Pith. sign in

REVIEW 3 major objections 5 minor 69 references

Leveraging a Simulator for Learning Causal Representations from Post-Treatment Covariates for CATE

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

Pith's one-line read Post-treatment covariates make CATE unidentifiable; a simulator helps only up to its mismatch with reality.

desk verdict Useful empirical study of simulator-assisted CATE, but Lemma 4's bound does not follow from the proof as written. read the letter →

arxiv 2502.05037 v1 pith:YXCNAUSW submitted 2025-02-07 cs.LG

classification cs.LG MSC 62D20
keywords CATEestimationpost-treatmentcovariatessimulator-basedcounterfactualsgeneralizationboundcontrastivelearningcausalrepresentationtreatmenteffectidentifiabilityreal-simulatordistributiongap
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 addresses a setting where covariates and outcomes are both collected after treatment, which makes the Conditional Average Treatment Effect (CATE) unidentifiable from observational data alone. Its central claim is that a simulator supplying synthetic counterfactual pairs can restore the ability to estimate CATE, but only to the degree the simulator matches the real world; the paper proves a generalization bound that splits CATE error into a minimizable training part and an irreducible real-simulator mismatch part. Recognizing that mismatch is what makes the approach distinct: instead of trusting simulator counterfactuals wholesale, it builds an estimator, SimPONet, whose loss is exactly the minimizable portion of the bound, and whose weight on simulator representations is lowered when the simulator's factual predictions fall short. If the bound holds, the practical consequence is a principled recipe for when and how much to trust an off-the-shelf simulator when estimating treatment effects.

What carries the argument

The argument is carried by Lemma 4's two-bracket bound and the three distances it is built from: $d_{x|t}(f,g)$ is the expected squared distance between two representation extractors under the real treated-covariate distribution, $d_z(\tau,\tau^S)$ is the expected squared difference between real and simulated treatment-effect functions over the latent distribution $P_Z$, and $d_h$ is the same effect-distance composed with the rotation $h$ that contrastive learning leaves undetermined. The proof chain runs: contrastive learning on the simulator's paired counterfactual covariates recovers the simulator's inverse map $f^S$ up to a diffeomorphism $h$ (Lemma 3, building on Von Kügelgen et al.); on the unit hypersphere $h$ is a rotation (Zimmermann et al.), whose isometry lets the proof trade $d_{x|t}(h\circ f_t, h\circ f^S_t)$ for $d_{x|t}(f_t, f^S_t)$; and a $(a+b+c)^2 \le 3(a^2+b^2+c^2)$ expansion then separates factual error, simulator-reachable terms, and simulator-specific constants. SimPONet's loss (Eq. 3) is the direct instantiation of the minimizable bracket: a factual regression on real data, an $\ell^2$ pull of $\hat f_t$ toward the contrastively learned $\tilde f^S_t$, and an effect-consistency loss on the simulator's counterfactual outcome differences, with the representation weight $\lambda_f$ dropped when the simulator's representations give worse factual predictions than raw covariates.

What would settle it

With the linear DGP of Section 5.2, every quantity in Lemma 4 can be computed in closed form, so one could evaluate the claimed right-hand side of the bound and compare it against the realized CATE error of the jointly trained estimator. If the error exceeds the bound under strong treatment dependence on the latent $Z$, where $P(z\mid t)$ deviates most from $P_Z$, then the marginalization step in the proof of Lemma 4 fails and the bound as stated does not follow from the derivation.

Watch

Extended reading notes

Core claim

The paper's central result is Lemma 4, a bound saying that the CATE error of a jointly trained estimator is at most $8E^t_F + 12 d_h(\hat\tau, \tilde\tau^S) + 12K_\tau^2 d_{x|t}(\hat f_t, \tilde f^S_t)$ plus an irreducible bracket $12 d_z(\tau, \tau^S) + 12K_\tau^2 d_{x|t}(f_t, f^S_t)$ fixed by the gap between the real and simulated DGPs. The second bracket cannot be minimized by any training procedure; the first bracket can. The authors then show that three natural estimators each fail exactly when one of the two DGPs is taken at face value: SimOnly trusts the simulator entirely and fails when it diverges from reality, RealOnly ignores the simulator and works only if treatment leaves covariates unchanged, and Real$\mu$Sim$f$ imports simulator representations wholesale and fails when covariate generation differs. SimPONet, whose loss instantiates the minimizable terms, combines real factual supervision with simulator representation and effect regularizers, and the experiments across linear, Gaussian-process, IHDP, and ACIC settings show it matching or beating the better of its two data sources while degrading gracefully as the simulator gap grows.

Load-bearing premise

The paper's bound is derived by replacing the treatment-conditioned latent distribution $P(z\mid t)$ with the marginal $P_Z$ in the distance terms; this is only valid if treatment assignment is independent of the latent $Z$, which contradicts the paper's own data-generating process where $T$ depends on $Z$, unless one additionally assumes the likelihood ratio $P_Z/P(z\mid t)$ is bounded.

Editorial extensions

If this is right

  • If Lemma 4 holds, a practitioner can blend real and simulated data without knowing in advance which is better: the estimator's error is bounded by a fixed simulator gap plus training terms the loss can drive down, so SimPONet never degrades dramatically relative to the weaker of its two data sources.
  • Each of the three simpler baselines collapses in a specific identifiable regime (SimOnly when the simulator does not match reality, RealOnly when treatment changes covariates, Real$\mu$Sim$f$ when covariate generation differs), whereas the joint objective avoids all three failure modes.
  • The $\tau^S$ regularizer is load-bearing: without it, the representation pull alone drives $\hat f_t$ to $\tilde f^S_t$ and SimPONet degenerates into Real$\mu$Sim$f$, so the simulator's treatment-effect signal, not just its representation signal, is what carries the improvement.
  • The factual-error comparison that sets $\lambda_f$ gives SimPONet a data-dependent switch: on data like IHDP, where the simulator's representations hurt factual prediction, the method automatically reduces their influence, which explains its stable performance across varied real-simulator gaps.
  • Limited real data is where the simulator pays off most: at 10 percent training size SimPONet's margin over baselines is largest, indicating that the simulator regularizers most effectively substitute for the missing real counterfactual supervision when the real sample is small.

Reading between the lines

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

  • The bound suggests an estimable diagnostic the authors do not develop: with a small labeled pool of real counterfactual pairs, one could measure $d_z(\tau,\tau^S)$ and $d_{x|t}(f_t,f^S_t)$ directly and decide up front whether a simulator is worth using at all.
  • The same two-bracket decomposition should transfer to other sources of synthetic counterfactuals, such as image or text counterfactual generators, whenever the covariate maps are approximately invertible; the paper's non-invertible MLP experiments suggest the diffeomorphism condition is not a hard practical barrier.
  • A testable refinement of the $\lambda_f$ heuristic would set it continuously as a decreasing function of the real-versus-simulator factual-error gap instead of the binary drop to $10^{-4}$, which the bound's $d_{x|t}(\hat f_t,\tilde f^S_t)$ term suggests should track representation quality.
  • Lemma 1's impossibility implies that any purely observational method for this setting must import external paired information somewhere; treating that information as a simulator with a measurable gap may serve as a template for other unidentifiable causal estimands.
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

3 major / 5 minor

Summary. The paper studies Conditional Average Treatment Effect (CATE) estimation in a setting where both covariates X and outcomes Y are observed after treatment T is assigned, with a latent representation Z generating X through treatment-specific diffeomorphic maps and with T depending on Z. It proves that CATE is not identifiable from post-treatment X alone, analyzes three baseline estimators (SimOnly, RealOnly, and RealµSimf), derives a generalization bound (Lemma 4) for a joint real/simulator training objective, and proposes the SimPONet method, whose loss combines factual regression on real data, a representation-matching regularizer, and a simulator-based CATE regularizer. Experiments on synthetic linear DGPs, semi-synthetic IHDP, and ACIC datasets compare SimPONet against CATENets baselines and report that SimPONet often achieves the lowest mean CATE error.

Significance. If the main theoretical claim were established, the paper would make a useful contribution: it would decompose CATE error under post-treatment covariates into a minimizable part (factual error and two distance terms to simulator estimates) and an irreducible real-simulator mismatch, thereby justifying principled use of imperfect simulators. The paper also gives a clean population-level analysis of three natural baselines, releases code, and provides a broad experimental comparison. However, the proof of Lemma 4 in Appendix A.5.2 contains a load-bearing distribution mismatch and an unsupported rotation step, so the theoretical foundation for SimPONet is not currently established. The empirical results are suggestive but are weakened by a manually adjusted hyperparameter for IHDP.

major comments (3)
  1. [Appendix A.5.2, proof of Lemma 4] After applying (a+b+c)^2 <= 3(a^2+b^2+c^2), the proof reaches the integral ∫_Z [τ(z)-hatτ(h(z))]^2 P(z|t) dz and relabels it as d_z(τ,hatτ∘h). But Definition 2 defines d_z as the expectation under the marginal PZ, not under P(z|t). Under the paper's own DGP (Section 3) with T ~ P(T|Z), these distributions differ: P(z|t) is proportional to P(T=t|z) PZ(z). Assumption A1 only ensures P(T=t|z)>0, not a bounded likelihood ratio P(z|t)/PZ(z); under strong selection this ratio can be arbitrarily large, so the stated bound can fail by an unbounded factor. The same substitution is made for the terms d_z(τ,τS) and d_h(hatτ,~τS). The proof should either define all distances conditionally on P(z|t) or add a bounded density-ratio assumption, e.g., sup_z P(z|t)/PZ(z) <= C, and multiply the affected terms by the corresponding constant.
  2. [Appendix A.5.2, final equality in the proof of Lemma 4] The proof asserts that d_{x|t}(h∘f_t, h∘f^S_t) = d_{x|t}(f_t, f^S_t) 'since h is a rotation transform'. Lemma 5 guarantees that h is a rotation only under the extra assumption Z⊂S^{n_z-1}, which is not stated in Lemma 4. Without that assumption h is only a diffeomorphism, and the equality is generally false. This step is needed to identify the 'irreducible' real-simulator mismatch term, so the lemma either needs an explicit sphere assumption or a Lipschitz bound on h in the statement of Lemma 4.
  3. [Section 5.3(a) and Algorithm 1] The paper claims SimPONet automatically adjusts the simulator's influence, but for IHDP the weight λ_f is set to 1e-4 'because contrastive learning struggled with these binary features', which is a manual, dataset-specific choice rather than the output of the decision rule in Algorithm 1. Please state the exact criterion (e.g., the threshold on the validation factual-error comparison) that led to this value, and report sensitivity to the default λ_f=1 for IHDP, or revise the description to acknowledge explicit tuning.
minor comments (5)
  1. [Section 2.2] The phrase 'Please refer (Pearl, 2015) to for a formal proof' contains a typo ('to' should be removed).
  2. [Section 5.2] The sentence 'We show the results comparing SimPONet with the three baselines in Table 13 where we observe: achieves either the best or second-best performance' has a missing subject and an incorrect table number; it should refer to Table 1.
  3. [Section 5.9] The sentence 'SimPONet outperforms achieves either comparable performance or manages to surpass the baseline methods' is redundant and should be reworded.
  4. [Definition 2 and Lemma 4] The notation d_h and d_h(z) is used inconsistently between Definition 2 and the proof of Lemma 4; please define the notation once and use it uniformly.
  5. [Table 9 (Symbols)] Several symbols in the table of notations, such as K_μ, K_{μS}, and K_{τS}, are never used in the paper, while K_τ is used without a matching Lipschitz assumption on h; please clean up the symbol table.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: Lemma 4 is a derived upper bound whose minimizable terms inspire the SimPONet loss; no target CATE quantity is fitted, and the load-bearing identifiability facts are external theorems.

full rationale

The derivation chain is self-contained in the relevant sense. Lemma 1 is a standard collider/unidentifiability argument. Lemma 3 and Lemma 5 recover the simulator's f^S and tau^S up to a diffeomorphism h, citing Von Kügelgen et al. (2021) and Zimmermann et al. (2021) as external sources, not the authors' own prior work. Lemma 4 then upper-bounds E^t_CATE by adding and subtracting simulator estimates and applying Lipschitz and (a+b+c)^2 inequalities; the second bracket is exactly the real-simulator mismatch terms d_z(tau,tau^S) and d_{x|t}(f_t,f^S_t) defined in Section 3, so the statement that CATE error grows with mismatch is a theorem consequence, not an empirical prediction fitted to the target. SimPONet's objective (Eq. 3) is the minimizable portion of that bound, and no parameter is tuned against real CATE labels: lambda_f is set from factual validation errors of RealOnly versus RealµSimf, and lambda_tau is fixed at 1. Self-citations (Nagalapatti et al. 2022; 2024b) appear as context or as an experimental baseline (PairNet), not as justification of the bound or the SimPONet objective. Appendix A.5.2 contains a genuine proof gap: the proof replaces the conditional P(z|t) with the marginal PZ when labeling the integrals as d_z and d_h, and it uses the rotation property of h without restating the Z in S^(nz-1) assumption; however, a proof gap is a correctness risk, not circularity, because the paper does not define a target quantity in terms of the bound or fit the bound to the target. Thus no circular step can be exhibited, and the score reflects only peripheral self-citations and the need to verify the Appendix proof.

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

The central claim rests on standard causal assumptions (positivity, diffeomorphism, sufficiency of Z), an external contrastive identifiability theorem, and one unflagged distributional identity (P(z|t)=PZ) inside the main proof. The method itself introduces no new physical or causal entities. The two loss weights are hyperparameters, one tuned per dataset and one fixed.

free parameters (2)
  • lambda_f = 1 or 1e-4
    Weight on the representation regularizer d(hat_f_t, tilde_f^S_t). Set to 1e-4 for IHDP because RealOnly's factual error was much lower than Real_mu_Sim_f's; otherwise default 1. This is a dataset-specific tuning based on validation factual error.
  • lambda_tau = 1
    Weight on the simulator treatment effect regularizer. Fixed to 1 because tuning would require counterfactual supervision on real data, which is unavailable. This is a chosen constant, not fitted to data.
assumptions (6)
  • domain assumption A1 Positivity: P(T=t|Z=z)>0 for all t and z.
    Standard overlap condition stated in Section 3 as required for identifying CATE.
  • domain assumption A2 Diffeomorphic covariate generation: g_t and g^S_t are diffeomorphisms.
    Stated in Section 3; needed for the theoretical identifiability of tau and for the change of variables in the proof of Lemma 4. The authors test robustness when this is violated in Section 5.6.
  • domain assumption A3 Identifiability of tau given Z: Z is a sufficient adjustment set blocking backdoor paths between T and Y.
    Stated in Section 3 as a weaker notion of unconfoundedness. It is the conceptual basis for recovering Z from X.
  • domain assumption Contrastive learning with paired counterfactual covariates recovers the latent Z up to a rotation (Theorem 4.4 of Von Kugelgen et al. 2021 and Proposition 2 of Zimmermann et al. 2021).
    This external theorem is invoked in Lemma 3 and Lemma 5 to justify that tilde_f^S = h o f^S and tilde_tau^S = tau^S o h^{-1}. The proof relies entirely on this cited result.
  • domain assumption The simulator DGP samples Z from the same distribution PZ as the real DGP.
    Stated in Section 3 under Simulator DGP. This ensures the latent space is shared, but it does not imply P(Z|T=t)=PZ.
  • ad hoc to paper Implicit assumption in Lemma 4 proof: P(Z|T=t)=PZ when replacing conditional integrals with dz and dh defined under PZ.
    In Appendix A.5.2 the proof writes integrals over P(z|t) and then directly substitutes dz(τ, hat_tau o h) and dh(hat_tau, tilde_tau^S), which are defined with the marginal PZ. No density ratio or independence assumption is stated. This contradicts the DGP where T depends on Z.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging a Simulator for Learning Causal Representations from Post-Treatment Covariates for CATE." pith.science (2026). https://pith.science/paper/YXCNAUSW

@misc{pith2026250205037,
  author       = {Pith},
  title        = {Pith review of: Leveraging a Simulator for Learning Causal Representations from Post-Treatment Covariates for CATE},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YXCNAUSW}},
  note         = {Machine review of arXiv:2502.05037}
}
read the original abstract

Treatment effect estimation involves assessing the impact of different treatments on individual outcomes. Current methods estimate Conditional Average Treatment Effect (CATE) using observational datasets where covariates are collected before treatment assignment and outcomes are observed afterward, under assumptions like positivity and unconfoundedness. In this paper, we address a scenario where both covariates and outcomes are gathered after treatment. We show that post-treatment covariates render CATE unidentifiable, and recovering CATE requires learning treatment-independent causal representations. Prior work shows that such representations can be learned through contrastive learning if counterfactual supervision is available in observational data. However, since counterfactuals are rare, other works have explored using simulators that offer synthetic counterfactual supervision. Our goal in this paper is to systematically analyze the role of simulators in estimating CATE. We analyze the CATE error of several baselines and highlight their limitations. We then establish a generalization bound that characterizes the CATE error from jointly training on real and simulated distributions, as a function of the real-simulator mismatch. Finally, we introduce SimPONet, a novel method whose loss function is inspired from our generalization bound. We further show how SimPONet adjusts the simulator's influence on the learning objective based on the simulator's relevance to the CATE task. We experiment with various DGPs, by systematically varying the real-simulator distribution gap to evaluate SimPONet's efficacy against state-of-the-art CATE baselines.

Figures

Figures reproduced from arXiv: 2502.05037 by the authors.

Figure 1
Figure 1. The Data Generating process for Real and Simulator. We present our setup in the top panel of [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Factual errors with p-values shown above bars. For IHDP, RealOnly consistently outperforms RealµSimf . We evaluated SimPONet against various baselines from the well-known CATENets (Curth et al., 2021), a benchmarking library for CATE estimation. Since the baseline methods are not designed to extract the causal representations, we provided them with representations extracted by simulated causal representation extract… view at source ↗
Figure 3
Figure 3. Comparing CATE errors under pre-treatment [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: We vary γτ , which controls the gap between the synthetic CATE, τ S , and the real CATE, τ . Each dataset is represented by a distinct color, where the pale version of the color indicates SimOnly and the darker version denotes SimPONet. For ACIC-7 and ACIC-26, as γτ in…
Figure 5
Figure 5. Figure 5: SimPONet’s model architecture. We present an overview of the SimPONet model architecture in [PITH_FULL_IMAGE:figures/full_fig_p020_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 48 canonical work pages

  1. [1]

    Estimating the labor market impact of voluntary military service using social security data on military applicants, 1995

    Joshua Angrist. Estimating the labor market impact of voluntary military service using social security data on military applicants, 1995

  2. [2]

    Estimating the effect of training programs on earnings

    Orley Ashenfelter. Estimating the effect of training programs on earnings. The Review of Economics and Statistics, pp.\ 47--57, 1978

  3. [3]

    Learning representations by maximizing mutual information across views

    Philip Bachman, R Devon Hjelm, and William Buchwalter. Learning representations by maximizing mutual information across views. Advances in neural information processing systems, 32, 2019

  4. [4]

    Controlling selection bias in causal inference

    Elias Bareinboim and Judea Pearl. Controlling selection bias in causal inference. In Artificial Intelligence and Statistics, pp.\ 100--108. PMLR, 2012

  5. [5]

    Recovering causal effects from selection bias

    Elias Bareinboim and Jin Tian. Recovering causal effects from selection bias. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015

  6. [6]

    D o C o G en: D omain counterfactual generation for low resource domain adaptation

    Nitay Calderon, Eyal Ben-David, Amir Feder, and Roi Reichart. D o C o G en: D omain counterfactual generation for low resource domain adaptation. In Smaranda Muresan, Preslav Nakov, and Aline Villavicencio (eds.), Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp.\ 7727--7746, Dublin, Irela...

  7. [7]

    Adversarial de-confounding in individualised treatment effects estimation

    Vinod K Chauhan, Soheila Molaei, Marzia Hoque Tania, Anshul Thakur, Tingting Zhu, and David A Clifton. Adversarial de-confounding in individualised treatment effects estimation. In International Conference on Artificial Intelligence and Statistics, pp.\ 837--849. PMLR, 2023

  8. [8]

    DISCO : Distilling counterfactuals with large language models

    Zeming Chen, Qiyue Gao, Antoine Bosselut, Ashish Sabharwal, and Kyle Richardson. DISCO : Distilling counterfactuals with large language models. In Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki (eds.), Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp.\ 5514--5528, Toronto, Canada, Jul...

Show all 69 references
  1. [9]

    Avoiding post-treatment bias in audit experiments

    Alexander Coppock. Avoiding post-treatment bias in audit experiments. Journal of Experimental Political Science, 6 0 (1): 0 1--4, 2019

  2. [10]

    Generalized adjustment under confounding and selection biases

    Juan Correa, Jin Tian, and Elias Bareinboim. Generalized adjustment under confounding and selection biases. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32, 2018

  3. [11]

    On inductive biases for heterogeneous treatment effect estimation

    Alicia Curth and Mihaela van der Schaar. On inductive biases for heterogeneous treatment effect estimation. Advances in Neural Information Processing Systems, 34: 0 15883--15894, 2021

  4. [12]

    In search of insights, not magic bullets: Towards demystification of the model selection dilemma in heterogeneous treatment effect estimation

    Alicia Curth and Mihaela van der Schaar. In search of insights, not magic bullets: Towards demystification of the model selection dilemma in heterogeneous treatment effect estimation. In International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii...

  5. [13]

    Really doing great at estimating cate? a critical look at ml benchmarking practices in treatment effect estimation

    Alicia Curth, David Svensson, Jim Weatherall, and Mihaela van der Schaar. Really doing great at estimating cate? a critical look at ml benchmarking practices in treatment effect estimation. In Thirty-fifth conference on neural information processing systems datasets and benchm...

  6. [14]

    Meal simulation model of the glucose-insulin system

    Chiara Dalla Man, Robert A Rizza, and Claudio Cobelli. Meal simulation model of the glucose-insulin system. IEEE Transactions on biomedical engineering, 54 0 (10): 0 1740--1749, 2007

  7. [15]

    Counterfactual mri generation with denoising diffusion models for interpretable alzheimer's disease effect detection

    Nikhil J Dhinagar, Sophia I Thomopoulos, Emily Laltoo, and Paul M Thompson. Counterfactual mri generation with denoising diffusion models for interpretable alzheimer's disease effect detection. bioRxiv, pp.\ 2024--02, 2024

  8. [16]

    Density estimation using real nvp

    Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. Density estimation using real nvp. arXiv preprint arXiv:1605.08803, 2016

  9. [17]

    CORE : A retrieve-then-edit framework for counterfactual data generation

    Tanay Dixit, Bhargavi Paranjape, Hannaneh Hajishirzi, and Luke Zettlemoyer. CORE : A retrieve-then-edit framework for counterfactual data generation. In Yoav Goldberg, Zornitsa Kozareva, and Yue Zhang (eds.), Findings of the Association for Computational Linguistics: EMNLP 202...

  10. [18]

    Minimax optimal nonparametric estimation of heterogeneous treatment e\ ffects

    Zijun Gao and Yanjun Han. Minimax optimal nonparametric estimation of heterogeneous treatment e\ ffects. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. L\ in (eds.), Advances in Neural Information Processing Systems, volume 33, pp.\ 21751--21762. Curran Associat...

  11. [19]

    On the transfer of inductive bias from simulation to the real world: a new disentanglement dataset

    Muhammad Waleed Gondal, Manuel Wuthrich, Djordje Miladinovic, Francesco Locatello, Martin Breidt, Valentin Volchkov, Joel Akpo, Olivier Bachem, Bernhard Sch \"o lkopf, and Stefan Bauer. On the transfer of inductive bias from simulation to the real world: a new disentanglement ...

  12. [20]

    Medjourney: Counterfactual medical image generation by instruction-learning from multimodal patient journeys

    Yu Gu, Jianwei Yang, Naoto Usuyama, Chunyuan Li, Sheng Zhang, Matthew P Lungren, Jianfeng Gao, and Hoifung Poon. Medjourney: Counterfactual medical image generation by instruction-learning from multimodal patient journeys. 2023

  13. [21]

    Counterfactual regression with importance sampling weights

    Negar Hassanpour and Russell Greiner. Counterfactual regression with importance sampling weights. In IJCAI, pp.\ 5880--5887, 2019 a

  14. [22]

    Learning disentangled representations for counterfactual regression

    Negar Hassanpour and Russell Greiner. Learning disentangled representations for counterfactual regression. In International Conference on Learning Representations, 2019 b

  15. [23]

    PEREIRA, and MARGIT TAVITS

    JONATHAN HOMOLA, MIGUEL M. PEREIRA, and MARGIT TAVITS. Fixed effects and post-treatment bias in legacy studies. American Political Science Review, 118 0 (1): 0 537–544, 2024. doi:10.1017/S0003055423001351

  16. [24]

    Extracting post-treatment covariates for heterogeneous treatment effect estimation

    Qiang Huang, Defu Cao, Yi Chang, and Yan Liu. Extracting post-treatment covariates for heterogeneous treatment effect estimation. 2023

  17. [25]

    Causal inference without balance checking: Coarsened exact matching

    Stefano M Iacus, Gary King, and Giuseppe Porro. Causal inference without balance checking: Coarsened exact matching. Political analysis, 20 0 (1): 0 1--24, 2012

  18. [26]

    Diffusion models for counterfactual explanations

    Guillaume Jeanneret, Lo \" c Simon, and Fr \'e d \'e ric Jurie. Diffusion models for counterfactual explanations. In Proceedings of the Asian Conference on Computer Vision, pp.\ 858--876, 2022

  19. [27]

    Learning causal effects via weighted empirical risk minimization

    Yonghan Jung, Jin Tian, and Elias Bareinboim. Learning causal effects via weighted empirical risk minimization. Advances in neural information processing systems, 33: 0 12697--12709, 2020

  20. [28]

    Deepmatch: Balancing deep covariate representations for causal inference using adversarial training

    Nathan Kallus. Deepmatch: Balancing deep covariate representations for causal inference using adversarial training. In International Conference on Machine Learning, pp.\ 5067--5077. PMLR, 2020

  21. [29]

    A survey on simulators for testing self-driving cars

    Prabhjot Kaur, Samira Taghavi, Zhaofeng Tian, and Weisong Shi. A survey on simulators for testing self-driving cars. In 2021 Fourth International Conference on Connected and Autonomous Driving (MetroCAD), pp.\ 62--70. IEEE, 2021

  22. [30]

    Towards optimal doubly robust estimation of heterogeneous causal effects

    Edward H Kennedy. Towards optimal doubly robust estimation of heterogeneous causal effects. arXiv preprint arXiv:2004.14497, 2020

  23. [31]

    A hard unsolved problem? post-treatment bias in big social science questions

    Gary King. A hard unsolved problem? post-treatment bias in big social science questions. In Hard Problems in Social Science” Symposium, April, volume 10, 2010

  24. [32]

    o ren R K \

    S \"o ren R K \"u nzel, Jasjeet S Sekhon, Peter J Bickel, and Bin Yu. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the national academy of sciences, 116 0 (10): 0 4156--4165, 2019

  25. [33]

    Explaining counterfactual images

    Oran Lang, Ilana Traynis, and Yun Liu. Explaining counterfactual images. Nature Biomedical Engineering, 2023. URL https://rdcu.be/dwVKK

  26. [34]

    Challenging common assumptions in the unsupervised learning of disentangled representations

    Francesco Locatello, Stefan Bauer, Mario Lucic, Gunnar Raetsch, Sylvain Gelly, Bernhard Sch \"o lkopf, and Olivier Bachem. Challenging common assumptions in the unsupervised learning of disentangled representations. In international conference on machine learning, pp.\ 4114--4...

  27. [35]

    a tsch, Bernhard Sch \

    Francesco Locatello, Michael Tschannen, Stefan Bauer, Gunnar R \"a tsch, Bernhard Sch \"o lkopf, and Olivier Bachem. Disentangling factors of variations using few labels, 2019 b . URL https://openreview.net/forum?id=SkGy6hjvPE

  28. [36]

    Generate your counterfactuals: Towards controlled counterfactual generation for text

    Nishtha Madaan, Inkit Padhi, Naveen Panwar, and Diptikalyan Saha. Generate your counterfactuals: Towards controlled counterfactual generation for text. In AAAI Conference on Artificial Intelligence, 2020. URL https://api.semanticscholar.org/CorpusID:228063841

  29. [37]

    Learning recourse on instance environment to enhance prediction accuracy

    Lokesh Nagalapatti, Guntakanti Sai Koushik, Abir De, and Sunita Sarawagi. Learning recourse on instance environment to enhance prediction accuracy. In Advances in Neural Information Processing Systems, 2022

  30. [38]

    Continuous treatment effect estimation using gradient interpolation and kernel smoothing

    Lokesh Nagalapatti, Akshay Iyer, Abir De, and Sunita Sarawagi. Continuous treatment effect estimation using gradient interpolation and kernel smoothing. Proceedings of the AAAI Conference on Artificial Intelligence, 38 0 (13): 0 14397--14404, Mar. 2024 a . doi:10.1609/aaai.v38...

  31. [39]

    Pairnet: Training with observed pairs to estimate individual treatment effect

    Lokesh Nagalapatti, Pranava Singhal, Avishek Ghosh, and Sunita Sarawagi. Pairnet: Training with observed pairs to estimate individual treatment effect. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024 . OpenReview.net, 2...

  32. [40]

    Vcnet and functional targeted regularization for learning causal effects of continuous treatments

    Lizhen Nie, Mao Ye, Qiang Liu, and Dan Nicolae. Vcnet and functional targeted regularization for learning causal effects of continuous treatments. arXiv preprint arXiv:2103.07861, 2021

  33. [41]

    Quasi-oracle estimation of heterogeneous treatment effects

    Xinkun Nie and Stefan Wager. Quasi-oracle estimation of heterogeneous treatment effects. Biometrika, 108 0 (2): 0 299--319, 2021

  34. [42]

    Adversarial balancing for causal inference

    Michal Ozery-Flato, Pierre Thodoroff, and Tal El-Hay. Adversarial balancing for causal inference. ArXiv, abs/1810.07406, 2018

  35. [43]

    Counterfactual image editing

    Yushu Pan and Elias Bareinboim. Counterfactual image editing. arXiv preprint arXiv:2403.09683, 2024

  36. [44]

    Deep structural causal models for tractable counterfactual inference

    Nick Pawlowski, Daniel Coelho de Castro, and Ben Glocker. Deep structural causal models for tractable counterfactual inference. Advances in neural information processing systems, 33: 0 857--869, 2020

  37. [45]

    Pearl and Cambridge University Press

    J. Pearl and Cambridge University Press. Causality: Models, Reasoning, and Inference. Cambridge University Press, 2000. ISBN 9780521773621. URL https://books.google.co.in/books?id=wnGU\_TsW3BQC

  38. [46]

    Conditioning on post-treatment variables

    Judea Pearl. Conditioning on post-treatment variables. Journal of Causal Inference, 3 0 (1): 0 131--137, 2015

  39. [47]

    Carl Edward Rasmussen and Christopher K. I. Williams. Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning). The MIT Press, 2005. ISBN 026218253X

  40. [48]

    Generating realistic natural language counterfactuals

    Marcel Robeer, Floris Bex, and Ad Feelders. Generating realistic natural language counterfactuals. In Marie-Francine Moens, Xuanjing Huang, Lucia Specia, and Scott Wen-tau Yih (eds.), Findings of the Association for Computational Linguistics: EMNLP 2021, pp.\ 3611--3625, Punta...

  41. [49]

    Estimation of regression coefficients when some regressors are not always observed

    James M Robins, Andrea Rotnitzky, and Lue Ping Zhao. Estimation of regression coefficients when some regressors are not always observed. Journal of the American statistical Association, 89 0 (427): 0 846--866, 1994

  42. [50]

    The central role of the propensity score in observational studies for causal effects

    Paul R Rosenbaum and Donald B Rubin. The central role of the propensity score in observational studies for causal effects. Biometrika, 70 0 (1): 0 41--55, 1983

  43. [51]

    Counterfactual generative networks

    Axel Sauer and Andreas Geiger. Counterfactual generative networks. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=BXewfAYMmJw

  44. [52]

    Perfect match: A simple method for learning representations for counterfactual inference with neural networks

    Patrick Schwab, Lorenz Linhardt, and Walter Karlen. Perfect match: A simple method for learning representations for counterfactual inference with neural networks. arXiv preprint arXiv:1810.00656, 2018

  45. [53]

    Learning counterfactual representations for estimating individual dose-response curves

    Patrick Schwab, Lorenz Linhardt, Stefan Bauer, Joachim M Buhmann, and Walter Karlen. Learning counterfactual representations for estimating individual dose-response curves. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pp.\ 5612--5619, 2020

  46. [54]

    Johansson, and David Sontag

    Uri Shalit, Fredrik D. Johansson, and David Sontag. Estimating individual treatment effect: generalization bounds and algorithms, 2016. URL https://arxiv.org/abs/1606.03976

  47. [55]

    Estimating individual treatment effect: generalization bounds and algorithms

    Uri Shalit, Fredrik D Johansson, and David Sontag. Estimating individual treatment effect: generalization bounds and algorithms. In International Conference on Machine Learning, pp.\ 3076--3085. PMLR, 2017

  48. [56]

    Adapting neural networks for the estimation of treatment effects

    Claudia Shi, David Blei, and Victor Veitch. Adapting neural networks for the estimation of treatment effects. Advances in neural information processing systems, 32, 2019

  49. [57]

    Matching methods for causal inference: A review and a look forward

    Elizabeth A Stuart. Matching methods for causal inference: A review and a look forward. Statistical science: a review journal of the Institute of Mathematical Statistics, 25 0 (1): 0 1, 2010

  50. [58]

    Designing counterfactual generators using deep model inversion

    Jayaraman Thiagarajan, Vivek Sivaraman Narayanaswamy, Deepta Rajan, Jia Liang, Akshay Chaudhari, and Andreas Spanias. Designing counterfactual generators using deep model inversion. Advances in Neural Information Processing Systems, 34: 0 16873--16884, 2021

  51. [59]

    u gelgen, Yash Sharma, Luigi Gresele, Wieland Brendel, Bernhard Sch \

    Julius Von K \"u gelgen, Yash Sharma, Luigi Gresele, Wieland Brendel, Bernhard Sch \"o lkopf, Michel Besserve, and Francesco Locatello. Self-supervised learning with data augmentations provably isolates content from style. Advances in neural information processing systems, 34:...

  52. [60]

    Wang, Natalia S

    Clinton J. Wang, Natalia S. Rost, and Polina Golland. Spatial-intensity transform gans for high fidelity medical image-to-image translation. In Anne L. Martel, Purang Abolmaesumi, Danail Stoyanov, Diana Mateus, Maria A. Zuluaga, S. Kevin Zhou, Daniel Racoceanu, and Leo Joskowi...

  53. [61]

    Optimal transport for treatment effect estimation

    Hao Wang, Jiajun Fan, Zhichao Chen, Haoxuan Li, Weiming Liu, Tianqiao Liu, Quanyu Dai, Yichao Wang, Zhenhua Dong, and Ruiming Tang. Optimal transport for treatment effect estimation. Advances in Neural Information Processing Systems, 36, 2024

  54. [62]

    Stable estimation of heterogeneous treatment effects

    Anpeng Wu, Kun Kuang, Ruoxuan Xiong, Bo Li, and Fei Wu. Stable estimation of heterogeneous treatment effects. In International Conference on Machine Learning, pp.\ 37496--37510. PMLR, 2023

  55. [63]

    Simglucose v0.2.1 (2018) [Online] , 2018

    Jinyu Xie. Simglucose v0.2.1 (2018) [Online] , 2018. URL https://github.com/jxx123/simglucose. Accessed on September 25, 2023

  56. [64]

    Representation learning for treatment effect estimation from observational data

    Liuyi Yao, Sheng Li, Yaliang Li, Mengdi Huai, Jing Gao, and Aidong Zhang. Representation learning for treatment effect estimation from observational data. Advances in Neural Information Processing Systems, 31, 2018

  57. [65]

    Ganite: Estimation of individualized treatment effects using generative adversarial nets

    Jinsung Yoon, James Jordon, and Mihaela Van Der Schaar. Ganite: Estimation of individualized treatment effects using generative adversarial nets. In International Conference on Learning Representations, 2018

  58. [66]

    Learning overlapping representations for the estimation of individualized treatment effects

    Yao Zhang, Alexis Bellot, and Mihaela Schaar. Learning overlapping representations for the estimation of individualized treatment effects. In International Conference on Artificial Intelligence and Statistics, pp.\ 1005--1014. PMLR, 2020

  59. [67]

    Lipton, Li Erran Li, and Eric P

    Yi-Fan Zhang, Hanlin Zhang, Zachary C. Lipton, Li Erran Li, and Eric P. Xing. Exploring transformer backbones for heterogeneous treatment effect estimation, 2022. URL https://arxiv.org/abs/2202.01336

  60. [68]

    Zimmermann, Yash Sharma, Steffen Schneider, Matthias Bethge, and Wieland Brendel

    Roland S. Zimmermann, Yash Sharma, Steffen Schneider, Matthias Bethge, and Wieland Brendel. Contrastive learning inverts the data generating process. 139: 0 12979--12990, 2021. URL http://proceedings.mlr.press/v139/zimmermann21a.html

  61. [69]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.