pith. sign in

arxiv: 2606.19594 · v1 · pith:NLDWFBHZnew · submitted 2026-06-17 · 💻 cs.LG

Unsupervised Causal Abstractions Discovery

Th\'eo Saulus , Simon Lacoste-Julien , Dhanya Sridhar This is my paper

Pith reviewed 2026-06-26 20:43 UTC · model grok-4.3

classification 💻 cs.LG
keywords causal abstractionslow-rank graphsstructural causal modelsunsupervised learningidentifiabilitycausal discoveryhigh-level models
0
0 comments X

The pith

Observations generated by a low-rank graph induce latents that form a causal abstraction.

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

The paper addresses the complementary problem to hypothesis testing for causal abstractions: learning a high-level structural causal model directly from low-level measurements. It shows that observations from a low-rank graph induce latents forming such an abstraction, supplies identifiability results for those latents, and gives a practical objective for recovering the high-level model. A sympathetic reader would care because this turns causal abstraction into an unsupervised discovery task rather than requiring an expert to propose and validate a candidate high-level model in advance.

Core claim

The paper claims that observations generated by a low-rank graph induce latents that form a causal abstraction, provides identifiability results about these latents, and proposes a practical objective to learn the high-level SCM from low-level measurements.

What carries the argument

The low-rank graph structure, which generates observations whose induced latents form a causal abstraction of the low-level system.

If this is right

  • High-level structural causal models can be learned directly from low-level data without expert-proposed candidates.
  • The induced latents are identifiable under the low-rank graph assumption.
  • A practical objective exists for recovering the high-level model in an unsupervised manner.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This approach might enable automated abstraction discovery in domains such as biology or social sciences where manual hypothesis generation is expensive.
  • Integration with neural representation learning could extend the method to high-dimensional data while preserving causal semantics.
  • Relaxing the low-rank condition to other structured graph families could broaden applicability if similar induction properties hold.

Load-bearing premise

The low-level system must be generated according to a low-rank graph structure.

What would settle it

Generating observations from a causal graph that lacks low-rank structure and checking whether the induced latents still form a valid causal abstraction would test the claim; failure to form an abstraction would support the necessity of the low-rank condition.

Figures

Figures reproduced from arXiv: 2606.19594 by Dhanya Sridhar, Simon Lacoste-Julien, Th\'eo Saulus.

Figure 1
Figure 1. Figure 1: Illustration of H (in red), L (in green), as well as the mechanisms FX (in blue) and FZ (in orange). where PaX ⊆ Z and PaZ ⊆ X . These mechanisms imply the following low-level mecha￾nisms over L = (X , U, FL) and high-level mechanisms over H = (Z, U, FH) when collapsing on either type of node (see [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of Boolean matrix factorization. [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
read the original abstract

Causal abstractions formalize when a high-level structural causal model (SCM) captures the interventional behavior of a lower-level SCM. Existing applications of this notion largely follow a hypothesis-testing paradigm: an expert proposes a candidate high-level model and then evaluates if the low-level system implements it. We study the complementary problem of learning a high-level model directly from low-level measurements. Our contributions leverage hypotheses from low-rank causal discovery, and can be summarized as follows: (1) we show that observations generated by a low-rank graph induce latents that form a causal abstraction, (2) we provide identifiability results about these latents, and (3) we propose a practical objective to learn this high-level SCM.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 1 minor

Summary. The manuscript addresses unsupervised discovery of causal abstractions. It claims that observations generated by a low-rank graph induce latents forming a causal abstraction of the low-level system, establishes identifiability results for these latents, and proposes a practical objective for learning the corresponding high-level SCM, all by leveraging hypotheses from low-rank causal discovery.

Significance. If the identifiability results hold under the stated low-rank assumption, the work would provide a data-driven route to high-level causal models without expert-proposed candidates, extending causal discovery into the abstraction setting. The explicit leverage of low-rank graph structure as a hypothesis (rather than a derived claim) is a clear strength.

minor comments (1)
  1. [Abstract] Abstract: the three listed contributions are stated at a high level; the main text should include an early, self-contained statement of the precise conditions under which the induced latents are identifiable (e.g., rank assumptions, observational vs. interventional data).

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of our work on unsupervised causal abstractions discovery and for recommending minor revision. We appreciate the recognition that our identifiability results, if they hold under the low-rank assumption, offer a data-driven approach to high-level causal models.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper's main claims rest on external hypotheses from low-rank causal discovery literature rather than any self-referential definition, fitted parameter renamed as prediction, or load-bearing self-citation chain. The abstract explicitly positions the low-rank graph structure as a leveraged assumption, not something derived or fitted within the paper. No equations or steps in the provided text reduce the claimed induction of causal abstractions or identifiability results to the inputs by construction. This is the normal case of a self-contained derivation against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are detailed beyond the high-level reliance on low-rank causal discovery hypotheses.

axioms (1)
  • domain assumption Low-level observations are generated by a low-rank graph structure
    Invoked to induce latents that form a causal abstraction (contribution 1)

pith-pipeline@v0.9.1-grok · 5645 in / 1079 out tokens · 24654 ms · 2026-06-26T20:43:34.590207+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

129 extracted references · 63 canonical work pages · 11 internal anchors

  1. [1]

    Algebraic Priors for Approximately Equivariant Networks

    Parameter-Free Approximate Equivariance for Tasks with Finite Group Symmetry , author =. 2025 , month = jun, number =. doi:10.48550/arXiv.2506.08244 , urldate =. arXiv , keywords =:2506.08244 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.08244 2025

  2. [2]

    Invariant Risk Minimization

    Arjovsky, Martin and Bottou, L. Invariant. 2020 , month = mar, number =. doi:10.48550/arXiv.1907.02893 , urldate =. arXiv , keywords =:1907.02893 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1907.02893 2020

  3. [3]

    Battiloro, Claudio and Karaismailo. E(n). 2025 , month = feb, number =. doi:10.48550/arXiv.2405.15429 , urldate =. arXiv , keywords =:2405.15429 , primaryclass =

    work page doi:10.48550/arxiv.2405.15429 2025

  4. [4]

    Proceedings of the aaai conference on artificial intelligence , volume=

    Abstracting causal models , author=. Proceedings of the aaai conference on artificial intelligence , volume=

  5. [5]

    Bengio, Yoshua and Deleu, Tristan and Rahaman, Nasim and Ke, Rosemary and Lachapelle, S. A. 2019 , month = feb, number =. doi:10.48550/arXiv.1901.10912 , urldate =. arXiv , keywords =:1901.10912 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1901.10912 2019

  6. [6]

    Brown, Noam and Lerer, Adam and Gross, Sam and Sandholm, Tuomas , year =. Deep. doi:10.48550/arXiv.1811.00164 , urldate =. arXiv , keywords =:1811.00164 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1811.00164

  7. [7]

    Learning

    Buchholz, Simon and Rajendran, Goutham and Rosenfeld, Elan and Aragam, Bryon and Sch. Learning

  8. [8]

    Cadei, Riccardo and Demirel, Ilker and Bartolomeis, Piersilvio De and Lindorfer, Lukas and Cremer, Sylvia and Schmid, Cordelia and Locatello, Francesco , year =. Causal. doi:10.48550/arXiv.2502.06343 , urldate =. arXiv , langid =:2502.06343 , primaryclass =

    work page doi:10.48550/arxiv.2502.06343

  9. [9]

    Causal Feature Learning: An Overview , shorttitle =

    Chalupka, Krzysztof and Eberhardt, Frederick and Perona, Pietro , year =. Causal Feature Learning: An Overview , shorttitle =. Behaviormetrika , volume =. doi:10.1007/s41237-016-0008-2 , urldate =

    work page doi:10.1007/s41237-016-0008-2

  10. [10]

    Artificial intelligence and statistics , pages=

    Multi-level cause-effect systems , author=. Artificial intelligence and statistics , pages=. 2016 , organization=

    2016

  11. [11]

    Exploring

    Chen, Siyi and Zhang, Huijie and Guo, Minzhe and Lu, Yifu and Wang, Peng and Qu, Qing , year =. Exploring. doi:10.48550/arXiv.2409.02374 , urldate =. arXiv , keywords =:2409.02374 , primaryclass =

    work page doi:10.48550/arxiv.2409.02374

  12. [12]

    Language

    Chen, Haolin and Feng, Yihao and Liu, Zuxin and Yao, Weiran and Prabhakar, Akshara and Heinecke, Shelby and Ho, Ricky and Mui, Phil and Savarese, Silvio and Xiong, Caiming and Wang, Huan , year =. Language. doi:10.48550/arXiv.2411.04282 , urldate =. arXiv , keywords =:2411.04282 , primaryclass =

    work page doi:10.48550/arxiv.2411.04282

  13. [13]

    Conreux, Louis and Soulaire, Tom , abstract =. The

  14. [14]

    Cui, Jingyi and Wen, Hongwei and Wang, Yisen , year =. An. doi:10.48550/arXiv.2505.22196 , urldate =. arXiv , keywords =:2505.22196 , primaryclass =

    work page doi:10.48550/arxiv.2505.22196

  15. [15]

    Cunningham, Hoagy and Ewart, Aidan and Riggs, Logan and Huben, Robert and Sharkey, Lee , year =. Sparse. doi:10.48550/arXiv.2309.08600 , urldate =. arXiv , keywords =:2309.08600 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2309.08600

  16. [16]

    , year =

    Daunhawer, Imant and Bizeul, Alice and Palumbo, Emanuele and Marx, Alexander and Vogt, Julia E. , year =. Identifiability. doi:10.48550/arXiv.2303.09166 , urldate =. arXiv , langid =:2303.09166 , primaryclass =

    work page doi:10.48550/arxiv.2303.09166

  17. [17]

    ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=

    Factorization of binary matrices: Rank relations, uniqueness and model selection of boolean decomposition , author=. ACM Transactions on Knowledge Discovery from Data (TKDD) , volume=. 2022 , publisher=

    2022

  18. [18]

    Probable

    Eastwood, Cian and Robey, Alexander and Singh, Shashank and von K. Probable. 2023 , month = aug, number =. doi:10.48550/arXiv.2207.09944 , urldate =. arXiv , keywords =:2207.09944 , primaryclass =

    work page doi:10.48550/arxiv.2207.09944 2023

  19. [19]

    2025 , month = may, journal =

    Learning to Estimate Sample-Specific Transcriptional Networks for 7,000 Tumors , author =. 2025 , month = may, journal =. doi:10.1073/pnas.2411930122 , urldate =

    work page doi:10.1073/pnas.2411930122 2025

  20. [20]

    Friedman, Dan and Dieng, Adji Bousso , abstract =. The

  21. [21]

    2021 , month = oct, number =

    Learning Disentangled Representations via Product Manifold Projection , author =. 2021 , month = oct, number =. doi:10.48550/arXiv.2103.01638 , urldate =. arXiv , keywords =:2103.01638 , primaryclass =

    work page doi:10.48550/arxiv.2103.01638 2021

  22. [22]

    G. On. 2022 , month = sep, urldate =

    2022

  23. [23]

    Gat, Itai and Remez, Tal and Shaul, Neta and Kreuk, Felix and Chen, Ricky T. Q. and Synnaeve, Gabriel and Adi, Yossi and Lipman, Yaron , year =. Discrete. doi:10.48550/arXiv.2407.15595 , urldate =. arXiv , keywords =:2407.15595 , primaryclass =

    work page doi:10.48550/arxiv.2407.15595

  24. [24]

    Advances in neural information processing systems , volume=

    Causal abstractions of neural networks , author=. Advances in neural information processing systems , volume=

  25. [25]

    Journal of Machine Learning Research , volume=

    Causal abstraction: A theoretical foundation for mechanistic interpretability , author=. Journal of Machine Learning Research , volume=

  26. [26]

    Causal Learning and Reasoning , pages=

    Finding alignments between interpretable causal variables and distributed neural representations , author=. Causal Learning and Reasoning , pages=. 2024 , organization=

    2024

  27. [27]

    Independent

    Ghosh, Shubhangi and Gresele, Luigi and von K. Independent. 2023 , month = dec, number =. doi:10.48550/arXiv.2312.13438 , urldate =. arXiv , langid =:2312.13438 , primaryclass =

    work page doi:10.48550/arxiv.2312.13438 2023

  28. [28]

    2022 , month = feb, number =

    Independent Mechanism Analysis, a New Concept? , author =. 2022 , month = feb, number =. doi:10.48550/arXiv.2106.05200 , urldate =. arXiv , langid =:2106.05200 , primaryclass =

    work page doi:10.48550/arxiv.2106.05200 2022

  29. [29]

    Leibo, and Yali Du

    Guo, Zihao and Willis, Richard and Shi, Shuqing and Tomilin, Tristan and Leibo, Joel Z. and Du, Yali , year =. doi:10.48550/arXiv.2503.14576 , urldate =. arXiv , keywords =:2503.14576 , primaryclass =

    work page doi:10.48550/arxiv.2503.14576

  30. [30]

    Towards a Definition of Disentangled Representations

    Higgins, Irina and Amos, David and Pfau, David and Racaniere, Sebastien and Matthey, Loic and Rezende, Danilo and Lerchner, Alexander , year =. Towards a. doi:10.48550/arXiv.1812.02230 , urldate =. arXiv , keywords =:1812.02230 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1812.02230

  31. [31]

    Amortizing intractable inference in large language models.arXiv preprint arXiv:2310.04363, 2023

    Amortizing Intractable Inference in Large Language Models , author =. 2024 , month = mar, number =. doi:10.48550/arXiv.2310.04363 , urldate =. arXiv , keywords =:2310.04363 , primaryclass =

    work page doi:10.48550/arxiv.2310.04363 2024

  32. [32]

    Huang, Wei and Han, Andi and Chen, Yongqiang and Cao, Yuan and Xu, Zhiqiang and Suzuki, Taiji , abstract =. On the

  33. [33]

    doi:10.48550/arXiv.2401.06781 , urldate =

    Huang, Chenghao and Cao, Yanbo and Wen, Yinlong and Zhou, Tao and Zhang, Yanru , year =. doi:10.48550/arXiv.2401.06781 , urldate =. arXiv , langid =:2401.06781 , primaryclass =

    work page doi:10.48550/arxiv.2401.06781

  34. [34]

    2017 , month = apr, journal =

    Normality and Actual Causal Strength , author =. 2017 , month = apr, journal =. doi:10.1016/j.cognition.2017.01.010 , urldate =

    work page doi:10.1016/j.cognition.2017.01.010 2017

  35. [35]

    2025 , month = feb, number =

    What Is Causal about Causal Models and Representations? , author =. 2025 , month = feb, number =. doi:10.48550/arXiv.2501.19335 , urldate =. arXiv , langid =:2501.19335 , primaryclass =

    work page doi:10.48550/arxiv.2501.19335 2025

  36. [36]

    Kamb and S

    An Analytic Theory of Creativity in Convolutional Diffusion Models , author =. 2025 , month = jun, number =. doi:10.48550/arXiv.2412.20292 , urldate =. arXiv , langid =:2412.20292 , primaryclass =

    work page doi:10.48550/arxiv.2412.20292 2025

  37. [37]

    Learning to induce causal structure.arXiv preprint arXiv:2204.04875, 2022

    Ke, Nan Rosemary and Chiappa, Silvia and Wang, Jane and Goyal, Anirudh and Bornschein, Jorg and Rey, Melanie and Weber, Theophane and Botvinic, Matthew and Mozer, Michael and Rezende, Danilo Jimenez , year =. Learning to. doi:10.48550/arXiv.2204.04875 , urldate =. arXiv , keywords =:2204.04875 , primaryclass =

    work page doi:10.48550/arxiv.2204.04875

  38. [38]

    When and

    Kempf, Elias and Schrodi, Simon and Argus, Max and Brox, Thomas , year =. When and. doi:10.48550/arXiv.2502.09507 , urldate =. arXiv , keywords =:2502.09507 , primaryclass =

    work page doi:10.48550/arxiv.2502.09507

  39. [39]

    Regularizing

    Kim, Hyunsu and Lee, Hyungi and Yang, Hongseok and Lee, Juho , year =. Regularizing. doi:10.48550/arXiv.2306.00356 , urldate =. arXiv , keywords =:2306.00356 , primaryclass =

    work page doi:10.48550/arxiv.2306.00356

  40. [40]

    Krueger, David and Caballero, Ethan and Jacobsen, Joern-Henrik and Zhang, Amy and Binas, Jonathan and Zhang, Dinghuai and Priol, Remi Le and Courville, Aaron , year =. Out-of-. doi:10.48550/arXiv.2003.00688 , urldate =. arXiv , keywords =:2003.00688 , primaryclass =

    work page doi:10.48550/arxiv.2003.00688 2003

  41. [41]

    Synergies between

    Lachapelle, S. Synergies between. 2023 , month = jun, number =. doi:10.48550/arXiv.2211.14666 , urldate =. arXiv , keywords =:2211.14666 , primaryclass =

    work page doi:10.48550/arxiv.2211.14666 2023

  42. [42]

    Langlais, Pierre-Carl and Chizhov, Pavel and Nee, Mattia and Hinostroza, Carlos Rosas and Girard, Ir. Even

  43. [43]

    Learning to Estimate Sample-Specific Transcriptional Networks for 7,000 Tumors , doi =

  44. [44]

    Multi-agent Reinforcement Learning in Sequential Social Dilemmas

    Leibo, Joel Z. and Zambaldi, Vinicius and Lanctot, Marc and Marecki, Janusz and Graepel, Thore , year =. Multi-Agent. doi:10.48550/arXiv.1702.03037 , urldate =. arXiv , keywords =:1702.03037 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1702.03037

  45. [45]

    2024 , month = nov, number =

    Lei, Anson and Sch. 2024 , month = nov, number =. doi:10.48550/arXiv.2411.06890 , urldate =. arXiv , keywords =:2411.06890 , primaryclass =

    work page doi:10.48550/arxiv.2411.06890 2024

  46. [46]

    Equivariance via

    Lin, Yuchao and Helwig, Jacob and Gui, Shurui and Ji, Shuiwang , year =. Equivariance via. doi:10.48550/arXiv.2406.07598 , urldate =. arXiv , keywords =:2406.07598 , primaryclass =

    work page doi:10.48550/arxiv.2406.07598

  47. [47]

    Lipman, Yaron and Havasi, Marton and Holderrieth, Peter and Shaul, Neta and Le, Matt and Karrer, Brian and Chen, Ricky T. Q. and. Flow. 2024 , month = dec, number =. doi:10.48550/arXiv.2412.06264 , urldate =. arXiv , keywords =:2412.06264 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2412.06264 2024

  48. [48]

    Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations

    Locatello, Francesco and Bauer, Stefan and Lucic, Mario and R. Challenging. 2019 , month = jun, number =. doi:10.48550/arXiv.1811.12359 , urldate =. arXiv , langid =:1811.12359 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1811.12359 2019

  49. [49]

    Advances in Neural Information Processing Systems , volume=

    Large-scale differentiable causal discovery of factor graphs , author=. Advances in Neural Information Processing Systems , volume=

  50. [50]

    Amortized

    Lorch, Lars and Sussex, Scott and Rothfuss, Jonas and Krause, Andreas and Sch. Amortized. 2022 , month = dec, number =. doi:10.48550/arXiv.2205.12934 , urldate =. arXiv , keywords =:2205.12934 , primaryclass =

    work page doi:10.48550/arxiv.2205.12934 2022

  51. [51]

    Compositional

    Mahajan, Divyat and Pezeshki, Mohammad and Arnal, Charles and Mitliagkas, Ioannis and Ahuja, Kartik and Vincent, Pascal , year =. Compositional. doi:10.48550/arXiv.2410.06303 , urldate =. arXiv , langid =:2410.06303 , primaryclass =

    work page doi:10.48550/arxiv.2410.06303

  52. [52]

    Maurer, Andreas and Pontil, Massimiliano and. The. 2016 , month = mar, number =. doi:10.48550/arXiv.1505.06279 , urldate =. arXiv , keywords =:1505.06279 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1505.06279 2016

  53. [53]

    Signal Processing , volume=

    Boolean decomposition of binary matrices using a post-nonlinear mixture approach , author=. Signal Processing , volume=. 2021 , publisher=

    2021

  54. [54]

    Neller, Todd W and Lanctot, Marc , langid =. An

  55. [55]

    2020 , month = oct, number =

    Learning Explanations That Are Hard to Vary , author =. 2020 , month = oct, number =. doi:10.48550/arXiv.2009.00329 , urldate =. arXiv , langid =:2009.00329 , primaryclass =

    work page doi:10.48550/arxiv.2009.00329 2020

  56. [56]

    Emergence of

    Park, Core Francisco and Okawa, Maya and Lee, Andrew and Tanaka, Hidenori and Lubana, Ekdeep Singh , year =. Emergence of. doi:10.48550/arXiv.2406.19370 , urldate =. arXiv , keywords =:2406.19370 , primaryclass =

    work page doi:10.48550/arxiv.2406.19370

  57. [57]

    Suitability

    Pouget, Ang. Suitability. 2025 , month = may, number =. doi:10.48550/arXiv.2505.22356 , urldate =. arXiv , keywords =:2505.22356 , primaryclass =

    work page doi:10.48550/arxiv.2505.22356 2025

  58. [58]

    Priol, R. An. 2021 , month = feb, number =. doi:10.48550/arXiv.2005.09136 , urldate =. arXiv , langid =:2005.09136 , primaryclass =

    work page doi:10.48550/arxiv.2005.09136 2021

  59. [59]

    and Balestriero, Randall and Brendel, Wieland and Klindt, David , year =

    Reizinger, Patrik and Bizeul, Alice and Juhos, Attila and Vogt, Julia E. and Balestriero, Randall and Brendel, Wieland and Klindt, David , year =. Cross-. doi:10.48550/arXiv.2410.21869 , urldate =. arXiv , keywords =:2410.21869 , primaryclass =

    work page doi:10.48550/arxiv.2410.21869

  60. [60]

    Decision-

    Reyes, Emilio , abstract =. Decision-

  61. [61]

    , year =

    Roeder, Geoffrey and Metz, Luke and Kingma, Diederik P. , year =. On. doi:10.48550/arXiv.2007.00810 , urldate =. arXiv , keywords =:2007.00810 , primaryclass =

    work page doi:10.48550/arxiv.2007.00810 2007

  62. [62]

    Scherrer, Nino and Goyal, Anirudh and Bauer, Stefan and Bengio, Yoshua and Ke, Nan Rosemary , year =. On the. doi:10.48550/arXiv.2206.04620 , urldate =. arXiv , keywords =:2206.04620 , primaryclass =

    work page doi:10.48550/arxiv.2206.04620

  63. [63]

    Bridging the Inference Gap in

    Senellart, Agathe and Allassonni. Bridging the Inference Gap in. 2025 , month = feb, number =. doi:10.48550/arXiv.2502.03952 , urldate =. arXiv , keywords =:2502.03952 , primaryclass =

    work page doi:10.48550/arxiv.2502.03952 2025

  64. [64]

    Shafer, Glenn and Kogan, Alex and Spirtes, Peter , langid =. A

  65. [65]

    Directed

    Sim, Aaron and Wiatrak, Maciej and Brayne, Angus and Creed, P. Directed. 2021 , month = jun, number =. doi:10.48550/arXiv.2106.08678 , urldate =. arXiv , langid =:2106.08678 , primaryclass =

    work page doi:10.48550/arxiv.2106.08678 2021

  66. [66]

    and Knutins, Maksis and Ziyin, Liu and Geisz, Daniel and Fetterman, Abraham J

    Simon, James B. and Knutins, Maksis and Ziyin, Liu and Geisz, Daniel and Fetterman, Abraham J. and Albrecht, Joshua , year =. On the. doi:10.48550/arXiv.2303.15438 , urldate =. arXiv , langid =:2303.15438 , primaryclass =

    work page doi:10.48550/arxiv.2303.15438

  67. [67]

    Anderson and Sebe, Nicu and Welling, Max , year =

    Song, Yue and Keller, T. Anderson and Sebe, Nicu and Welling, Max , year =. Flow. doi:10.48550/arXiv.2309.13167 , urldate =. arXiv , keywords =:2309.13167 , primaryclass =

    work page doi:10.48550/arxiv.2309.13167

  68. [68]

    Soudry, Daniel and Hoffer, Elad and Nacson, Mor Shpigel and Gunasekar, Suriya and Srebro, Nathan , year =. The. doi:10.48550/arXiv.1710.10345 , urldate =. arXiv , keywords =:1710.10345 , primaryclass =

    work page doi:10.48550/arxiv.1710.10345

  69. [69]

    Squires, Chandler and Yun, Annie and Nichani, Eshaan and Agrawal, Raj and Uhler, Caroline , year =. Causal. doi:10.48550/arXiv.2207.01237 , urldate =. arXiv , keywords =:2207.01237 , primaryclass =

    work page doi:10.48550/arxiv.2207.01237

  70. [70]

    Identifying

    Syrota, Stas and Zainchkovskyy, Yevgen and Xi, Johnny and. Identifying. 2025 , month = may, number =. doi:10.48550/arXiv.2502.13757 , urldate =. arXiv , keywords =:2502.13757 , primaryclass =

    work page doi:10.48550/arxiv.2502.13757 2025

  71. [71]

    Disentangling

    Vafidis, Pantelis and Bhargava, Aman and Rangel, Antonio , year =. Disentangling. doi:10.48550/arXiv.2407.11249 , urldate =. arXiv , langid =:2407.11249 , primaryclass =

    work page doi:10.48550/arxiv.2407.11249

  72. [72]

    Wang, Qizhou and Lin, Yong and Chen, Yongqiang and Schmidt, Ludwig and Han, Bo and Zhang, Tong , year =. A. doi:10.48550/arXiv.2403.11497 , urldate =. arXiv , langid =:2403.11497 , primaryclass =

    work page doi:10.48550/arxiv.2403.11497

  73. [73]

    Wildberger, Jonas and Guo, Siyuan and Bhattacharyya, Arnab and Sch. On the. 2023 , month = feb, journal =

    2023

  74. [74]

    Propensity

    Xi, Johnny and Osea, Jana and Xu, Zuheng and Hartford, Jason , year =. Propensity. doi:10.48550/arXiv.2404.01595 , urldate =. arXiv , keywords =:2404.01595 , primaryclass =

    work page doi:10.48550/arxiv.2404.01595

  75. [75]

    Permutation

    Xu, Hengyuan and Xiang, Liyao and Ye, Hangyu and Yao, Dixi and Chu, Pengzhi and Li, Baochun , year =. Permutation. doi:10.48550/arXiv.2304.07735 , urldate =. arXiv , keywords =:2304.07735 , primaryclass =

    work page doi:10.48550/arxiv.2304.07735

  76. [77]

    Marrying

    Yao, Dingling and Muller, Caroline and Locatello, Francesco , year =. Marrying. doi:10.48550/arXiv.2405.13888 , urldate =. arXiv , langid =:2405.13888 , primaryclass =

    work page doi:10.48550/arxiv.2405.13888

  77. [78]

    Yao, Dingling and Xu, Danru and Lachapelle, S. Multi-. 2024 , month = mar, number =. doi:10.48550/arXiv.2311.04056 , urldate =. arXiv , langid =:2311.04056 , primaryclass =

    work page doi:10.48550/arxiv.2311.04056 2024

  78. [79]

    Yao, Dingling and Huang, Shimeng and Cadei, Riccardo and Zhang, Kun and Locatello, Francesco , year =. The. doi:10.48550/arXiv.2505.17708 , urldate =. arXiv , langid =:2505.17708 , primaryclass =

    work page doi:10.48550/arxiv.2505.17708

  79. [80]

    Unifying

    Yao, Dingling and Rancati, Dario and Cadei, Riccardo and Fumero, Marco and Locatello, Francesco , year =. Unifying. doi:10.48550/arXiv.2409.02772 , urldate =. arXiv , keywords =:2409.02772 , primaryclass =

    work page doi:10.48550/arxiv.2409.02772

  80. [81]

    Yu, Haizi and Mineyev, Igor , abstract =. A

Showing first 80 references.