Bounded primary path
The trusted continuity estimator remains capped and nonrecursive rather than turning successful rescue into future primary state.
Current research frontier · Phase 11 · P14R
Independent rescue + robust uncertainty transfer.
A frozen simulation study with high protected availability and coverage, plus one lateral tail-efficiency failure that remains visible.
P14R passed every required seen-transfer gate. Protected validation reached 98.53% useful availability, 96.17% lateral 95% coverage, 95.82% altitude 95% coverage, and 94.63% rescue recovery. The study closed without final confirmation because one preregistered protected component failed.
Protected availability98.53%
Lateral 95% coverage96.17%
Altitude 95% coverage95.82%
Rescue recovery94.63%
What changed
Three bounded changes define P14R.
The primary estimator remains bounded. Rescue stays independent. Uncertainty transfer becomes robust across grouped compound shift.
Independent rescue
Primary-unavailable truth-visible observations can be recovered without allowing the rescue path to silently become the new anchor.
Robust uncertainty envelope
Groupwise conformal uncertainty is calibrated for compound shift instead of depending on one fragile severity partition.
Fresh seen transfer
Availability recovered without giving up coverage.
Every required seen-transfer family passed before protected validation. This supports robustness under the frozen protocol; it is not a safety verdict.
Useful availability97.72%
Lateral 95% coverage94.76%
Altitude 95% coverage94.79%
Calibration MACE0.00512
Lateral p95 width / p95 error2.169×
Rescue coverage95.06 / 95.01%
Protected boundary
One locked lateral-tail component stopped final confirmation.
Protected validation passed every required family except the preregistered H4 lateral p95 interval-width / p95-error limit. The threshold was not loosened, and the final P15 holdout remained unexposed.
2.435×Lateral p95 interval width / p95 errorFrozen maximum: 2.25×.
P15-v2 seed 869869: not exposed.
P15-v2 seed 869869: not exposed.
Provenance
Frozen evidence, not a rewritten success story.
The study is closed under the Phase 11 protocol. Future tuning requires a new preregistered phase with fresh development, transfer, and protected evidence.
Scientific head
58b06089a621264afb886f6aee2acaacf8a8709cCandidate SHA-256
bba1b7be7565b470af80db502e6fa5d969bc86c712e322557c06b30e49bd9c86Candidate artifact#9271909463
Claim boundary
Strongest so far does not mean safety accepted.
Phase 11 supports a narrower claim: bounded continuity plus independent rescue achieved high availability and robust coverage across fresh transfer and protected validation, while protected shift still exposed excessive lateral tail width.
simulation_only=true · safety_acceptance=false · controller_tuning_allowed=false