Platt calibration of the memory-note Likelihood, each tied to the exact model it was fitted on (see
FindDecisionCalibration in @memberjunction/ai-core-plus): the MJ decision model that answered
(ModelName, from modelInfo.modelName) and the model its driver reports behind it
(ResolvedModel, from DecisionResult.ResolvedModel). Fitted on the memory-note gate measurement:
Jev at its pinned APIName, typesafe/jev-1.13-20260917, which it reports back as the
resolved model.
Only models that beat the proxy self-report scorer (see MEMORY_NOTE_MIN_PROBABILITY) are
listed. LLM Decision was measured and left out: at any threshold it kept more wrong and speculative
notes than the proxy's confidence >= 80 (at a calibrated 0.6, 11.5% of wrong notes). A batch answered by any other model, including Jev at another version,
falls back to the self-reported confidence rule, as a failed decision does. Refit, and add the new
pair, whenever a model, its version or the question changes.
Platt calibration of the memory-note Likelihood, each tied to the exact model it was fitted on (see
FindDecisionCalibrationin@memberjunction/ai-core-plus): the MJ decision model that answered (ModelName, frommodelInfo.modelName) and the model its driver reports behind it (ResolvedModel, fromDecisionResult.ResolvedModel). Fitted on the memory-note gate measurement:APIName,typesafe/jev-1.13-20260917, which it reports back as the resolved model.Only models that beat the proxy self-report scorer (see MEMORY_NOTE_MIN_PROBABILITY) are listed. LLM Decision was measured and left out: at any threshold it kept more wrong and speculative notes than the proxy's
confidence >= 80(at a calibrated 0.6, 11.5% of wrong notes). A batch answered by any other model, including Jev at another version, falls back to the self-reported confidence rule, as a failed decision does. Refit, and add the new pair, whenever a model, its version or the question changes.