Research map / Probabilistic and scientific modeling / Treatment effects
Latent-variable causal inference
Infers individual effects using deep latent-variable and structural-equation models.
91 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Treatment effects, keyword cluster “causal effects · causal inference · iv”
State History Tried 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.
30 papers in 2023–24 and 33 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 22.9% to 20.1%.
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 · Long-term Causal Effects via Behavioral Game Theory (NeurIPS 2016 · 7 citations)
- 2017 · Causal Effect Inference with Deep Latent-Variable Models (NeurIPS 2017 · 281 citations)
- 2018 · Causal Inference with Noisy and Missing Covariates via Matrix Factorization (NeurIPS 2018 · 34 citations)
- 2019 · Counterfactual Randomization: Rescuing Experimental Studies from Obscured Confounding (AAAI 2019 · 11 citations)
- 2020 · Estimating Causal Effects Using Weighting-Based Estimators (AAAI 2020 · 20 citations)
- 2021 · Bounding Causal Effects on Continuous Outcome (AAAI 2021 · 17 citations)
- 2022 · Validating Causal Inference Methods (ICML 2022 · 11 citations)
- 2023 · Estimating Geographic Spillover Effects of COVID-19 Policies from Large-Scale Mobility Networks (AAAI 2023 · 16 citations)
- 2024 · Addressing Hidden Confounding with Heterogeneous Observational Datasets for Recommendation (NeurIPS 2024 · 3 citations)
- 2025 · Adjustment for Confounding using Pre-Trained Representations (ICML 2025 · 0 citations)
- 2026 · Towards a Holistic Understanding of Selection Bias for Causal Effect Identification (ICML 2026 · 0 citations)
Earliest papers here
- Long-term Causal Effects via Behavioral Game Theory (NeurIPS 2016 · 7 citations)
- Observational-Interventional Priors for Dose-Response Learning (NeurIPS 2016 · 3 citations)
- Causal Effect Inference with Deep Latent-Variable Models (NeurIPS 2017 · 281 citations)
What has been tried
70 papers here report 519 measured changes (a component added, removed or swapped, compared with a baseline): helped 59%, no clear effect 12%, hurt 22%, 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 |
|---|---|---|---|---|
| Use the same text instance to infer both proxies | Fully synthetic · bias | 0.004 -> 0.130 | Hurt | Proximal Causal Inference With Text Data (NeurIPS 2024 · 1 citation) |
| increase the number of non-focal LIWC categories from 2 to 9 | Amazon semi-synthetic linear outcome · proximity to ground-truth effect | Helped | Isolated Causal Effects of Natural Language (ICML 2025 · 0 citations) | |
| Increase polynomial model degree beyond the true degree | Synthetic polynomial data with unmeasured confounding · power | Hurt | Falsification of Unconfoundedness by Testing Independence of Causal Mechanisms (ICML 2025 · 0 citations) | |
| Vary α across the stated sensitivity range | XBOT and POAT sensitivity analysis · treatment effect estimate | 0.50 to 0.33 | Mixed | A Cautionary Tale on Integrating Studies with Disparate Outcome Measures for Causal Inference (NeurIPS 2025 · 0 citations) |
| Use the differentiable Gumbel relaxation instead of combinatorial subset search | Synthetic experiments with known DAGs · MAE | No clear effect | Doubly robust identification of treatment effects from multiple environments (ICLR 2025 · 0 citations) | |
| Add HSIC regularization | BC and Flickr, within-sample and out-of-sample · PEHE | 0.2478±0.0197 -> 0.2093±0.014; 0.1456±0.0129 -> 0.1265±0.0126; 0.194±0.0018 -> 0.1358±0.0026; 0.0673±0.0025 -> 0.0588±0. | Helped | Telling Peer Direct Effects from Indirect Effects in Observational Network Data (ICML 2025) |
| Add the treatment residual to the second-stage regression | Demand, α = 10, β = 10, S = 0 · MSE | 5.233±0.289 -> 1.278±0.155 | Helped | Two-Stage Shadow Inclusion Estimation: An IV Approach for Causal Inference under Latent Confounding and Collider Bias (ICML 2024) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 63 papers; from OpenAlex and the papers' affiliations).
- LMU Munich · 5 papers
- Peking University · 4 papers
- Massachusetts Institute of Technology · 3 papers
- Zhejiang University · 3 papers
- Swiss Federal Institute of Technology Lausanne · 3 papers
- Ludwig-Maximilians-Universität München · 3 papers
- Columbia University · 3 papers
- University of South Australia · 2 papers
- Stanford University · 2 papers
- Yokohama National University · 2 papers
Use it
Download BibTeX of 40 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
- Proximal and deep causal effect estimation (105 papers)
- Instrumental variable estimation (19 papers)
Elsewhere in the topic
- Meta-learners and partial identification (78 papers)
- Benchmarking treatment effect models (77 papers)
- Distributional treatment effects (51 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).