Curriculum actions
Selects the intensity of hard-example mining and data selection regimes during training.
Public, patent-safe technical highlights
QuantX
Training control that is budgeted, uncertainty-aware, and auditable.
QuantX is a training-time meta-controller designed for high-stakes tabular systems where performance must be balanced with operational constraints and reproducibility requirements. It operates at the level of training policy decisions rather than changing the user-facing business model interface.
Trace concept
state → action → reward → constraint signal → logged decision\n\nPurpose: curriculum / sampling / hard-example mining decisions\nBoundary: no proprietary equations or controller internals disclosed here
Budget and stability limits
Designed to balance performance with time, memory, and reproducibility constraints.
Decision traces
Logs training decisions so they can be archived alongside model artifacts for forensic review.
Intentionally non-enabling public page
This page avoids disclosure of proprietary internal state representations, loss formulations, controller update equations, or implementation details. It is intended for public orientation only.