Research map / Probabilistic and scientific modeling / Treatment effects
Benchmarking treatment effect models
Evaluates limits and representation learning methods for heterogeneous treatment effect estimation.
77 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Treatment effects, keyword cluster “treatment effects · heterogeneous treatment · average treatment”
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.
22 papers in 2023–24 and 22 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 16.8% to 13.4%.
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
- 2017 · Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes (NeurIPS 2017 · 152 citations)
- 2018 · GANITE: Estimation of Individualized Treatment Effects using Generative Adversarial Nets (ICLR 2018 · 179 citations)
- 2019 · Adapting Neural Networks for the Estimation of Treatment Effects (NeurIPS 2019 · 52 citations)
- 2020 · Learning Counterfactual Representations for Estimating Individual Dose-Response Curves (AAAI 2020 · 68 citations)
- 2021 · Causal-BALD: Deep Bayesian Active Learning of Outcomes to Infer Treatment-Effects from Observational Data (NeurIPS 2021 · 9 citations)
- 2022 · Imbalance-Aware Uplift Modeling for Observational Data (AAAI 2022 · 9 citations)
- 2023 · Learning Instrumental Variable from Data Fusion for Treatment Effect Estimation (AAAI 2023 · 4 citations)
- 2024 · Contrastive Balancing Representation Learning for Heterogeneous Dose-Response Curves Estimation (AAAI 2024 · 8 citations)
- 2025 · LLM-Driven Treatment Effect Estimation Under Inference Time Text Confounding (NeurIPS 2025 · 1 citation)
- 2026 · Auxiliary Gene Learning: Spatial Gene Expression Estimation by Auxiliary Gene Selection (AAAI 2026 · 1 citation)
Earliest papers here
- Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes (NeurIPS 2017 · 152 citations)
- Estimating individual treatment effect: generalization bounds and algorithms (ICML 2017 · 131 citations)
- Matching on Balanced Nonlinear Representations for Treatment Effects Estimation (NeurIPS 2017 · 36 citations)
What has been tried
56 papers here report 517 measured changes (a component added, removed or swapped, compared with a baseline): helped 65%, no clear effect 9%, 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 |
|---|---|---|---|---|
| add balancing representation (BR) to CRNet | Data-1, Data-2, Data-5, Data-10, IHDP-1, News-2, News-4, News-8, News-16 · MISE | Data-1: 2.04 ± 0.5 -> 1.69 ± 0.5; Data-2: 2.56 ± 1.0 -> 2.07 ± 0.8; Data-5: 4.76 ± 1.2 -> 3.05 ± 0.7; Data-10: 9.69 ± 5. | Helped | Contrastive Balancing Representation Learning for Heterogeneous Dose-Response Curves Estimation (AAAI 2024 · 8 citations) |
| Add the GI and GP counterfactual losses to factual training | TCGA-0, TCGA-1, TCGA-2, IHDP, NEWS · CF error | 0.18 (0.10) -> 0.15; 0.09 (0.08) -> 0.09; 0.17 (0.03) -> 0.12; 2.05 (0.00) -> 1.89; 1.09 (0.00) -> 1.07 | Helped | Continuous Treatment Effect Estimation Using Gradient Interpolation and Kernel Smoothing (AAAI 2024 · 3 citations) |
| add cross-attention to ACFR | News out-of-sample · MISE | 1.58 ± 0.15 -> 1.12 ± 0.12 | Helped | Adversarially Balanced Representation for Continuous Treatment Effect Estimation (AAAI 2024 · 2 citations) |
| add DIVE instead of final-stage patient/graph concatenation | four downstream TEE datasets · AUC | Helped | KG-TREAT: Pre-training for Treatment Effect Estimation by Synergizing Patient Data with Knowledge Graphs (AAAI 2024 · 2 citations) | |
| Vary the prompt strategy among factual, narrative, and symptom-focused prompts | IST and MIMIC-III · PEHE | No clear effect | LLM-Driven Treatment Effect Estimation Under Inference Time Text Confounding (NeurIPS 2025 · 1 citation) | |
| Add counterfactual coverage maximization to FCCM- | Toy · sqrt PEHE | +0%, +17%, +18%, +18%, +16% | Helped | Enhancing Treatment Effect Estimation via Active Learning: A Counterfactual Covering Perspective (ICML 2025 · 0 citations) |
| Use multiple optimization trials to tune the evaluation models | IHDP, ACIC, Twins, News · √PEHE | Helped | AutoCATE: End-to-End, Automated Treatment Effect Estimation (ICML 2025 · 0 citations) | |
| vary the latent confounder and noise dimensions | Synthetic data · causal-effect estimate | No clear effect | Coupling Generative Modeling and an Autoencoder with the Causal Bridge (NeurIPS 2025 · 0 citations) | |
| replace the best constant α with adaptive shrinkage | Hippocampus R · Mean AE | 0.257 (±0.007) -> 0.243 (±0.003) | Helped | Adaptive Shrinkage Estimation for Personalized Deep Kernel Regression in Modeling Brain Trajectories (ICLR 2025 · 0 citations) |
| Add individualized fine-tuning of patient-specific ODE constants | Synthetic longitudinal treatment-effect datasets · counterfactual prediction performance | Helped | ODE Discovery for Longitudinal Heterogeneous Treatment Effects Inference (ICLR 2024 · 0 citations) | |
| Add propensity-score regularization to the reward | Uganda · outcome error | 1.701±0.161 -> 1.662±0.136 | Helped | DISCRET: Synthesizing Faithful Explanations For Treatment Effect Estimation (ICML 2024 · 0 citations) |
| Add sequential targeting | Complex synthetic survival, τ = 10, 20, 30 · RMSE | 0.0263 -> 0.0571; 0.0332 -> 0.0379; 0.0533 -> 0.0571 | Hurt | Longitudinal Targeted Minimum Loss-based Estimation with Temporal-Difference Heterogeneous Transformer (ICML 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 13 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 13 papers
- MIMIC-III · 2 of 13 papers
Baselines
- DFPV · 2 of 13 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 44 papers; from OpenAlex and the papers' affiliations).
- University of Cambridge · 6 papers
- University of California, Los Angeles · 5 papers
- Emory University · 4 papers
- Ludwig-Maximilians-Universität München · 4 papers
- LMU Munich · 4 papers
- Zhejiang University · 3 papers
- University of California, Berkeley · 3 papers
- Tsinghua University · 2 papers
- Peking University · 2 papers
- Mohamed bin Zayed University of Artificial Intelligence · 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
- Meta-learners and partial identification (78 papers)
- Distributional treatment effects (51 papers)
Elsewhere in the topic
- Proximal and deep causal effect estimation (105 papers)
- Latent-variable causal inference (91 papers)
- Instrumental variable estimation (19 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).