six-category validation checklist. This checks unit consistency, physical plausibility, mathematical accuracy, and literature comparison before accepting results.

Yes, you can use the three-agent consensus workflow to check DEM parameter calibration. The Validator verifies physical plausibility and mathematical accuracy, while the Replicator independently reproduces the calibration results to ensure they meet agreement thresholds.

When independent reproduction of analytical results fails, the system triggers an escalation path with retry limits. The Consensus Gate issues a FAIL or CONDITIONAL_PASS decision, requiring failure-resolution steps to address discrepancies before accepting the experiment protocol.

During scientific reproducibility verification, the Validator checks six categories: unit consistency, physical plausibility, ASTM standard compliance, mathematical accuracy, cross-consistency, and literature comparison. These checks prevent hallucinations and ensure calculations are verified.