Research map / Time series / Causal and generative time series
Generative temporal point processes
Uses diffusion, intensity-free and meta-learning methods for temporal point processes.
53 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Causal and generative time series, keyword cluster “conformal · causal discovery · predictions”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
14 papers in 2023–24 and 16 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 9.0% to 7.0%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Infinite Hidden Semi-Markov Modulated Interaction Point Process (NeurIPS 2016)
- 2017 · Predicting User Activity Level In Point Processes With Mass Transport Equation (NeurIPS 2017 · 10 citations)
- 2018 · Learning Conditional Generative Models for Temporal Point Processes (AAAI 2018 · 42 citations)
- 2019 · Event Cameras, Contrast Maximization and Reward Functions: An Analysis (CVPR 2019 · 93 citations)
- 2020 · Intensity-Free Learning of Temporal Point Processes (ICLR 2020 · 22 citations)
- 2021 · UNIPoint: Universally Approximating Point Processes Intensities (AAAI 2021 · 8 citations)
- 2022 · DeepGPD: A Deep Learning Approach for Modeling Geospatio-Temporal Extreme Events (AAAI 2022 · 18 citations)
- 2023 · C-NTPP: Learning Cluster-Aware Neural Temporal Point Process (AAAI 2023 · 5 citations)
- 2024 · EasyTPP: Towards Open Benchmarking Temporal Point Processes (ICLR 2024 · 6 citations)
- 2025 · EBS-EKF: Accurate and High Frequency Event-based Star Tracking (CVPR 2025 · 7 citations)
- 2026 · In-Context Learning of Temporal Point Processes with Foundation Inference Models (ICLR 2026 · 0 citations)
Earliest papers here
- Infinite Hidden Semi-Markov Modulated Interaction Point Process (NeurIPS 2016)
- Predicting User Activity Level In Point Processes With Mass Transport Equation (NeurIPS 2017 · 10 citations)
- Variational Policy for Guiding Point Processes (ICML 2017 · 8 citations)
What has been tried
40 papers here report 378 measured changes (a component added, removed or swapped, compared with a baseline): helped 70%, no clear effect 10%, hurt 13%, mixed 7%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add intensity-dependent centroiding offset correction | Velocity Sweep 1 · about difference (arcseconds) | 83.1 -> 60.3 | Helped | EBS-EKF: Accurate and High Frequency Event-based Star Tracking (CVPR 2025 · 7 citations) |
| Add input-dependent dynamics | Real-world ablation experiments · performance | Helped | Deep Continuous-Time State-Space Models for Marked Event Sequences (NeurIPS 2025 · 1 citation) | |
| Increase HYPRO weighted proposals from 2 to 32 | Taxi, Taobao, and StackOverflow-V2 · avg OTD | Taxi: 20.35 -> 18.81; Taobao: 39.73 -> 36.96; StackOverflow-V2: 39.84 -> 38.84 | Helped | Decomposable Transformer Point Processes (NeurIPS 2024 · 1 citation) |
| add thresholding to mark prediction | Retweet rare marks · macro-F1 | 0.0333 -> 0.2010 | Helped | Addressing Mark Imbalance in Integration-free Marked Temporal Point Processes (NeurIPS 2025 · 0 citations) |
| jointly model event types and inter-arrival times through interacting denoising processes rather than independent processes | Amazon · OTD | 40.49 ± 0.60 -> 37.73 ± 0.20 | Helped | Interacting Diffusion Processes for Event Sequence Forecasting (ICML 2024 · 0 citations) |
| Use the alternative influence function instead of the original | MIMIC-II · log-likelihood | -2.045 -> -2.150 | Hurt | Residual TPP: A Unified Lightweight Approach for Event Stream Data Analysis (ICML 2025) |
| Add Gaussian noise and use denoising score matching | Earthquake · CS-time | Helped | Beyond Point Prediction: Score Matching-based Pseudolikelihood Estimation of Neural Marked Spatio-Temporal Point Process (ICML 2024) | |
| Add the Brownian-motion diffusion term | Earthquake · NLL | 1.594 -> 1.305 | Helped | Neural Jump-Diffusion Temporal Point Processes (ICML 2024) |
What reviewers will expect
Datasets and baselines named in the 9 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Retweet · 4 of 9 papers
- StackOverflow · 4 of 9 papers
- MIMIC-II · 3 of 9 papers
- Amazon · 3 of 9 papers
- Taobao · 3 of 9 papers
- Taxi · 2 of 9 papers
- Last.fm · 2 of 9 papers
Baselines
- THP · 5 of 9 papers
- RMTPP · 4 of 9 papers
- SAHP · 4 of 9 papers
- AttNHP · 4 of 9 papers
- NHP · 3 of 9 papers
- MHP · 2 of 9 papers
- FullyNN · 2 of 9 papers
- IFTPP · 2 of 9 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 30 papers; from OpenAlex and the papers' affiliations).
- Ant Group · 2 papers
- Georgia Institute of Technology · 2 papers
- Alibaba Group · 2 papers
- Amazon · 2 papers
- GE Healthcare · 2 papers
Use it
Download BibTeX of 37 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Probabilistic and conformal forecasting (121 papers)
- State-space and latent dynamics forecasting (94 papers)
- Dynamical system reconstruction (90 papers)
- Robust forecasting and evaluation (91 papers)
Elsewhere in the topic
- Irregular series and imputation (59 papers)
- Diffusion models for time series (55 papers)
- Neural differential equations (52 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).