TempoPFN: synthetic pre-training of linear RNNs for zero-shot forecasting

Accepted at NeurIPS 2026, the paper introduces a library of synthetic time-series generators and augmentations to train a DeltaProduct-based linear RNN foundation model; the synthetic data has already been used in Toto-2 and TS-ICL.

• Is it worth staying for NeurIPS workshops without registration?2026-10-17
• Master's grad weighs self-funding a NeurIPS workshop trip2026-10-17
• NeurIPS 2026 Sydney waitlist draws concern2026-10-15
• ML PhD Student Asks What Industry Internships Really Want2026-10-14
• Community Discusses How to Write Top-Conference Papers2026-10-14