Formal clearance record for the operating-characteristic (false-REFUTE) calibration run of SIP-PREREG-03, a pre-registered Bayesian equivalence study of LLM factual-commitment retention on the retrieval axis (OSF.IO/Q45RC). The calibration run establishes a false-REFUTE operating characteristic of 0.046 (Wilson 95% CI 0.0373, 0.0567, half-width 0.0097) on 1,803 converged replicates (1.21% refusal), bound to golden configuration 91af3858 — consistent with the registered prediction of ≈0.05. The test apparatus was cleared through a two-node adversarial pass on independent hardware: PIN-B refused-path count-identity and resume-equivalence under a mid-replicate crash were each verified rather than assumed. The record documents the cleared apparatus (four hash-pinned modules and the sealed golden config), the adversarial clearance ledger, three documentation conditions, and a stratified refusal report. This is instrument onus discharged — the operating characteristic of the test — not a substantive empirical claim, which is retired only by the registered confirmatory run on real data plus independent reproduction. Verification provenance is reported as method, not evidence (AI-Neutrality Principle): the load-bearing result is reproducible bit-for-bit by re-running the adjudication on the registered data, without trusting the multi-node pipeline.
Smith et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: