aiqbench analysis
LLapDiff addresses the long-standing trade-off in irregular multivariate time series by combining continuous-time generative modeling with Laplace domain parameterization to avoid re-gridding distortions and solver drift. This approach enables horizon-wide forecasting and robust missing-value imputation without sequential integration over physical time, offering a significant methodological advance for complex temporal data. The model's acceptance as an ICML 2026 Spotlight underscores its potential to set new benchmarks in handling sparse, irregular observation streams.
Latent Laplace Diffusion for Irregular Multivariate Time Series
arXiv:2605.19805v2 Announce Type: replace-cross Abstract: Irregular multivariate time series impose a trade-off for long-horizon forecasting: discrete methods can distort temporal structure via re-gridding, while continuous-time models often require sequential solvers prone to drift. To bridge this gap, we present Latent Laplace Diffusion (LLapDiff), a generative framework that models the target as a low-dimensional latent trajectory, enabling horizon-wide generation without step-by-step integration over physical time. We guide the reverse process utilizing a stable modal parameterization motivated by stochastic port-Hamiltonian dynamics, and parameterize its mean evolution in the Laplace domain via learnable complex-conjugate poles, enabling direct evaluation over irregular timestamps. We also link continuous dynamics to irregular observations through renewal-averaging analysis, which maps sampling gaps to effective event-domain poles and motivates a gap-aware history summarizer. Extensive experiments show that LLapDiff improves over baselines in long-horizon forecasting, and its continuous-time generative nature supports missing-value imputation by querying the same model at historical timestamps. Code is available at https://github.com/pixelhero98/LLapDiffusion.
Abstract:Irregular multivariate time series impose a trade-off for long-horizon forecasting: discrete methods can distort temporal structure via re-gridding, while continuous-time models often require sequential solvers prone to drift. To bridge this gap, we present Latent Laplace Diffusion (LLapDiff), a generative framework that models the target as a low-dimensional latent trajectory, enabling horizon-wide generation without step-by-step integration over physical time. We guide the reverse process utilizing a stable modal parameterization motivated by stochastic port-Hamiltonian dynamics, and parameterize its mean evolution in the Laplace domain via learnable complex-conjugate poles, enabling direct evaluation over irregular timestamps. We also link continuous dynamics to irregular observations through renewal-averaging analysis, which maps sampling gaps to effective event-domain poles and motivates a gap-aware history summarizer. Extensive experiments show that LLapDiff improves over baselines in long-horizon forecasting, and its continuous-time generative nature supports missing-value imputation by querying the same model at historical timestamps. Code is available at this https URL.
Accepted as a Spotlight at ICML 2026. The Version of Record will appear in Proceedings of Machine Learning Research (PMLR). 27 pages, 5 figures. Code: this https URL
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
arXiv:2605.19805 [cs.LG]
(or arXiv:2605.19805v2 [cs.LG] for this version)
Full text under the publisher's open licence · 197 words
Related coverage
LogSig-SSM: Time-Series Modelling with Multi-Scale Log-Signature Compression for State-Space Models
arXiv · Oct 7, 2026
Scan-Parallel Variational Inference for event-driven Latent Jump Models
arXiv · Oct 12, 2026
QuITE: Query-Based Irregular Time Series Embedding
arXiv · Jun 3, 2026
FADTI: Fourier and Attention Driven Diffusion for Multivariate Time Series Imputation
arXiv · Sep 5, 2026