Research map / Time series / Causal and generative time series
ZeroDiff: Zero-Shot Time Series Reconstruction via Informed-Prior Diffusion
Yingda Fan, Dan Lu, Xiaowei Jia
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 informed prior (replace N(0,I) endpoint) (ablation) | ZeroDiff · Streamflow | NSE: fails (NSE<0) -> 0.596 | Improved |
| add VAE moment estimation (ablation) | ZeroDiff · Methane | NSE: -0.655 -> 0.788 | Improved |
| add VAE moment estimation (ablation) | ZeroDiff · Streamflow | NSE: 0.377 -> 0.596 | Improved |
| add VAE moment estimation (ablation) | ZeroDiff · Solar | NSE: 0.839 -> 0.846 | No clear effect |
| add diffusion calibration on top of prior Phi (ablation) | ZeroDiff · Methane | NSE: 0.481 -> 0.788 | Improved |
| add bidirectional denoiser (ablation) | ZeroDiff · Streamflow | NSE: 0.463 -> 0.596 | Improved |
| add bidirectional denoiser (ablation) | ZeroDiff · Solar | NSE: 0.519 -> 0.846 | Improved |
| add moment-guided weighting w_k (ablation) | ZeroDiff · Streamflow | NSE: 0.365 -> 0.463 | Improved |
| add moment-guided weighting w_k (ablation) | ZeroDiff · Solar | NSE: 0.518 -> 0.519 | No clear effect |
| add estimated moments as denoiser input (ablation) | ZeroDiff · Methane | NSE: 0.731 -> 0.788 | Improved |
| add estimated moments as denoiser input (ablation) | ZeroDiff · Streamflow | NSE: 0.551 -> 0.596 | Improved |
| swap LSTM backbone for Mamba | ZeroDiff · CAMELS | NSE: 0.596 -> 0.667 | Improved |
Closest papers by meaning
- Zero-shot Imputation with Foundation Inference Models for Dynamical Systems (ICLR 2025 · similarity 0.88)
- Analyzing and Guiding Zero-Shot Posterior Sampling in Diffusion Models (NeurIPS 2026 · similarity 0.88)
- True Zero-Shot Inference of Dynamical Systems Preserving Long-Term Statistics (NeurIPS 2025 · similarity 0.87)
- A Minimal Interpretable Architecture for Zero-Shot Reconstruction of Dynamical Systems (NeurIPS 2026 · similarity 0.87)
- ABC: Any-Subset Autoregression via Non-Markovian Diffusion Bridges in Continuous Time and Space (NeurIPS 2026 · similarity 0.86)
- Zero-shot forecasting of chaotic systems (ICLR 2025 · 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.
