Cadence uses a time-series foundation model for error-bounded lossy compression of demand data

Roberto Tacconelli · hf · 2026-09-09

Cadence pairs a time-series foundation model with an adaptive arithmetic coder for error-bounded lossy compression of demand time series, substantially outperforming classical predictors.

The authors also report negative results: the method does not work for lossless coding or under cross-batch determinism requirements, clarifying its applicability boundaries.

Original post →

More from Research

Research channel →