Optimal Policy Adaptation under Covariate Shift

Peng Wu; Qinwei Yang; Ruocheng Guo; Xueqing Liu; Zhaoqing Tian

REVIEW 3 major objections 5 minor 45 references

Optimal Policy Adaptation under Covariate Shift

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that a doubly robust, semiparametrically efficient estimator of the target reward lets a policy trained on source data approximate the oracle policy for a target domain under covariate shift.

desk verdict Standard doubly robust policy evaluation under covariate shift, wrapped in overstated novelty; the regret bound in Theorem 3 does not hold as written. read the letter →

arxiv 2501.08067 v2 pith:3PN7E4T4 submitted 2025-01-14 cs.LG

Xueqing Liu , Qinwei Yang , Zhaoqing Tian , Ruocheng Guo , Peng Wu This is my paper · ORCID

classification cs.LG

keywords policylearningcovariateshifttransferdoublyrobustestimationsemiparametricefficiencyefficientinfluencefunctionoptimaltreatmentcausalinference

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the optimal treatment policy for a target population, where only covariates are observed, can be learned reliably from a fully observed source population even when the two populations have different covariate distributions. The authors derive the efficient influence function for the target reward and build a doubly robust estimator that reaches the semiparametric efficiency bound, then learn the policy by maximizing this estimated reward. They prove consistency, asymptotic normality, an explicit bias formula, and a finite-sample generalization error bound for the learned policy, and their experiments show the estimator beats direct regression and inverse-probability weighting on simulated and real-world data. If correct, the result gives a principled way to transfer treatment rules from study populations to new target settings with a quantifiable closeness to the oracle policy.

What carries the argument

The efficient influence function $\varphi_R$ of the target reward, whose variance is the semiparametric efficiency bound, and the estimator $\hat R_{SE}(\pi)$ built from it. The estimator combines a source residual term weighted by inverse propensity and inverse sampling odds, a target outcome-regression term, and a centering term; double robustness is visible in the bias formula, which is a product of the outcome-regression error and the propensity/sampling-score error. This object is the load-bearing mechanism because it fixes the form of the estimator, supplies the asymptotic variance benchmark, and yields the regret bound for the learned policy.

What would settle it

Obtain a target dataset that also contains outcomes, and on the overlap support test whether source-fitted outcome regressions for each arm predict target outcomes as well as target-fitted regressions; a systematic gap falsifies transportability and shows the SE estimator is biased. In simulation, generate a concept shift by changing only the target outcome function while keeping covariate distributions, and verify that the bias of $\hat R_{SE}(\pi)$ grows with the size of the shift.

Watch

Extended reading notes

Core claim

Under Assumptions 1 and 2 — unconfoundedness and overlap in the source, plus transportability of conditional outcome means and overlap of the sampling score — the target reward $R(\pi)$ is identified as $E[\pi(X)\mu_1(X) + (1-\pi(X))\mu_0(X) \mid G=0]$. The paper's central claim is that the estimator $\hat R_{SE}(\pi)$, formed from the efficient influence function $\varphi_R$, is doubly robust (unbiased if either the outcome regressions are correct or the propensity and sampling scores are correct), consistent, asymptotically normal, and semiparametric efficient: $\sqrt{n}(\hat R_{SE}(\pi) - R(\pi)) \xrightarrow{d} N(0, \mathrm{Var}(\varphi_R))$. The learned policy $\hat\pi = \arg\max_{\pi\in\Pi}\hat R_{SE}(\pi)$ then satisfies the generalization bound of Theorem 3, which makes its reward approach that of the oracle threshold policy $\pi^*_0(x)=1\{\tau(x)\ge 0\}$ as the sample grows.

Load-bearing premise

The load-bearing premise is transportability, Assumption 2(i): the conditional outcome means $E[Y(a)|X]$ are the same in the source and target domains; since target data contain no outcomes, any violation of this assumption makes $R(\pi)$ unidentified and biases every estimator built from source data.

Editorial extensions

If this is right

With source data containing outcomes and target data with covariates only, the SE estimator estimates the target reward with smaller mean squared error than direct or IPW estimators, and the learned policy yields higher true reward and lower regret in the paper's experiments.
Under the product-rate conditions on nuisance estimates, no regular estimator can have asymptotic variance below $\mathrm{Var}(\varphi_R)$, so the approach is optimal within the semiparametric model.
The finite-sample bound in Theorem 3 implies that when squared residuals are bounded and nuisance estimates are accurate, the regret of the learned policy relative to the oracle decays as the sample size grows.
The same efficient-influence-function construction applies to the whole-domain reward $V(\pi)$, giving a doubly robust estimator for policies intended to generalize across the combined source–target population.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If transportability fails in a structured way — for instance, the outcome shift is a known function of covariates — the same influence-function machinery could be adapted to estimate the shift and the target reward jointly, an extension the paper does not spell out.
A small labeled target sample could provide a practical check of Assumption 2(i): compare source-imputed outcome regressions with target-fitted ones on the overlap support, and treat a significant discrepancy as evidence that the reported reward estimates are biased.
The abstract also announces a sensitivity analysis for the case of simultaneous covariate and concept shift; the provided manuscript text does not develop that analysis, so it stands as a stated aim rather than a delivered result.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick read on 2501.08067. The honest headline: the policy evaluation part is standard but competently done; the policy learning part has a load-bearing gap in Theorem 3, and the abstract promises a sensitivity analysis that is not in the paper.

What is actually new is close to nothing. The efficient influence function for the reward under covariate shift with a covariate-only target sample appears in Uehara et al. 2020, which the paper itself cites. The double robustness and semiparametric efficiency claims are standard, and from what I can see the proofs are correct. The bias expression in Proposition 2 is useful and clearly derived. The extension to the whole-domain reward V(pi) is straightforward. So give credit where it is earned: the derivations are clean, the estimator is sound, and the asymptotic statements check out.

The soft spots are real and in proportion. Theorem 3's fluctuation term B(D0,D1,eta,Pi) sums over i=1..n terms involving A_i and Y_i, but target units have neither; as written it is not even computable. If you restrict the sum to source units, it captures only source residual variance and ignores the target sampling variance of the estimated reward. With oracle nuisances, B=0 while hat_R_SE(pi) is still random because of the target-only sample, so the high-probability regret bound cannot hold. This matters because Theorem 3 is the only formal bridge from reward estimation to the claim that the learned policy approximates the optimal policy. That bridge is broken.

Also, the abstract advertises a novel sensitivity analysis for concept shift; the body of this version has no such section. That is a gap between promise and delivery. And while the paper says policy transfer is "rarely discussed" or "underexplored," it cites Uehara et al. 2020, which contains the central estimator; the positioning should be explicit about what is being added beyond that work. The experiments are fine but even the "real-world" dataset uses simulated treatments and outcomes, so it is still a simulation.

I want to be clear: I am not saying the estimator is wrong. The EIF, double robustness, and efficiency results look right. But the headline contributions beyond Uehara et al. are the generalization bound and the sensitivity analysis, and neither is currently supported.

Recommendation: send it to a serious referee. The evaluation half is solid and worth publishing after revision, but the authors need to fix Theorem 3, add or remove the sensitivity analysis, and re-position against prior work. A referee can sort that out.

Referee Report

3 major / 5 minor

Summary. The paper studies policy transfer under covariate shift: given a source sample with covariates, treatment, and outcome, and a target sample with covariates only, the authors aim to estimate the target-domain reward of a treatment policy and to learn the policy maximizing it. Under unconfoundedness, overlap, and transportability assumptions, the paper identifies the reward, derives its efficient influence function and semiparametric efficiency bound, constructs a doubly robust and semiparametric efficient estimator, and proposes optimizing this estimator to learn the target-optimal policy. It also states a finite-sample generalization bound for the learned policy and reports simulation and real-covariate experiments. The abstract additionally promises a sensitivity analysis under concept shift, but the body does not contain such a section.

Significance. If the central claims hold, the paper makes a useful contribution to policy learning with unlabeled target data: the efficient influence function and the doubly robust estimator are standard in spirit but are carefully adapted to the two-sample source/target design, and the efficiency result is nontrivial. The explicit bias expression and the intended generalization bound are also valuable. The reward-estimation half of the paper appears sound; the policy-learning half currently rests on a flawed generalization bound. Because that flaw is local and appears fixable, the overall contribution is defensible after a major revision.

major comments (3)

[Section 4.2, Theorem 3] The fluctuation term B(D0,D1,eta,Pi) is not well-defined: the sum runs over i=1,...,n and contains Y_i - hat_mu_{A_i}(X_i), while units in D0 have neither A_i nor Y_i. If the sum is intended to run only over D1, then B is a source-only residual bound and omits the target component of hat_R_SE(pi), which is a sample average over n0 target units and fluctuates at order n0^{-1/2}. Even with oracle nuisance functions and zero outcome noise, B=0 while hat_R_SE(pi) is still random, so the claimed 1-eta high-probability inequality cannot hold as stated. The same issue affects part (ii) of the theorem.
[Section 4.2, Proposition 2] Proposition 2 defines Bias(hat_R_SE(pi)) as |E[hat_R_SE(pi)] - R(pi)|, but the displayed expression is a sum over the observed X_i and is therefore random rather than a deterministic bias. If the intent is a conditional bias given X_1,...,X_n, that conditioning must be stated, and the high-probability treatment in Theorem 3 must account for the fluctuation of this data-dependent bias term. As written, the bias term in Theorem 3 is ambiguous and the proof does not close this gap.
[Section 4.1, Theorem 2 and Algorithm 1] Theorem 2 states product-rate conditions on the nuisance estimates, but Algorithm 1 fits hat_mu, hat_e, and hat_s once on D1 and then evaluates the source correction terms on the same units. Standard double/debiased machine-learning results require cross-fitting or explicit Donsker/empirical-process conditions for such product-rate conditions to imply asymptotic normality and semiparametric efficiency. The theorem and its proof should state which mechanism is used; otherwise the efficiency claim is not justified for the implemented procedure.

minor comments (5)

[Abstract] The abstract promises a sensitivity analysis for simultaneous covariate and concept shifts, but the main text contains no sensitivity-analysis section or results; this claimed contribution should either be added or removed.
[Assumption 2(ii)] Assumption 2(ii) states 0<s(X)<1 only for X in the source domain, but the IPW and efficient-influence-function terms multiply by (1-s(X))/s(X) evaluated at target covariate values; the overlap condition should be stated for the union of the supports of the source and target covariate distributions.
[Section 4.2, Theorem 3] The theorem assumes a finite hypothesis space Pi, while Algorithm 1 and the examples in Section 2.2 describe continuous parametric policy classes; the relationship between the finite-Pi bound and the actual optimization procedure should be clarified.
[Section 5, Table 1] The table caption describes the reported quantities as estimated rewards, but the evaluation metric defined in Section 5.1 is the true reward computed from held-out target potential outcomes; the wording should be aligned with the definition.
[Section 5.2] The text calls the Communities and Crime experiment a real-world dataset, but the treatment and outcomes are simulated and only the covariates are real; this should be stated in the main text rather than only in the data-generation appendix description.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: identifiability, EIF, double robustness, and policy learning are derived from stated assumptions without reusing fitted quantities as predictions.

full rationale

The paper's central derivation chain is not circular. The target reward R(π) is identified directly from Assumptions 1–2 through Eq. (3), and the efficient influence function in Theorem 1 is derived from that identified functional using standard semiparametric theory (Tsiatis 2006; Newey 1990), not from the estimator itself. The constructed estimator R̂_SE(π) is a plug-in AIPW-style estimator whose double robustness (Proposition 1) and semiparametric efficiency (Theorem 2) are checked against the true nuisance functions under product-rate conditions; no fitted parameter is renamed as a prediction. The policy step, π̂ = argmax_{π∈Π} R̂_SE(π), is the standard empirical welfare maximization procedure, and the regret bound in Theorem 3 is an attempt to control R(π̂) − R̂(π̂) plus bias terms, not a quantity forced by construction. The self-citations in the paper (e.g., Wu et al. 2025, Wu and Mao 2025, Yang et al. 2024) appear in related-work and assumption-recipe contexts alongside many independent references, and none is the sole load-bearing justification for the paper's central results, so they do not constitute circularity. Separately, the skeptic's concern about Theorem 3 is a substantive correctness issue—B(D0,D1,η,Π) is written with source-side residuals involving Yi and Ai and does not appear to capture the n0^{-1/2} target-only sampling fluctuation in R̂_SE(π)—but that is an error or gap in the generalization analysis, not a circular step, because the bound is not used as an input to the estimator or policy.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The method rests on standard causal identifiability assumptions, including unconfoundedness, overlap, and transportability, plus regularity conditions for semiparametric efficiency. No free parameters are fit as part of the core derivation; the experimental generative model is not part of the theoretical claim. The main fragility is transportability of conditional outcome means, which cannot be checked from the target data because target outcomes are not observed.

assumptions (5)

standard math Consistency and the potential-outcome model: $Y = A Y(1) + (1-A) Y(0)$ (Section 2.1).
Bridges observed outcomes to potential outcomes; standard in causal inference and used throughout the paper.
domain assumption Assumption 1: unconfoundedness $(Y(1), Y(0)) \perp\!\!\perp A \mid X, G=1$ and overlap $0 < e_1(X) < 1$ in the source domain.
Identifies source-domain conditional outcome means from observed treatment and outcome data; untestable from data.
domain assumption Assumption 2: transportability $(Y(0), Y(1)) \perp\!\!\perp G \mid X$ and sampling-score overlap $0 < s(X) < 1$.
Transfers conditional outcome means to the target domain; this is the critical untestable premise that makes the target reward identified.
domain assumption Semiparametric regularity and nuisance convergence conditions in Theorem 2, e.g. $\|\hat e_1 - e_1\|_2 \|\hat\mu_a - \mu_a\|_2 = o_P(n^{-1/2})$.
Needed for asymptotic normality and efficiency of the estimator; standard in double machine learning but not verified in the experiments.
domain assumption Finite hypothesis space $\Pi$ and boundedness assumptions for the generalization bound in Theorem 3.
The bound uses a union bound over finite $\Pi$ and treats a data-dependent square-root term as a valid concentration term, which is not fully justified in the main text.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Optimal Policy Adaptation under Covariate Shift." pith.science (2026). https://pith.science/paper/3PN7E4T4

@misc{pith2026250108067,
  author       = {Pith},
  title        = {Pith review of: Optimal Policy Adaptation under Covariate Shift},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3PN7E4T4}},
  note         = {Machine review of arXiv:2501.08067}
}

read the original abstract

Transfer learning of prediction models has been extensively studied, while the corresponding policy learning approaches are rarely discussed. In this paper, we propose principled approaches for learning the optimal policy in the target domain by leveraging two datasets: one with full information from the source domain and the other from the target domain with only covariates. First, under the setting of covariate shift, we formulate the problem from a perspective of causality and present the identifiability assumptions for the reward induced by a given policy. Then, we derive the efficient influence function and the semiparametric efficiency bound for the reward. Based on this, we construct a doubly robust and semiparametric efficient estimator for the reward and then learn the optimal policy by optimizing the estimated reward. Moreover, we theoretically analyze the bias and the generalization error bound for the learned policy. Extensive experiments demonstrate that the approach not only estimates the reward more accurately but also yields a policy that closely approximates the theoretically optimal policy.

Figures

Figures reproduced from arXiv: 2501.08067 by the authors.

**Figure 2.** Comparison of three methods with different treatments in the target dataset [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 43 canonical work pages

[1]

Policy learning with observational data

[Athey and Wager, 2021] Susan Athey and Stefan Wager. Policy learning with observational data. Econometrica, 89(1):133–161,

work page 2021
[5]

Chernozhukov, D

[Chernozhukov et al., 2018] V . Chernozhukov, D. Chetverikov, M. Demirer, and et al. Double/de- biased machine learning for treatment and structural parameters. The Econometrics Journal, 21,

work page 2018
[6]

Causal inference methods for combin- ing randomized trials and observational studies: a review

[Colnet et al., 2024] B´en´edicte Colnet, Imke Mayer, Guan- hua Chen, and et al. Causal inference methods for combin- ing randomized trials and observational studies: a review. Statistical Science, 39(1):165–191,

work page 2024
[8]

Dahabreh, Sarah E

[Dahabreh et al., 2020] Issa J. Dahabreh, Sarah E. Robert- son, Jon A. Steingrimsson, Elizabeth A. Stuart, and Miguel A. Hern´an. Extending inferences from a random- ized trial to a new target population.Statistics in Medicine, 39:1999–2014,

work page 2020
[12]

Predicting the efficacy of future training programs using past experiences at other locations

[Hotz et al., 2005] V Joseph Hotz, Guido W Imbens, and Julie H Mortimer. Predicting the efficacy of future training programs using past experiences at other locations. Jour- nal of Econometrics, 125(1-2):241–270,

work page 2005
[16]

On the role of surrogates in the efficient estimation of treat- ment effects with limited outcome data

[Kallus and Mao, 2024] Nathan Kallus and Xiaojie Mao. On the role of surrogates in the efficient estimation of treat- ment effects with limited outcome data. Journal of the Royal Statistical Society Series B: Statistical Methodol- ogy, page qkae099,

work page 2024
[18]

Transfer learning techniques for medical image analysis: A review

[Kora et al., 2022] Padmavathi Kora, Chui Ping Ooi, Oliver Faust, and et al. Transfer learning techniques for medical image analysis: A review. Biocybernetics and Biomedical Engineering, 42(1):79–107,

work page 2022
[19]

Kosorok and Eric B

[Kosorok and Laber, 2019] Michael R. Kosorok and Eric B. Laber. Precision medicine. Annual Review of Statistics and Its Application, 6:263–86,

work page 2019

Show all 45 references

[21]

Generalizing study results: a poten- tial outcomes perspective

[Lesko et al., 2017] Catherine R Lesko, Ashley L Buchanan, Daniel Westreich, Jessie K Edwards, Michael G Hudgens, and Stephen R Cole. Generalizing study results: a poten- tial outcomes perspective. Epidemiology, 28(4):553–561,

2017
[22]

Policy learning with constraints in model-free rein- forcement learning: A survey

[Liu et al., 2021] Yongshuai Liu, Avishai Halev, and Xin Liu. Policy learning with constraints in model-free rein- forcement learning: A survey. In IJCAI,

2021
[23]

Effect estimates in randomized tri- als and observational studies: comparing apples with ap- ples

[Lodi et al., 2019] Sara Lodi, Andrew Phillips, Jens Lund- gren, and et al. Effect estimates in randomized tri- als and observational studies: comparing apples with ap- ples. American journal of epidemiology , 188(8):1569– 1577,

2019
[24]

Off-policy learning in two-stage recommender systems

[Ma et al., 2020] Jiaqi Ma, Zhe Zhao, Xinyang Yi, and et al. Off-policy learning in two-stage recommender systems. In WWW,

2020
[25]

Optimal dynamic treat- ment regimes

[Murphy, 2003] Susan A Murphy. Optimal dynamic treat- ment regimes. Journal of the Royal Statistical Society Se- ries B: Statistical Methodology, 65(2):331–355,

2003
[29]

Communities and crime

[Redmond, 2009] Michael Redmond. Communities and crime. UCI Machine Learning Repository, doi.org/10. 24432/C53W3X,

2009
[33]

The use of propensity scores to assess the generalizability of results from randomized trials

[Stuart et al., 2011] Elizabeth A Stuart, Stephen R Cole, Catherine P Bradshaw, and Philip J Leaf. The use of propensity scores to assess the generalizability of results from randomized trials. Journal of the Royal Statistical Society Series A: Statistics in Society , 174(2):369–386,

2011
[35]

Kar- vanen

[Tikka et al., 2019] Santtu Tikka, Antti Hyttinen, and J. Kar- vanen. Causal effect identification from multiple in- complete data sources: A general search-based approach. arXiv:1902.01073,

2019 arXiv
[36]

Semiparametric theory and missing data, volume

[Tsiatis, 2006] Anastasios A Tsiatis. Semiparametric theory and missing data, volume

2006
[38]

Deep transfer learn- ing for crop yield prediction with remote sensing data

[Wang et al., 2018] Anna X Wang, Caelin Tran, Nikhil De- sai, David Lobell, and Stefano Ermon. Deep transfer learn- ing for crop yield prediction with remote sensing data. In SIGCAS,

2018
[39]

A novel approach combined transfer learning and deep learning to predict tmb from histology image

[Wang et al., 2020] Liansheng Wang, Yudi Jiao, Ying Qiao, Nianyin Zeng, and Rongshan Yu. A novel approach combined transfer learning and deep learning to predict tmb from histology image. Pattern Recognition Letters , 135:244–248,

2020
[40]

The promises of multiple experiments: Identifying joint distri- bution of potential outcomes

[Wu and Mao, 2025] Peng Wu and Xiaojie Mao. The promises of multiple experiments: Identifying joint distri- bution of potential outcomes. arXiv:2504.20470,

2025 arXiv
[42]

Quantifying individual risk for binary outcome: Bounds and inference

[Wu et al., 2024a] Peng Wu, Peng Ding, Zhi Geng, and Yue Li. Quantifying individual risk for binary outcome: Bounds and inference. arXiv:2402.10537,

arXiv
[43]

On the comparative analysis of average treatment effects estimation via data combination

[Wu et al., 2025] Peng Wu, Shanshan Luo, and Zhi Geng. On the comparative analysis of average treatment effects estimation via data combination. Journal of the American Statistical Association,

2025
[44]

Learning the optimal policy for balancing short-term and long-term re- wards

[Yang et al., 2024] Qinwei Yang, Xueqing Liu, Yan Zeng, Ruocheng Guo, Yang Liu, and Peng Wu. Learning the optimal policy for balancing short-term and long-term re- wards. NeurIPS,

2024
[45]

A comprehensive survey on transfer learning

[Zhuang et al., 2020] Fuzhen Zhuang, Zhiyuan Qi, Keyu Duan, Dongbo Xi, Yongchun Zhu, Hengshu Zhu, Hui Xiong, and Qing He. A comprehensive survey on transfer learning. Proceedings of the IEEE, 109(1):43–76, 2020

2020
[1974]

On the ap- plication of probability theory to agricultural experiments

[Splawa-Neyman, 1990] Jerzy Splawa-Neyman. On the ap- plication of probability theory to agricultural experiments. essay on. Statistical Science, 5(4,465-480),

1990
[1983]

Estimating causal effects of treatments in randomized and nonrandomized studies

[Rubin, 1974] Donald B Rubin. Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 66(5):688,

1974
[1990]

Causal diagrams for empirical re- search

[Pearl, 1995] Judea Pearl. Causal diagrams for empirical re- search. Biometrika, 82(4):669–688,

1995
[1995]

Transfer learning enables the molecular transformer to predict regio-and stereoselective reactions on carbohy- drates

[Pesciullesi et al., 2020] Giorgio Pesciullesi, Philippe Schwaller, Teodoro Laino, and Jean-Louis Reymond. Transfer learning enables the molecular transformer to predict regio-and stereoselective reactions on carbohy- drates. Nature Communications, 11(1):4874,

2020
[2003]

Semiparametric effi- ciency bounds

[Newey, 1990] Whitney K Newey. Semiparametric effi- ciency bounds. Journal of Applied Econometrics, 5(2):99– 135,

1990
[2005]

[Hu et al., 2023] W. Hu, X.H. Zhou, and P. Wu. Identifica- tion and estimation of treatment effects on long-term out- comes in clinical trials with external observational data. Statistica Sinica,

2023
[2006]

Off-policy evaluation and learning for exter- nal validity under a covariate shift

[Uehara et al., 2020] Masatoshi Uehara, Masahiro Kato, and Shota Yasui. Off-policy evaluation and learning for exter- nal validity under a covariate shift. NeurIPS, 33:49–61,

2020
[2009]

[Rosenbaum and Rubin, 1983] P. R. Rosenbaum and D. B. Rubin. The central role of the propensity score in observa- tional studies for causal. Biometric, 70:41–55,

1983
[2011]

Scalability in perception for au- tonomous driving: Waymo open dataset

[Sun et al., 2020] Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, and et al. Scalability in perception for au- tonomous driving: Waymo open dataset. In CVPR,

2020
[2014]

[Hartman et al., 2015] Erin Hartman, Richard Grieve, Roland Ramsahai, and Jasjeet S Sekhon. From sample average treatment effect to population average treatment effect on the treated: combining experimental with observational studies to estimate population treatment effects. J...

2015
[2015]

Hern ´an and J

[Hern´an and Robins, 2020] M.A. Hern ´an and J. M. Robins. Causal Inference: What If . Boca Raton: Chapman and Hall/CRC,

2020
[2016]

Who should be treated? empirical welfare max- imization methods for treatment choice

[Kitagawa and Tetenov, 2018] Toru Kitagawa and Aleksey Tetenov. Who should be treated? empirical welfare max- imization methods for treatment choice. Econometrica, 86(2):591–616,

2018
[2017]

Generalizing evi- dence from randomized trials using inverse probability of sampling weights

[Buchanan et al., 2018] Ashley L Buchanan, Michael G Hudgens, Stephen R Cole, and et al. Generalizing evi- dence from randomized trials using inverse probability of sampling weights. Journal of the Royal Statistical Society Series A: Statistics in Society, 181(4):1193–1209,

2018
[2018]

Knowledge-based conversational recommender systems enhanced by dialogue policy learning

[Chen and Sun, 2021] Keyu Chen and Shiliang Sun. Knowledge-based conversational recommender systems enhanced by dialogue policy learning. In IJCKG, pages 10–18,

2021
[2019]

A survey on recent ad- vances and challenges in reinforcement learning methods for task-oriented dialogue policy learning

[Kwan et al., 2023] Wai-Chung Kwan, Hong-Ru Wang, Hui- Min Wang, and Kam-Fai Wong. A survey on recent ad- vances and challenges in reinforcement learning methods for task-oriented dialogue policy learning. Machine Intel- ligence Research, 20(3):318–334,

2023
[2020]

A survey on concept drift adaptation.ACM Computing Surveys, 46(4):1–37,

[Gama et al., 2014] Jo˜ao Gama, Indr ˙e ˇZliobait˙e, Albert Bifet, and et al. A survey on concept drift adaptation.ACM Computing Surveys, 46(4):1–37,

2014
[2021]

Per- sonalized diabetes management using electronic medical records

[Bertsimas et al., 2017] Dimitris Bertsimas, Nathan Kallus, Alexander M Weinstein, and Ying Daisy Zhuo. Per- sonalized diabetes management using electronic medical records. Diabetes Care, 40(2):210–217,

2017
[2022]

Long-term causal inference under persistent confounding via data combination

[Imbens et al., 2024] Guido Imbens, Nathan Kallus, Xiaojie Mao, and Yuhao Wang. Long-term causal inference under persistent confounding via data combination. Journal of the Royal Statistical Society Series B: Statistical Method- ology,

2024
[2023]

Semi-supervised learning for medical image classification using imbalanced training data

[Huynh et al., 2022] Tri Huynh, Aiden Nibali, and Zhen He. Semi-supervised learning for medical image classification using imbalanced training data. Computer methods and programs in biomedicine, 216:106628,

2022
[2024]

Generalized adjustment under confounding and se- lection biases

[Correa et al., 2018] Juan Correa, Jin Tian, and Elias Barein- boim. Generalized adjustment under confounding and se- lection biases. In AAAI, pages 6355–6342,

2018
[2025]

On the opportunity of causal learn- ing in recommendation systems: Foundation, estimation, prediction and challenges

[Wu et al., 2022] Peng Wu, Haoxuan Li, Yuhao Deng, Wen- jie Hu, Quanyu Dai, Zhenhua Dong, Jie Sun, Rui Zhang, and Xiao-Hua Zhou. On the opportunity of causal learn- ing in recommendation systems: Foundation, estimation, prediction and challenges. In IJCAI,

2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Policy learning with observational data

[Athey and Wager, 2021] Susan Athey and Stefan Wager. Policy learning with observational data. Econometrica, 89(1):133–161,

work page 2021

[2] [5]

Chernozhukov, D

[Chernozhukov et al., 2018] V . Chernozhukov, D. Chetverikov, M. Demirer, and et al. Double/de- biased machine learning for treatment and structural parameters. The Econometrics Journal, 21,

work page 2018

[3] [6]

Causal inference methods for combin- ing randomized trials and observational studies: a review

[Colnet et al., 2024] B´en´edicte Colnet, Imke Mayer, Guan- hua Chen, and et al. Causal inference methods for combin- ing randomized trials and observational studies: a review. Statistical Science, 39(1):165–191,

work page 2024

[4] [8]

Dahabreh, Sarah E

[Dahabreh et al., 2020] Issa J. Dahabreh, Sarah E. Robert- son, Jon A. Steingrimsson, Elizabeth A. Stuart, and Miguel A. Hern´an. Extending inferences from a random- ized trial to a new target population.Statistics in Medicine, 39:1999–2014,

work page 2020

[5] [12]

Predicting the efficacy of future training programs using past experiences at other locations

[Hotz et al., 2005] V Joseph Hotz, Guido W Imbens, and Julie H Mortimer. Predicting the efficacy of future training programs using past experiences at other locations. Jour- nal of Econometrics, 125(1-2):241–270,

work page 2005

[6] [16]

On the role of surrogates in the efficient estimation of treat- ment effects with limited outcome data

[Kallus and Mao, 2024] Nathan Kallus and Xiaojie Mao. On the role of surrogates in the efficient estimation of treat- ment effects with limited outcome data. Journal of the Royal Statistical Society Series B: Statistical Methodol- ogy, page qkae099,

work page 2024

[7] [18]

Transfer learning techniques for medical image analysis: A review

[Kora et al., 2022] Padmavathi Kora, Chui Ping Ooi, Oliver Faust, and et al. Transfer learning techniques for medical image analysis: A review. Biocybernetics and Biomedical Engineering, 42(1):79–107,

work page 2022

[8] [19]

Kosorok and Eric B

[Kosorok and Laber, 2019] Michael R. Kosorok and Eric B. Laber. Precision medicine. Annual Review of Statistics and Its Application, 6:263–86,

work page 2019