Signal, structure, and language
Radio-frequency records paired with grounded text, machine-readable labels, realized values, and generation provenance.
Roadmap
A staged path toward Signal Foundation Models that connect raw RF, structured scene truth, and qualified natural-language descriptions.
“Now” describes the present training contract. Later phases broaden annotation tasks and RF domains; they are not prerequisites for evaluating the released collection.
Radio-frequency records paired with grounded text, machine-readable labels, realized values, and generation provenance.
Extend grounded language from record description toward comparison, explanation, and qualified cross-domain reasoning.
New domains promoted only after their definitions and pilot evidence support scrutiny.
Released text annotations are paired with traceable declared record and scene evidence. Future work extends the annotation tasks and signal domains. Radar, satellite and navigation, spectrum environments, and wireless sensing remain research directions without published specifications or schedules.