Pith. sign in

REVIEW 5 major objections 5 minor 17 references

Distributionally Robust Policy Learning under Concept Drifts

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper shows that when distribution shift is restricted to the conditional reward law—concept drift—a debiased cross-fitted estimator of the worst-case policy value is asymptotically normal, and a policy maximizing it achieves the…

desk verdict Solid debiased estimation and a clean lower bound, but the minimax optimality headline does not follow from the paper's own entropy bound; likely repairable but needs a major fix. read the letter →

arxiv 2412.14297 v2 pith:ZSLXGDNT submitted 2024-12-18 cs.LG stat.ML

Jingyuan Wang , Zhimei Ren , Ruohan Zhan , Zhengyuan Zhou This is my paper · ORCID
classification cs.LGstat.ML MSC 62C2068T05
keywords distributionallyrobustoptimizationconceptdriftpolicylearningdoublyestimationofflinecontextualbanditsminimaxregretKLdivergencecross-fitting
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

The paper studies a sharper version of distributionally robust policy learning: instead of guarding against arbitrary shifts of the joint covariate-outcome distribution, it assumes the drift is confined to the conditional outcome law $P(Y \mid X)$—the concept-drift case—with the covariate distribution unchanged. Its central claim is that knowing the drift source is not just convenient but statistically optimal: the proposed policy has worst-case regret $O(\kappa(\Pi)/\sqrt{n})$, with $\kappa(\Pi)$ the Hamming entropy integral of the policy class, and a matching lower bound $\Omega(\sqrt{\mathrm{Ndim}(\Pi)/n})$ shows no learner can do better. To reach that rate the paper builds a doubly-robust, cross-fitted estimator of the worst-case policy value that is asymptotically normal even when nuisance functions converge slower than root-$n$, and a computational shortcut that estimates one nuisance function per action rather than per policy, making the search over an infinite policy class tractable. If the paper is right, practitioners facing concept drift can obtain the same sample efficiency as standard policy learning while hedging against worst-case conditional shifts, without the conservatism of joint-shift robust methods.

What carries the argument

The argument rests on three mechanisms. First, strong duality (Lemma 2.3) converts the worst-case conditional expectation under the KL uncertainty set into the minimization over $(\alpha, \eta)$ of the loss $\ell(x, y; \theta) = \alpha \exp(-(y+\eta)/\alpha - 1) + \eta + \alpha\delta$, so the robust value becomes the optimum of an empirical risk minimization problem. Second, a doubly-robust cross-fitted estimator—an importance-weighted residual $\hat{G} - \hat{g}$ plus a regression correction $\hat{g}$—removes the first-order bias of plugging in slowly converging nuisance estimates, with the four error terms (I)–(IV) vanishing at $o_P(n^{-1/2})$ under Assumption 3.4. Third, for policy learning, the observation that $\theta^*_{\pi(x)}(x) = \theta^*_a(x)$ for the chosen action $a$ reduces the infinite family of policy-dependent nuisances to $M$ action-dependent ones, allowing a chaining argument over the Hamming entropy integral $\kappa(\Pi)$ to bound the uniform estimation error and yield the regret rate, together with the matching Natarajan-dimension lower bound.

What would settle it

Generate data from the Section 5 simulation with a smooth conditional reward (satisfying Assumption B.1) but choose the sieve dimension so that $\hat{\theta}$ converges at exactly the $n^{-1/4}$ $L_2$ boundary; Theorem 3.5 predicts $\sqrt{n}(\hat{V}_\delta - V_\delta)$ is still asymptotically normal. If coverage of the resulting Wald intervals visibly degrades as $n$ grows, or if the empirical regret of Algorithm 2 acquires an extra logarithmic factor in $n$, the claimed double robustness and the rate would be contradicted.

Watch

Extended reading notes

Core claim

This paper establishes that robust policy learning under concept drift—where only the conditional reward law $P(Y \mid X)$ may shift while the covariate distribution stays fixed—admits a minimax-optimal learner for general covariate spaces. Concretely, Algorithm 2, which maximizes a debiased cross-fitted estimate of the KL-robust policy value $V_\delta(\pi)$, satisfies with probability at least $1 - \beta$ that $R_\delta(\hat{\pi}_{\mathrm{LN}}) \leq C_0(65 + 8\kappa(\Pi) + \sqrt{\log(1/\beta)})/\sqrt{n}$ (Theorem 4.3), where $\kappa(\Pi)$ is the Hamming entropy integral of the policy class; and no algorithm can have worst-case regret better than $\Omega(\sqrt{\mathrm{Ndim}(\Pi)/(n\epsilon)})$ for $\delta \leq 0.2$ (Theorem 4.6), so the rate is tight up to a log factor in the covariate dimension. On the evaluation side, the paper proves $\sqrt{n}(\hat{V}_\delta(\pi) - V_\delta(\pi)) \to N(0, \sigma_\pi^2)$ even when the nuisance parameters are estimated at slower-than-root-$n$ rates, as long as the cross-fitted rates of Assumption 3.4 are met.

Load-bearing premise

Everything rests on Assumption 3.4: the sieve-estimated $(\alpha, \eta)$ optimizer must converge to the truth in $L_2$ faster than $n^{-1/4}$ and uniformly in $L_\infty$, and the product of the propensity-score and conditional-mean estimation errors must be $o_P(n^{-1/2})$; if these rates slip, both the root-$n$ normality of the value estimator and the $\kappa(\Pi)/\sqrt{n}$ regret bound fail.

Editorial extensions

If this is right

  • Knowing the drift source shrinks the uncertainty set: the concept-drift worst-case value $V_\delta(\pi)$ is no smaller than the joint-shift worst-case value, so policies trained under concept drift provably dominate joint-shift robust policies when no covariate shift is present.
  • The regret upper bound $O(\kappa(\Pi)/\sqrt{n})$ and lower bound $\Omega(\sqrt{\mathrm{Ndim}(\Pi)/n})$ are matching in sample size and policy-class complexity, establishing minimax optimality for general covariate spaces.
  • The policy value estimator is doubly robust: it remains root-$n$ asymptotically normal when either the propensity score or the conditional-mean nuisance is consistent with product rate $n^{-1/2}$, and even when the $(\alpha, \eta)$ optimizer converges at $n^{-1/4}$ in $L_2$.
  • The same machinery extends to identifiable covariate shifts combined with concept drift: with density-ratio weights $r(X) = dQ_X/dP_X$, the estimator keeps asymptotic normality and the regret bound holds under the modified Assumption E.1.
  • Algorithm 2 is computationally feasible for infinite policy classes because the ERM nuisance is solved once per action rather than per policy, so policy search over $\Pi$ reduces to maximizing a single score.

Reading between the lines

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

  • If the paper's rate is right, the source-aware factorization principle likely extends beyond KL divergence: any divergence whose dual adds an $(\alpha, \eta)$ term to the loss should inherit the same debiasing recipe, making the machinery a template for other $f$-divergences.
  • The per-action shortcut $\theta^*_{\pi(x)}(x) = \theta^*_a(x)$ suggests a natural bridge to continuous action spaces—replacing the sum over $a \in [M]$ with an action-indexed model $\theta^*(x, a)$—which the paper leaves as a future direction but its own decomposition makes immediate.
  • Because the estimator's asymptotic variance $\sigma_\pi^2$ is given explicitly, one can construct Wald confidence intervals for $V_\delta(\pi)$ and use them for robust policy comparisons; the paper does not develop this inferential tool, but its variance formula makes it a direct byproduct.
  • Taking $\delta \to 0$, the robust value $V_\delta(\pi)$ converges to the plain policy value, so the regret bound should recover standard offline learning rates such as $\sqrt{\mathrm{VCdim}/n}$; checking this limit against known results would validate the constants.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper studies distributionally robust policy learning under concept drift, where only the conditional reward distribution Y|X is allowed to shift within a KL ball, while the covariate distribution is fixed. The authors derive a pointwise dual representation of the robust policy value, propose a cross-fitted, doubly robust estimator for V_delta(pi), prove root-n asymptotic normality under slower-than-root-n nuisance estimation rates, and then learn a policy by maximizing this estimated value over a class Pi. They state an upper bound of order kappa(Pi)/sqrt(n) on the sub-optimality gap and a lower bound of order sqrt(Ndim(Pi)/(n epsilon)), claiming minimax optimality. Numerical experiments compare the proposed Algorithm 2 with a joint-shift DRO benchmark on simulated and real data.

Significance. If the results are correct, this is a valuable contribution: it extends concept-drift-robust policy learning from the finite-covariate setting of Mu et al. (2022) to general covariate spaces, gives a doubly robust policy-value estimator with a clean rate condition, and provides both upper and lower regret bounds. The paper is also commendably explicit: the appendices contain the strong-duality derivation, the full regret decomposition, the Rademacher chaining argument, and a lower-bound construction, and the authors provide code for the main algorithm. However, the central minimax optimality claim is not actually supported by the theorems as written, because the stated relationship between kappa(Pi) and Ndim(Pi) does not imply a matching bound; this and several proof-level gaps require correction before the central claims are established.

major comments (5)
  1. [Section 4.3, Remark 4.7, Table 1] The claimed matching lower bound is not established by the displayed theorems. Theorem 4.3 gives R_delta(pi_hat^LN) = O((65 + 8 kappa(Pi) + sqrt(log(1/beta)))/sqrt(n)), while Theorem 4.6 gives Omega(sqrt(Ndim(Pi)/(n epsilon))). To compare these, one needs an upper bound on kappa(Pi) in terms of Ndim(Pi). Remark 4.5 and Table 1 state kappa(Pi) = O(sqrt(log d) Ndim(Pi)) and O(log(d) Ndim(Pi)), respectively; either relation would imply an upper bound O(Ndim(Pi)/sqrt(n)), which is strictly worse than the lower bound for Ndim(Pi) > 1. The standard metric-entropy bound log N_H(epsilon, Pi) <= C Ndim(Pi) log(1/epsilon) + O(Ndim(Pi) log M) would yield kappa(Pi) = O(sqrt(Ndim(Pi))) and would restore the matching claim, but this bound is neither stated nor proved in the paper. Please add the entropy lemma and use it explicitly in Remark 4.7.
  2. [Section 2.1, Lemma 2.3] The KL uncertainty set is not mathematically well-defined as written. Definition 2.2 defines KL divergence for distributions, but D_KL(Q_Y|X || P_Y|X) for conditional distributions is not the same object; it is either a pointwise quantity for each x or an averaged conditional KL. The proof of Lemma 2.3 in Eq. (11) uses the pointwise constraint D_KL(Q_{Y|X=x} || P_{Y|X=x}) <= delta with x-dependent multipliers alpha*(x), eta*(x). If the averaged conditional KL is intended, the dual with x-dependent alpha and eta is not valid. The paper should explicitly define the uncertainty set as {Q_Y|X : D_KL(Q_{Y|X=x} || P_{Y|X=x}) <= delta for P_X-almost every x}, since that is the formulation used in all subsequent derivations.
  3. [Appendix D.4, Theorem 4.6] The lower-bound construction does not satisfy Assumption 2.1 with the epsilon appearing in the theorem statement. The proof sets pi_0(f_1(x_j) | x_j) = pi_0(f_{-1}(x_j) | x_j) = epsilon/2, so the minimum propensity in the constructed instance is epsilon/2, not epsilon. If epsilon is the overlap lower bound in Assumption 2.1, these hard instances are outside the class P over which the supremum is taken. The construction must either use propensities at least epsilon and adjust the KL calculation, or the theorem must state the lower bound in terms of a separate parameter for the overlap of the hard instances. This is a load-bearing gap in the proof of the main lower bound.
  4. [Appendix D.3.2, Lemma D.2, Eq. (17)-(18)] There is an arithmetic inconsistency in the Rademacher bound constants. Lemma D.2's statement gives E_epsilon sup <= (sqrt(sum c_i^2)/n)(32 + 4 kappa(Pi)), but its proof concludes with (sqrt(sum c_i^2)/n)(64 + 8 kappa(Pi)) because Lambda = 2 sqrt(sum c_i^2). In the effective-term argument, Eq. (17) introduces an outer factor 2, then the text writes E_epsilon sup <= 2 C_0 (32 + 4 kappa)/sqrt(n), and Eq. (18) concludes with C_0(64 + 8 kappa + sqrt(log(1/beta)))/sqrt(n). These constants do not follow from one another. The rate is unaffected, but the proof of the constant 65 in Theorem 4.3 is not correct as written and should be repaired.
  5. [Assumption 3.4, Appendix D.2] The proofs of Terms (II) and (III) in the proof of Theorem 3.5 require a lower bound on the estimated propensity score. For example, the text says 'due to the lower bound on pi_0 and bpi_0' when bounding Term (II), and Term (III) uses factors of epsilon^{-4} that assume bpi_0 is bounded below by epsilon. Assumption 3.4 only imposes L2 convergence of bpi_0 to pi_0 and does not state any lower bound on bpi_0. Please add a clipping condition such as bpi_0 >= epsilon/2 (or an equivalent bounded-inverse-moment condition) to Assumption 3.4, or modify the variance bounds so that they do not require pointwise lower bounds on the estimator.
minor comments (5)
  1. [Section 5] There is a typo in 'the real-wrold dataset' that should read 'real-world dataset'.
  2. [Appendix D.4] In the lower-bound construction, the expression (1 - epsilon)/(K - 2) should use the number of actions M, not the number of folds K; K is not defined in that proof.
  3. [Appendix E, Theorem E.3] The statement of Theorem E.3 appears to be missing a division by sqrt(n); it should read C(kappa(Pi) + sqrt(log(1/beta)))/sqrt(n).
  4. [Table 1, Remark 4.5] The two stated relationships between kappa(Pi) and Ndim(Pi) are inconsistent with each other (O(log(d) Ndim) versus O(sqrt(log d) Ndim)). These should be unified after the correct entropy bound is added.
  5. [Section 2.1, Eq. (1)] The notation P_Y|X and Q_Y|X should be defined explicitly as Markov kernels, and the quantifier 'for P_X-almost every x' should be added to the KL constraint to remove the ambiguity noted in the major comments.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the upper bound is proved by chaining over Hamming balls and the lower bound by explicit hard instances; overlapping-author citations provide auxiliary lemmas, not the target results.

full rationale

The paper's central derivation chain is not circular. The robust policy value Vδ(π) is defined through a KL uncertainty set, and Lemma 2.3 obtains its dual representation by strong duality rather than by assuming the value estimator. The debiased estimator in Algorithm 1 is justified in Theorem 3.5 by a four-term decomposition (Terms (I)–(IV) in Appendix D.2), each shown to be oP(n−1/2) under the explicit rate conditions in Assumption 3.4; Assumption 3.4 is a stated input, not a restatement of the conclusion. The regret upper bound in Theorem 4.3 follows from the standard decomposition Rδ(π̂) ≤ 2 supπ∈Π |V̂δ(π)−Vδ(π)| and a Rademacher-chaining argument over Hamming balls, so κ(Π) enters as a complexity bound rather than as a fitted or renamed quantity. The lower bound in Theorem 4.6 is an independent minimax construction over Natarajan-shattered points with explicit Bernoulli hard instances; it does not invoke the upper bound and its constants are proof devices, not calibrated inputs. The paper does rely on prior work by overlapping authors (Jin et al. 2022b for Proposition 2.5 and sieve rates; Cai et al. 2020 for the κ–Natarajan relation), but these are parameter-free theoretical results whose assumptions do not include the present target claims, so under the review rules they constitute independent evidence rather than circularity. The apparent mismatch between the cited κ(Π)=O(Ndim(Π)) relation and the Ω(√Ndim) lower bound is a correctness or quantitative-support concern about the minimax-optimality claim, not a circular reduction: no displayed equation is identical to its input and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

No new particles, mediators, forces, or entities are introduced. The derivation relies on standard causal assumptions, a mild atomlessness condition, uniform bounds on dual optimizers, and double-machine-learning rate assumptions. The only user-chosen number is the robustness radius δ, which is an input to the problem rather than a fitted constant.

free parameters (2)
  • δ (KL uncertainty radius) = user-specified; 0.05, 0.1, 0.2 in simulations
    Controls the size of the concept-drift uncertainty set. It is not fitted to the data, but the paper itself notes in Section 5 that empirical performance depends substantially on this choice.
  • lower threshold for α in numerical ERM = 0.001
    Used to enforce Proposition 2.5 in the implementation with cubic splines. It is an implementation constant, not part of the main theoretical result.
assumptions (6)
  • domain assumption Assumption 2.1: unconfoundedness, overlap, and bounded reward support
    Standard causal inference assumptions in offline policy learning; they guarantee identifiability of policy values and valid importance weighting. Stated in Section 2.
  • domain assumption Assumption 2.4: no atom at the essential infimum of the reward conditional distribution
    Ensures the dual optimizer α* is strictly positive (Proposition 2.5). The paper argues it holds for all continuous distributions.
  • domain assumption Assumption 3.3: uniform bounds on the dual optimizers α* and η*
    Needed for boundedness of the loss and constants in the regret bound. The paper says it follows from continuity on compact X.
  • domain assumption Assumption 3.4: nuisance estimators converge at specified rates
    Double-machine-learning style rate conditions: product rate γ1+γ2≥1/2 for π0 and g, and bθ at n^{-1/4} in L2 plus L∞ consistency. Load-bearing for all debiasing terms.
  • domain assumption Assumption B.1: conditional reward distributions are smooth in x
    Used to justify the sieve convergence rates for the ERM optimizer, imported from Jin et al. (2022b).
  • standard math Strong duality for the KL-constrained inner problem
    Convex duality with Slater's condition, Theorem 8.6.1 of Luenberger (1997), invoked in Lemma 2.3 and Appendix D.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributionally Robust Policy Learning under Concept Drifts." pith.science (2026). https://pith.science/paper/ZSLXGDNT

@misc{pith2026241214297,
  author       = {Pith},
  title        = {Pith review of: Distributionally Robust Policy Learning under Concept Drifts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZSLXGDNT}},
  note         = {Machine review of arXiv:2412.14297}
}
abstract

Distributionally robust policy learning aims to find a policy that performs well under the worst-case distributional shift, and yet most existing methods for robust policy learning consider the worst-case joint distribution of the covariate and the outcome. The joint-modeling strategy can be unnecessarily conservative when we have more information on the source of distributional shifts. This paper studies a more nuanced problem -- robust policy learning under the concept drift, when only the conditional relationship between the outcome and the covariate changes. To this end, we first provide a doubly-robust estimator for evaluating the worst-case average reward of a given policy under a set of perturbed conditional distributions. We show that the policy value estimator enjoys asymptotic normality even if the nuisance parameters are estimated with a slower-than-root-$n$ rate. We then propose a learning algorithm that outputs the policy maximizing the estimated policy value within a given policy class $\Pi$, and show that the sub-optimality gap of the proposed algorithm is of the order $\kappa(\Pi)n^{-1/2}$, where $\kappa(\Pi)$ is the entropy integral of $\Pi$ under the Hamming distance and $n$ is the sample size. A matching lower bound is provided to show the optimality of the rate. The proposed methods are implemented and evaluated in numerical studies, demonstrating substantial improvement compared with existing benchmarks.

Figures

Figures reproduced from arXiv: 2412.14297 by the authors.

Figure 1
Figure 1. Empirical robust policy value V¯δ of policies πbLN, πbSNLN on the real-world dataset, over 50 seeds. Shading corresponds to 95% confidence intervals. being Gaussian, for a ∈ [3]. The training datasets Dtrain are generated with an unknown given behavior policy π0 over 50 random seeds. We similarly generate 100 testing datasets Dtest of size 10,000. The details are given in Appendix C. Real-world Dataset. We consider … view at source ↗
Figure 2
Figure 2. The Mean Square Error (MSE) of the estimated policy value by Algorithm 1. The x-axis is the number of samples used by Algorithm 1, and the y-axis is the mean squared error (MSE) of the policy value estimator. Computation Details. The experiments were run on the following cloud servers: (i) an Intel Xeon Platinum 8160 @ 2.1 GHz with 766GB RAM and 96 CPU x 2.1 GHz; (ii) an Intel Xeon Platinum 8160 @ 2.1 GHz with 1.5TB… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 9 canonical work pages

  1. [1]

    and Ren, Z

    Ai, J. and Ren, Z. Not all distributional shifts are equal: Fine-grained robust conformal inference. arXiv preprint arXiv:2402.13042,

    arXiv

  2. [10]

    Learning from a biased sample

    Sahoo, R., Lei, L., and Wager, S. Learning from a biased sample. arXiv preprint arXiv:2209.01754,

    arXiv

  3. [13]

    Notation We use [n] to denote the discrete set {1, 2, · · ·, n} for any n ∈ Z

    12 Distributionally Robust Policy Learning under Concept Drifts A. Notation We use [n] to denote the discrete set {1, 2, · · ·, n} for any n ∈ Z. We use argmin and argmax to denote the minimizers and maximizers; if the minimizer or the maximizer cannot be attained, we project it back to the feasible set. We denote the usual p-norm as ∥ · ∥p. Denote P to b...

    work page 1982

  4. [14]

    Computation Details

    The x-axis is the number of samples used by Algorithm 1, and the y-axis is the mean squared error (MSE) of the policy value estimator. Computation Details. The experiments were run on the following cloud servers: (i) an Intel Xeon Platinum 8160 @ 2.1 GHz with 766GB RAM and 96 CPU x 2.1 GHz; (ii) an Intel Xeon Platinum 8160 @ 2.1 GHz with 1.5TB RAM and 96 ...

    work page 1997

  5. [15]

    sup π∈Π 1 n X i∈[n] h(Zi, π(Xi)) − E h h(Zi, π(Xi)) i # = E

    By Equation (14), we have |h(Zi, π(Xi))| ≤6 ε · ¯α · exp ¯η α − 1 + ¯η + ¯αδ =: C0( ¯α, α, ¯η, δ, ε). 20 Distributionally Robust Policy Learning under Concept Drifts Meanwhile, we have write sup π∈Π 1 K KX k=1 − ˜V (k) δ (π) − Vδ(π) = sup π∈Π 1 K KX k=1 1 |D(k)| X i∈D(k) h Zi; π(Xi) = sup π∈Π 1 n nX i=1 h Zi; π(Xi) . Next, we define f (z1, . . . , zn; π) ...

    work page 2019

  6. [16]

    hard instances

    that will be used in the proof. For any p, q∈ [0, 1], define D(p ∥ q) = p log p q + (1 − p) log 1 − p 1 − q , and gδ(q) = inf p:DKL(p ∥ q)≤δ p, Lemma D.3 (Adapted from Lemma A17 of Si et al. (2023)) . For δ ≤ 0.2, gδ(q) is differentiable and g′ δ(q) ≥ 1/2 for q ∈ [0.4, 0.6]. Note that our definition of gδ(q) is slightly different from that in (Si et al., ...

    work page 2023

  7. [32]

    = 2 pPn i=1 ci(zi)2 n (4κ(Π) + 32). F.3. Proof of Lemma F.2 Fix γ >0. If NH (γ2, Π) = ∞, the lemma is trivially true. Otherwise, let N0 = NH (γ2; Π). For any realization z1, . . . , zn, define (π∗ i,1, π∗ i,2) = argmax π1,π2 |h(zi, π1(xi)) − h(zi, π2(xi))| . Implicitly, (π∗ i,1, π∗ i,2) depends on zi. For an arbitrary positive integer m and i ∈ [n], we de...

    work page 2022

  8. [1964]

    Diagnosing model performance under distribution shift

    Namkoong, H., Yadlowsky, S., et al. Diagnosing model performance under distribution shift. arXiv preprint arXiv:2303.02011,

    arXiv

Show all 17 references
  1. [1993]

    Distributionally robust policy evaluation under general covariate shift in contextual bandits

    Guo, Y ., Liu, H., Yue, Y ., and Liu, A. Distributionally robust policy evaluation under general covariate shift in contextual bandits. arXiv preprint arXiv:2401.11353 ,

    arXiv

  2. [1996]

    Bounds on the conditional and average treatment effect with unobserved confounding factors

    Yadlowsky, S., Namkoong, H., Basu, S., Duchi, J., and Tian, L. Bounds on the conditional and average treatment effect with unobserved confounding factors. arXiv preprint arXiv:1808.09521,

    arXiv

  3. [2012]

    S., and Lee, J

    Zhang, Z., Zhan, W., Chen, Y ., Du, S. S., and Lee, J. D. Optimal multi-distribution learning. arXiv preprint arXiv:2312.05134,

    arXiv

  4. [2013]

    Multi-action of- fline policy learning with bayesian optimization

    Cai, F., Qu, Z., Xia, L., and Zhou, Z. Multi-action of- fline policy learning with bayesian optimization. arXiv preprint arXiv:2003.07545v1,

    2003 arXiv

  5. [2016]

    Localized debiased ma- chine learning: Efficient inference on quantile treatment effects and beyond

    Kallus, N., Mao, X., and Uehara, M. Localized debiased ma- chine learning: Efficient inference on quantile treatment effects and beyond. arXiv preprint arXiv:1912.12945,

    1912 arXiv

  6. [2018]

    On the need for a language describing distribution shifts: Illustrations on tabular datasets

    Liu, J., Wang, T., Cui, P., and Namkoong, H. On the need for a language describing distribution shifts: Illustrations on tabular datasets. arXiv preprint arXiv:2307.05284 ,

    arXiv

  7. [2019]

    Doubly robust policy evaluation and learning

    Dudík, M., Langford, J., and Li, L. Doubly robust policy evaluation and learning. arXiv preprint arXiv:1103.4601,

    arXiv

  8. [2021]

    without

    Jin, Y ., Ren, Z., Yang, Z., and Wang, Z. Policy learning" without”overlap: Pessimism and generalized empirical bernstein’s inequality. arXiv preprint arXiv:2212.09900, 2022a. Jin, Y ., Ren, Z., and Zhou, Z. Sensitivity analysis under the f-sensitivity models: a distributional...

    arXiv

  9. [2023]

    and Udell, M

    Kallus, N. and Udell, M. Dynamic assortment per- sonalization in high dimensions. arXiv preprint arXiv:1610.05604,

    arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.