Research map / Probabilistic and scientific modeling / Treatment effects
Proximal and deep causal effect estimation
Estimates treatment effects with proximal learning, transformers and balancing for structured or continuous treatments.
105 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 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.
27 papers in 2023–24 and 44 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 20.6% to 26.8%.
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 · Recommendations as Treatments: Debiasing Learning and Evaluation (ICML 2016 · 138 citations)
- 2017 · Deep IV: A Flexible Approach for Counterfactual Prediction (ICML 2017)
- 2018 · Generalized Adjustment Under Confounding and Selection Biases (AAAI 2018 · 37 citations)
- 2019 · Validating Causal Inference Models via Influence Functions (ICML 2019 · 36 citations)
- 2020 · Estimating the Effects of Continuous-valued Interventions using Generative Adversarial Networks (NeurIPS 2020 · 50 citations)
- 2021 · VCNet and Functional Targeted Regularization For Learning Causal Effects of Continuous Treatments (ICLR 2021 · 17 citations)
- 2022 · Causal Transformer for Estimating Counterfactual Outcomes (ICML 2022 · 18 citations)
- 2023 · Direct Heterogeneous Causal Learning for Resource Allocation Problems in Marketing (AAAI 2023 · 21 citations)
- 2024 · Causal Walk: Debiasing Multi-Hop Fact Verification with Front-Door Adjustment (AAAI 2024 · 11 citations)
- 2025 · Alleviating Dual Biases in Recommendation (Student Abstract) (AAAI 2025 · 0 citations)
- 2026 · Invariant Feature Learning for Counterfactual Watch-time Prediction in Video Recommendation (AAAI 2026 · 1 citation)
Earliest papers here
- Recommendations as Treatments: Debiasing Learning and Evaluation (ICML 2016 · 138 citations)
- Deep IV: A Flexible Approach for Counterfactual Prediction (ICML 2017)
- Generalized Adjustment Under Confounding and Selection Biases (AAAI 2018 · 37 citations)
What has been tried
75 papers here report 595 measured changes (a component added, removed or swapped, compared with a baseline): helped 65%, no clear effect 10%, hurt 18%, 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 |
|---|---|---|---|---|
| remove causal intervention | PolitiHop · accuracy | 80.00 -> 77.00 | Hurt | Causal Walk: Debiasing Multi-Hop Fact Verification with Front-Door Adjustment (AAAI 2024 · 11 citations) |
| Use adaptive rather than random kernel-function selection | Coat, Music, Product · Debiasing performance | Helped | Debiased Collaborative Filtering with Kernel-Based Causal Balancing (ICLR 2024 · 7 citations) | |
| Set α=1, using purely causal loss instead of predictive loss | T1DEXI post-exercise glucose prediction · RMSE | Hurt | Hybrid$^2$ Neural ODE Causal Modeling and an Application to Glycemic Response (ICML 2024 · 2 citations) | |
| add the InfoNCE loss | Cancer simulation, γ=1, τ=1–10 · NRMSE | 1.05 -> 1.07 | Helped | Causal Contrastive Learning for Counterfactual Regression Over Time (NeurIPS 2024 · 1 citation) |
| Use the accelerated pseudo-outcome approximation instead of the exact fixed-point pseudo-outcomes | VCM and HTE · estimation accuracy | No clear effect | Generalized Random Forests Using Fixed-Point Trees (ICML 2025 · 0 citations) | |
| Add the minimization technique to Cumulant | G3 · relative absolute error | Helped | Causal Effect Identification in lvLiNGAM from Higher-Order Cumulants (ICML 2025 · 0 citations) | |
| Replace Transformer with RNNs for intensity estimation | Synthetic data, T=64, M=1 · relative error rate (RER) | Hurt | Transformer-Based Spatial-Temporal Counterfactual Outcomes Estimation (ICML 2025 · 0 citations) | |
| Vary α; use an intermediate regularization value rather than an arbitrarily small α | Linear simulation, κ = γ = 1 · nCER | Helped | An Analysis of Causal Effect Estimation using Outcome Invariant Data Augmentation (NeurIPS 2025 · 0 citations) | |
| Add attention | Synthetic, τ=5, β1=0.0–2.0 · RMSE | 0.50 ± 0.00 -> 0.33 ± 0.00; 0.46 ± 0.00 -> 0.35 ± 0.00; 0.51 ± 0.00 -> 0.40 ± 0.00; 0.45 ± 0.01 -> 0.44 ± 0.00; 0.47 ± 0 | Mixed | GST-UNet: A Neural Framework for Spatiotemporal Causal Inference with Time-Varying Confounding (NeurIPS 2025 · 0 citations) |
| add PPL decision loss | Marketing Data I · EOM | 1.0000 -> 1.0167 | Helped | Bi-Level Decision-Focused Causal Learning for Large-Scale Marketing Optimization: Bridging Observational and Experimental Data (NeurIPS 2025 · 0 citations) |
| add causal alignment loss | LIDC-IDRI · CAM precision | 0.057 (0.07) -> 0.587 (0.08) | Helped | Learning Causal Alignment for Reliable Disease Diagnosis (ICLR 2025 · 0 citations) |
| Add the L3 perturbation loss | BC(homo), Within Sample and Out-of Sample AME, ASE, and ATE · ε_average (MAE) | Helped | Doubly Robust Causal Effect Estimation under Networked Interference via Targeted Learning (ICML 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 24 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- IHDP · 2 of 24 papers
- Jobs · 2 of 24 papers
Baselines
- DragonNet · 4 of 24 papers
- T-learner · 3 of 24 papers
- Causal Forest · 2 of 24 papers
- S-Learner · 2 of 24 papers
- TARNet · 2 of 24 papers
- CFRNet · 2 of 24 papers
- DESCN · 2 of 24 papers
- X-learner · 2 of 24 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 71 papers; from OpenAlex and the papers' affiliations).
- University of Cambridge · 5 papers
- Beijing Technology and Business University · 4 papers
- Peking University · 4 papers
- Mohamed bin Zayed University of Artificial Intelligence · 3 papers
- Ludwig-Maximilians-Universität München · 3 papers
- LMU Munich · 3 papers
- Massachusetts Institute of Technology · 3 papers
- Inria · 3 papers
- Google · 3 papers
- Nanjing University · 2 papers
Use it
Download BibTeX of 38 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
- Latent-variable causal inference (91 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).