Initial window
Higher variance while the model observes environment and enrollment context.
Adaptive modeling
Running
Per-user adaptation with enrollment boundaries — same principles as live verification.
Signal fusion
Learning
Combines scorer outputs; optional refinement from labeled batches.
Calibration
Stabilizing
Threshold convergence toward stable per-user posture.
Anti-spoof posture
Active
Replay-aware sessions and explicit reject paths for synthetic or replay-like traffic.
Calibration trajectory
Calibration timeline
Consolidation
Scores and thresholds move toward a repeatable decision band per subject.
Stable band
Reduced spread on trusted attempts — ready to compare against production SLOs.
Operational metrics
Synthetic / spoof flags
0
Items flagged by fusion and anti-spoof heuristics during the scripted run.
Replay handling
0%
Indicates whether duplicate-frame paths were blocked under session rules.
Quality index
0%
Weighted accept and partial-credit mix from the corpus (Lab-only view).
Genuine-user relief
0%
Change in reject rate from early to late segments on labeled genuine rows.
Workload summary
Subjects
0
Active calibration
0
Plateaued
0
Labeled non-genuine rows
0
Summary notes
Scenario trends
Validator reference (JSON)
{}