VetIOS publishes the CIRE methodology as a versioned telemetry and conformance specification. Its signals describe output structure and publication state; clinical reliability still requires outcome-linked and external validation. The reference implementation remains subject to the VetIOS licence.
Differential concentration score
phi_hat = 1 - H(D) / ln(|D|), where H(D) = -sum(p_i * ln(p_i)) over normalized differential probabilities
Higher values mean the differential distribution is more concentrated. Concentration does not establish correctness, calibration, or clinical actionability.
Input quality risk score
input_m_hat = 0.40 * missingness + 0.35 * contradiction_rate + 0.25 * out_of_distribution_rate
Higher values mean the clinical input is incomplete, contradictory, or outside the supported operating envelope.
Rolling runtime signal drift
delta_rolling = EMA(delta_hat, alpha = 0.1)
Tracks short-horizon movement in the runtime signal compared with prior inference behavior.
Runtime signal volatility
sigma_delta = standard_deviation(last 50 delta_hat values)
Higher values indicate less stable runtime signal behavior across recent events.
Runtime perturbation score
cps = 0.40 * (1 - phi_hat / phi0) + 0.35 * max(0, -delta_rolling) / phi0 + 0.25 * sigma_delta / phi0
Composite runtime score used to place an output inside a configured publication envelope. It does not measure diagnostic correctness.
The runtime signals are within the configured publication envelope.
Publish the differential with its runtime status; retain clinician review and outcome validation requirements.
The inference should be interpreted with clinician review because runtime stability has weakened.
Surface uncertainty and recommend additional diagnostic confirmation.
The inference is unstable enough to require explicit caution.
Show the result as decision support only and avoid workflow automation.
The inference is outside the safe publication envelope.
Suppress the clinical answer and require human review or better input.
The specification is public; the managed VetIOS infrastructure remains the commercial product. Partners can implement the runtime signal contract, while production deployments use VetIOS for hosted lineage, outcomes, model cards, network learning, and enterprise/government API access.