ResearchResearch paperLarge Language Models · Training & Scaling1 source · Oct 19, 2026

CTLite: Lightweight Irregular Multivariate Time Series Forecasting with Reusable Continuous-Time Summaries

We propose CTLite, a lightweight forecasting framework that decouples historical summarisation from learned prediction.

Irregular multivariate time series forecasting (IMTSF) requires capturing temporal dynamics and cross-variable dependencies from sparse, asynchronous observations.
Existing methods often rely on learned temporal encoders, coupling historical representation construction with model optimisation and repeatedly processing the same observations.
Specifically, a parameter-free continuous-time moment encoder summarises local levels and trends at multiple temporal scales while suppressing poorly supported slope estimates.
A compact readout combines damped local extrapolation with a low-rank cross-variable correction that accounts for observation freshness, target availability, and forecast horizon.

[1]CTLite: Lightweight Irregular Multivariate Time Series Forecasting with Reusable Continuous-Time Summaries
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 01:32 PM
We propose CTLite, a lightweight forecasting framework that decouples historical summarisation from learned prediction.
Irregular multivariate time series forecasting (IMTSF) requires capturing temporal dynamics and cross-variable dependencies from sparse, asynchronous observations.

Extractive summary: sentences quoted from the sources.

Before this