Research map / Time series / Causal and generative time series
KineFlow: Kinematic Second-Order Flow Matching for Time-Series Forecasting
Haiqi Jiang, Hui Xiong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add informative phase-space prior initialization instead of Gaussian noise (ablation) | KineFlow · ablation avg | MSE: 0.346 -> 0.302 | Improved |
| add informative phase-space prior initialization instead of Gaussian noise (ablation) | KineFlow · ablation avg | V-Err: 0.232 -> 0.166 | Improved |
| add velocity (second-order dynamics) to the state (ablation) | KineFlow · ablation avg | MSE: 0.333 -> 0.302 | Improved |
| add velocity (second-order dynamics) to the state (ablation) | KineFlow · ablation avg | CRPS: 0.337 -> 0.320 | Improved |
| add velocity (second-order dynamics) to the state (ablation) | KineFlow · ablation avg | V-Err: 0.214 -> 0.166 | Improved |
| initialize V0 from historical momentum instead of zero (ablation) | KineFlow · ablation avg | MSE: 0.311 -> 0.302 | Improved |
| initialize V0 from historical momentum instead of zero (ablation) | KineFlow · ablation avg | V-Err: 0.185 -> 0.166 | Improved |
| use KineFlow instead of TimeXer | TimeXer · standard benchmarks avg | MSE: 0.315 -> 0.269 | Improved |
| use KineFlow instead of TSFlow | TSFlow · standard benchmarks avg | CRPS: 0.332 -> 0.293 | Improved |
| use KineFlow instead of TimeXer | TimeXer · AEMO | MSE: 0.628 -> 0.400 | Improved |
| use KineFlow (0.90M params, few-shot) instead of Sundial zero-shot | Sundial · AEMO VIC to SA | MSE: 0.285 -> 0.246 | Improved |
| use KineFlow instead of Electricity-ETTm1 best baseline on ETTm1 MSE | TimeXer · ETTm1 | MSE: 0.382 -> 0.397 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 5 of 5 on most shared results.
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 6 of 6 shared results
- Diffusion-TS — Diffusion-TS: Interpretable Diffusion for General Time Series Generation (ICLR 2024) · better on 3 of 3 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 1 of 1 shared result
- Autoformer — Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting (NeurIPS 2021) · better on 1 of 1 shared result
- CSDI — CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation (NeurIPS 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Flow Matching with Gaussian Process Priors for Probabilistic Time Series Forecasting (ICLR 2025 · similarity 0.89)
- PrismFlow: Residual Dynamics for Flow Matching in Time-Series Generation (NeurIPS 2026 · similarity 0.88)
- TARFVAE: Efficient One-Step Generative Time Series Forecasting via TARFLOW based VAE (NeurIPS 2025 · similarity 0.87)
- Sundial: A Family of Highly Capable Time Series Foundation Models (ICML 2025 · similarity 0.87)
- Universal Time Series Generation with Neural Controlled Differential Equations (NeurIPS 2026 · similarity 0.87)
- Trajectory Flow Matching with Applications to Clinical Time Series Modelling (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
