Art of Novel Signals: Predicting and Forecasting with High Confidence
This solicitation's response window has closed.
Summary
DARPA seeks proposals to develop a predictive model that uses temporal knowledge graph forecasting with in‑context learning on novel multilingual and multimodal data, such as low‑resource radio broadcasts, to extend event forecasting from days to weeks with at least 90% precision. The effort focuses on building a pipeline that ingests noisy audio, applies automatic speech recognition (e.g., Whisper Large‑v3), and integrates the transcriptions into a temporal knowledge graph for early warning in data‑sparse regions.