Curriculum & sampling
Research on how training emphasizes examples and adjusts data-selection choices.
RESEARCH TECHNOLOGY
QuantX
Training decisions,
with a record of why.
Research on curriculum and sampling decisions under compute budgets and reproducibility requirements.
Training stateSampling decisionFeedback & recorded outcomeConceptual overview
THE FOCUS
Training is a sequence
of choices.
What to emphasize. How to sample.
How much compute to spend.
QuantX investigates training-time control for tabular machine learning. Its focus is the selection of curriculum and sampling actions, with feedback and decision traces recorded alongside the model workflow.
This is early-stage research technology. Contact us to discuss its scope, evaluation requirements, and potential fit for a research collaboration.
TECHNICAL OVERVIEW
Compute constraints
Consider time, memory, and training stability alongside model performance.
Decision traces
Record training decisions so they can be inspected with the resulting model artifacts.
PART OF THE RESEARCH
Explore the wider work.
VALTRIX brings our scientific-screening work together. Its public JARVIS example provides a concrete result and an inference package to inspect.
View VALTRIX evidenceLET’S WORK ON A REAL QUESTION
Bring the research.
Let’s discuss the next step.
For scientific R&D teams, research collaborators, and technical partners exploring materials or molecular screening.