pith. machine review for the scientific record. sign in

arxiv: 2605.06288 · v1 · submitted 2026-05-07 · 📊 stat.ME · cs.AI

Recognition: unknown

A Topological Sorting Criterion for Random Causal Directed Acyclic Graphs

Alexander G. Reisach, Antoine Chambaz, Gilles Blanchard, Sebastian Weichwald

Pith reviewed 2026-05-08 07:23 UTC · model grok-4.3

classification 📊 stat.ME cs.AI
keywords causal discoverydirected acyclic graphsrandom graphstopological sortingMarkov equivalence classErdős-Rényiscale-free networkscausal order
0
0 comments X

The pith

In standard random causal DAGs, the number of relatives increases monotonically along the causal order.

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

The paper shows that DAGs generated by first imposing a topological order and then sampling edges from Erdős-Rényi or scale-free models have a built-in pattern: the set of nodes reachable via open paths, called relatives, grows larger the further one moves along the causal order. This monotonic growth appears frequently in the random graphs used to benchmark causal discovery methods. As a result, estimating the number of relatives from data and sorting nodes by that estimate often recovers the true causal order. When the increase is strict, the graph has a unique Markov equivalence class. The authors propose time-series DAG sampling as one way to generate alternatives that may lack this pattern.

Core claim

In DAGs generated by imposing an order on Erdős-Rényi and scale-free random graphs, the set of nodes reachable via open paths, termed relatives, increases monotonically along the causal order. Sorting by the estimated number of relatives recovers the causal order. A strict increase of relatives along the causal order leads to a singular Markov equivalence class. Sampling time-series DAGs is proposed as a possible alternative generation method.

What carries the argument

The set of relatives (nodes reachable via open paths) and its monotonic increase in count along the imposed causal order in these random DAGs.

If this is right

  • Sorting nodes by estimated number of relatives provides an effective proxy for recovering the causal order in many common simulation settings.
  • A strict monotonic increase in relatives implies the Markov equivalence class is singular.
  • The pattern is prevalent under standard procedures for generating random causal DAGs.
  • Time-series DAG sampling offers an alternative that may avoid this monotonic property.

Where Pith is reading between the lines

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

  • Causal discovery algorithms tested on these DAGs may benefit from the implicit ordering information carried by relative counts, affecting how performance is interpreted.
  • Simple sorting on estimated relatives could serve as a baseline comparator for more complex causal order methods.
  • Using generation procedures without this monotonicity would create stricter tests for causal discovery algorithms.

Load-bearing premise

The monotonic increase in relatives is a reliable property of standard random DAG generation by imposing order then sampling edges, and the number of relatives can be accurately estimated from finite observational data.

What would settle it

A generated random DAG from the Erdős-Rényi or scale-free procedure where the number of relatives does not increase monotonically along the imposed order, or data showing that sorting by estimated relatives fails to recover the order accurately.

Figures

Figures reproduced from arXiv: 2605.06288 by Alexander G. Reisach, Antoine Chambaz, Gilles Blanchard, Sebastian Weichwald.

Figure 1
Figure 1. Figure 1: Rel-sortability of random ER and SF DAGs with view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of rel-sortability to other sortabilities in ER DAGs. view at source ↗
Figure 3
Figure 3. Figure 3: Comparative SID (lower is better) performance of rel-SortnRegress on ER DAGs. view at source ↗
Figure 4
Figure 4. Figure 4: A directed and cyclic summary graph and corresponding time-unrolled DAG for view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of rel-sortability to other sortabilities in SF DAGs. view at source ↗
Figure 6
Figure 6. Figure 6: Comparative SID (lower is better) performance of rel-SortnRegress on SF DAGs. view at source ↗
Figure 7
Figure 7. Figure 7: compares the different sortabilities on data from non-standardized SCMs as discussed in Section 2.2. We observe that sortability by variance is extremely high for ER and SF DAGs, corroborating the findings in Reisach et al. (2021), and suggesting information in the variance as an explanation for the performance difference of the DAGMA algorithm between the standardized settings (sSCM) in Figures 3a and 6a … view at source ↗
read the original abstract

Random directed acyclic graphs (DAGs) based on imposing an order on Erd\H{o}s-R\'enyi and scale free random graphs are widely used for evaluating causal discovery algorithms. We show that in such DAGs, the set of nodes reachable via open paths, termed relatives, increases monotonically along the causal order. We assess the prevalence of this pattern numerically, and demonstrate that it can be exploited for causal order recovery via sorting by the estimated number of relatives. We note that many simulations in the literature feature settings where this yields an excellent proxy for the causal order, and show that a strict increase of relatives along the causal order leads to a singular Markov equivalence class. We propose sampling time-series DAGs as a possible alternative and discuss implications for causal discovery algorithms and their evaluation on synthetic data.

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

2 major / 2 minor

Summary. The paper claims that in random causal DAGs generated by imposing a topological order on Erdős-Rényi or scale-free graphs, the set of nodes reachable via open paths (termed 'relatives') increases monotonically along the causal order. It numerically assesses the prevalence of this monotonicity, shows that sorting nodes by the estimated number of relatives can recover the causal order (and is an excellent proxy in many literature simulations), proves that strict monotonicity implies a singular Markov equivalence class, proposes time-series DAG sampling as an alternative generation method, and discusses implications for evaluating causal discovery algorithms.

Significance. If the monotonicity property holds in the population and the number of relatives can be reliably estimated, the result would identify a structural bias in standard synthetic DAG generators that makes causal order recovery trivial in many simulation settings, thereby improving the design and interpretation of benchmarks for causal discovery methods. The suggestion of time-series alternatives and the Markov equivalence result are useful contributions to the literature on random graph models for causal inference.

major comments (2)
  1. [numerical assessment section] The central claim that sorting by the estimated number of relatives recovers the causal order (abstract and numerical assessment section) is load-bearing on accurate recovery of d-connections from finite observational data, but the manuscript provides no details on the conditional independence tests used, sample sizes, or robustness to test errors and faithfulness violations; this leaves the practical utility of the sorting procedure unverified even when the population monotonicity holds.
  2. [Markov equivalence section] § on Markov equivalence: While the proof that strict increase of relatives along the order yields a singular equivalence class is noted, the manuscript does not quantify how often strict monotonicity occurs under the ER/scale-free generators or discuss whether the result extends beyond the specific random graph models considered.
minor comments (2)
  1. [abstract] The abstract states a numerical assessment of prevalence but the manuscript should include a table or figure with exact simulation parameters (number of nodes, edge probabilities, number of replicates) to allow reproducibility.
  2. [introduction] Notation for 'relatives' and 'open paths' should be defined more formally with reference to d-separation in the main text before the numerical results.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive referee report. We address each major comment below and will revise the manuscript accordingly to improve clarity and completeness.

read point-by-point responses

  1. Referee: The central claim that sorting by the estimated number of relatives recovers the causal order (abstract and numerical assessment section) is load-bearing on accurate recovery of d-connections from finite observational data, but the manuscript provides no details on the conditional independence tests used, sample sizes, or robustness to test errors and faithfulness violations; this leaves the practical utility of the sorting procedure unverified even when the population monotonicity holds.

    Authors: We agree that details on finite-sample estimation are needed to substantiate practical utility. The current numerical assessment demonstrates prevalence of the monotonicity property using true population d-connections, with sorting presented as a conceptual exploitation. We will revise the numerical assessment section to specify the conditional independence tests (e.g., partial correlation tests), sample sizes, and add simulations assessing robustness to test errors and faithfulness violations. revision: yes

  2. Referee: While the proof that strict increase of relatives along the order yields a singular equivalence class is noted, the manuscript does not quantify how often strict monotonicity occurs under the ER/scale-free generators or discuss whether the result extends beyond the specific random graph models considered.

    Authors: The numerical assessment already evaluates prevalence of monotonicity for ER and scale-free generators; we will revise to explicitly report the frequency of strict monotonicity. The Markov equivalence proof is general for any DAG with strict monotonicity in relatives and does not rely on the specific generators. We will add discussion clarifying this generality while noting time-series sampling as one alternative to avoid the bias in practice. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained from graph generation process

full rationale

The paper establishes a monotonicity property of the 'relatives' set (nodes reachable via open paths) directly from the standard random DAG generation procedure (impose topological order, then sample edges via ER or scale-free models). This is shown analytically for the population case and assessed via numerical prevalence checks on generated graphs. The proposal to sort by estimated relatives count for order recovery follows as an application, without any step where a fitted parameter or self-citation is redefined as the output. Estimation from data is discussed as a practical step but does not enter the core derivation by construction. No self-definitional loops, fitted-input predictions, or load-bearing self-citations appear in the chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The analysis relies on standard definitions from causal graphical models and random graph theory; no new free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption Random causal DAGs are generated by first imposing a total order on nodes and then sampling edges from Erdős-Rényi or scale-free models while respecting the order.
    This is the standard construction used in the causal discovery simulation literature referenced in the abstract.

pith-pipeline@v0.9.0 · 5437 in / 1482 out tokens · 44816 ms · 2026-05-08T07:23:14.269215+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

76 extracted references

  1. [1]

    Better Simulations for Validating Causal Discovery with the DAG-Adaptation of the Onion Method

    Bryan Andrews and Erich Kummerfeld. Better Simulations for Validating Causal Discovery with the DAG-Adaptation of the Onion Method. 2024 (cit. on pp. 2, 6)

    2024

  2. [2]

    Causal Discovery between time series

    Charles Assaad. Causal Discovery between time series. Theses. Universit´e Grenoble Alpes, July 2021 (cit. on p. 8)

    2021

  3. [3]

    Do we become wiser with time? On causal equivalence with tiered background knowledge

    Christine W Bang and Vanessa Didelez. Do we become wiser with time? On causal equivalence with tiered background knowledge. In:Uncertainty in Artificial Intelligence. PMLR. 2023, pp. 119–129 (cit. on p. 8)

    2023

  4. [4]

    Emergence of Scaling in Random Networks

    Albert-L´aszl´o Barab´asi and R ´eka Albert. Emergence of Scaling in Random Networks. In: Science286.5439 (1999), pp. 509–512 (cit. on p. 1)

    1999

  5. [5]

    DAGMA: Learning DAGs via M- matrices and a Log-Determinant Acyclicity Characterization

    Kevin Bello, Bryon Aragam, and Pradeep Ravikumar. DAGMA: Learning DAGs via M- matrices and a Log-Determinant Acyclicity Characterization. In:Advances in Neural Informa- tion Processing Systems35 (2022), pp. 8226–8239 (cit. on p. 6)

    2022

  6. [6]

    Gilles Blanchard, Nicolas Curien, Klara Krause, and Alexander G. Reisach. A phase transition in Erd˝os-Barak random graphs. arXiv preprint. 2025 (cit. on p. 2)

    2025

  7. [7]

    Kording, Karen Sachs, Alexandre Drouin, and Dhanya Sridhar

    Philippe Brouillard, Chandler Squires, Jonas Wahl, Konrad P. Kording, Karen Sachs, Alexandre Drouin, and Dhanya Sridhar. The Landscape of Causal Discovery Data: Grounding Causal Discovery in Real-World Applications. In:Causal Learning and Reasoning (CLeaR). 2025 (cit. on pp. 1, 8)

    2025

  8. [8]

    Increasing paths in random temporal graphs

    Nicolas Broutin, Nina Kamcev, and G´abor Lugosi. Increasing paths in random temporal graphs. In:The Annals of Applied Probability34.6 (2024), pp. 5498–5521 (cit. on p. 2)

    2024

  9. [9]

    CausalTime: Realistically Generated Time-series for Benchmarking of Causal Discovery

    Yuxiao Cheng, Ziqian Wang, Tingxiong Xiao, Qin Zhong, Jinli Suo, and Kunlun He. CausalTime: Realistically Generated Time-series for Benchmarking of Causal Discovery. In:The Twelfth International Conference on Learning Representations. 2023 (cit. on p. 8)

    2023

  10. [10]

    Philip A. Dawid. Beware of the DAG! In:Causality: Abjectives and Assessment. PMLR. 2010, pp. 59–86 (cit. on p. 2)

    2010

  11. [11]

    On the Evolution of Random Graphs

    Paul Erd˝os, Alfr´ed R´enyi, et al. On the Evolution of Random Graphs. In:Publ. Math. Inst. Hung. Acad. Sci5.1 (1960), pp. 17–60 (cit. on p. 1)

    1960

  12. [12]

    Causal chambers as a real-world physical testbed for AI methodology

    Juan L Gamella, Jonas Peters, and Peter B¨uhlmann. Causal chambers as a real-world physical testbed for AI methodology. In:Nature Machine Intelligence7.1 (2025), pp. 107–118 (cit. on p. 8)

    2025

  13. [13]

    Identifying independence in Bayesian networks

    Dan Geiger, Thomas Verma, and Judea Pearl. Identifying independence in Bayesian networks. In:Networks20.5 (1990), pp. 507–534 (cit. on p. 2)

    1990

  14. [14]

    The Case for Evaluating Causal Models Using Interventional Measures and Empirical Data

    Amanda Gentzel, Dan Garant, and David Jensen. The Case for Evaluating Causal Models Using Interventional Measures and Empirical Data. In:Advances in Neural Information Processing Systems (NeurIPS). V ol. 32. 2019 (cit. on p. 1). 10

    2019

  15. [15]

    causalAssembly: Generating Realistic Production Data for Benchmarking Causal Discovery

    Konstantin G¨obler, Tobias Windisch, Mathias Drton, Tim Pychynski, Martin Roth, and Steffen Sonntag. causalAssembly: Generating Realistic Production Data for Benchmarking Causal Discovery. In:Causal Learning and Reasoning. PMLR. 2024, pp. 609–642 (cit. on p. 8)

    2024

  16. [16]

    Maathuis, and Nicolai Meinshausen

    Christina Heinze-Deml, Marloes H. Maathuis, and Nicolai Meinshausen. Causal Structure Learning. In:Annual Review of Statistics and Its Application5 (2018), pp. 371–391 (cit. on p. 1)

    2018

  17. [17]

    Unitless Unrestricted Markov- Consistent SCM Generation: Better Benchmark Datasets for Causal Discovery

    Rebecca J Herman, Jonas Wahl, Urmi Ninad, and Jakob Runge. Unitless Unrestricted Markov- Consistent SCM Generation: Better Benchmark Datasets for Causal Discovery. In:Causal Learning and Reasoning. PMLR. 2025, pp. 1506–1531 (cit. on pp. 2, 6, 8)

    2025

  18. [18]

    Hoyer, Dominik Janzing, Joris M

    Patrik O. Hoyer, Dominik Janzing, Joris M. Mooij, Jonas Peters, Bernhard Sch¨olkopf, et al. Nonlinear causal discovery with additive noise models. In:Advances in Neural Information Processing Systems. V ol. 21. Citeseer. 2008, pp. 689–696 (cit. on p. 5)

    2008

  19. [19]

    Antti Hyttinen, Frederick Eberhardt, and Patrik O. Hoyer. Learning Linear Cyclic Causal Models with Latent Variables. In:Journal of Machine Learning Research13.109 (2012), pp. 3387–3439 (cit. on p. 8)

    2012

  20. [20]

    Lower bounds on the size of Markov equivalence classes

    Erik Jahn, Frederick Eberhardt, and Leonard J Schulman. Lower bounds on the size of Markov equivalence classes. In:arXiv preprint(2025) (cit. on p. 7)

    2025

  21. [21]

    Kitson, Anthony C

    Neville K. Kitson, Anthony C. Constantinou, Zhigao Guo, Yang Liu, and Kiattikun Chobtham. A survey of Bayesian Network structure learning. In:Artificial Intelligence Review(2023), pp. 1–94 (cit. on p. 1)

    2023

  22. [22]

    Causal inference and causal explanation with background knowledge

    Christopher Meek. Causal inference and causal explanation with background knowledge. In: Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence. UAI’95. Morgan Kaufmann Publishers Inc., 1995, pp. 403–410 (cit. on p. 7)

    1995

  23. [23]

    Mooij, Dominik Janzing, Tom Heskes, and Bernhard Sch¨olkopf

    Joris M. Mooij, Dominik Janzing, Tom Heskes, and Bernhard Sch¨olkopf. On Causal Discovery with Cyclic Additive Noise Models. In:Advances in Neural Information Processing Systems (NeurIPS). V ol. 24. 2011 (cit. on p. 8)

    2011

  24. [24]

    Benchmarking Constraint-Based Bayesian Structure Learning Algorithms: Role of Network Topology

    Radha Nagarajan and Marco Scutari. Benchmarking Constraint-Based Bayesian Structure Learning Algorithms: Role of Network Topology. 2025 (cit. on p. 2)

    2025

  25. [25]

    Stan- dardizing Structural Causal Models

    Weronika Ormaniec, Scott Sussex, Lars Lorch, Bernhard Sch¨olkopf, and Andreas Krause. Stan- dardizing Structural Causal Models. In:International Conference on Learning Representations. 2025 (cit. on pp. 2, 6)

    2025

  26. [26]

    Causality: Models, Reasoning, and Inference

    Judea Pearl. Causality: Models, Reasoning, and Inference. Cambridge University Press, 2009 (cit. on pp. 1, 2, 5)

    2009

  27. [27]

    Structural Intervention Distance for Evaluating Causal Graphs

    Jonas Peters and Peter B ¨uhlmann. Structural Intervention Distance for Evaluating Causal Graphs. In:Neural computation27.3 (2015), pp. 771–799 (cit. on p. 7)

    2015

  28. [28]

    Causal Inference on Time Series using Restricted Structural Equation Models

    Jonas Peters, Dominik Janzing, and Bernhard Sch ¨olkopf. Causal Inference on Time Series using Restricted Structural Equation Models. In:Advances in Neural Information Processing Systems (NeurIPS). V ol. 26. 2013 (cit. on p. 8)

    2013

  29. [29]

    Elements of Causal Inference: Foundations and Learning Algorithms

    Jonas Peters, Dominik Janzing, and Bernhard Sch ¨olkopf. Elements of Causal Inference: Foundations and Learning Algorithms. MIT Press, 2017 (cit. on p. 2)

    2017

  30. [30]

    Reisach, Christof Seiler, and Sebastian Weichwald

    Alexander G. Reisach, Christof Seiler, and Sebastian Weichwald. Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy to Game. In:Advances in Neural Informa- tion Processing Systems (NeurIPS). V ol. 34. 2021 (cit. on pp. 1, 3, 5, 7)

    2021

  31. [31]

    Reisach, Alberto Su ´arez, Sebastian Weichwald, and Antoine Chambaz

    Alexander G. Reisach, Alberto Su ´arez, Sebastian Weichwald, and Antoine Chambaz. The Case for Time in Causal DAGs. In:Philosophy of Science(2026) (cit. on p. 8)

    2026

  32. [32]

    Reisach, Myriam Tami, Christof Seiler, Antoine Chambaz, and Sebastian Weichwald

    Alexander G. Reisach, Myriam Tami, Christof Seiler, Antoine Chambaz, and Sebastian Weichwald. A Scale-Invariant Sorting Criterion to Find a Causal Order in Additive Noise Models. In:Advances in Neural Information Processing Systems (NeurIPS). V ol. 36. 2023 (cit. on pp. 2, 3, 5, 6)

    2023

  33. [33]

    Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models

    Paul Rolland, V olkan Cevher, Matth ¨aus Kleindessner, Chris Russell, Dominik Janzing, Bern- hard Sch ¨olkopf, and Francesco Locatello. Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models. In:International Conference on Machine Learning. PMLR, 2022, pp. 18741–18753 (cit. on p. 3)

    2022

  34. [34]

    On the number and size of Markov equivalence classes of random directed acyclic graphs

    Dominik Schmid and Allan Sly. On the number and size of Markov equivalence classes of random directed acyclic graphs. 2022 (cit. on p. 2). 11

    2022

  35. [35]

    Causation, Prediction, and Search

    Peter Spirtes, Clark Glymour, and Richard Scheines. Causation, Prediction, and Search. MIT Press, 2001 (cit. on pp. 1, 3, 5)

    2001

  36. [36]

    Ordering-Based Search: A Simple and Effective Algo- rithm for Learning Bayesian Networks

    Marc Teyssier and Daphne Koller. Ordering-Based Search: A Simple and Effective Algo- rithm for Learning Bayesian Networks. In:Proceedings of the Twenty-First Conference on Uncertainty in Artificial Intelligence. AUAI Press, 2005, pp. 584–590 (cit. on p. 5)

    2005

  37. [37]

    V owels, Necati Cihan Camgoz, and Richard Bowden

    Matthew J. V owels, Necati Cihan Camgoz, and Richard Bowden. D’Ya Like DAGs? A Survey on Structure Learning and Causal Discovery. In:ACM Computing Surveys55.4 (2022) (cit. on p. 1)

    2022

  38. [38]

    Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning

    Marcel Wien¨obst, Leonard Henckel, and Sebastian Weichwald. Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning. In:International Conference on Learning Representations. 2025 (cit. on p. 6). A Proof Details This section provides additional details on the proofs of Theorem 2, Theorem 3, and Theorem 4. A.1 Proof of Theorem 2 (Addit...

    2025

  39. [39]

    Better Simulations for Validating Causal Discovery with the DAG-Adaptation of the Onion Method

    Bryan Andrews and Erich Kummerfeld. Better Simulations for Validating Causal Discovery with the DAG-Adaptation of the Onion Method. 2024 (cit. on p. 16)

    2024

  40. [40]

    BayesDAG: Gradient-Based Posterior Inference for Causal Discovery

    Yashas Annadani, Nick Pawlowski, Joel Jennings, Stefan Bauer, Cheng Zhang, and Wenbo Gong. BayesDAG: Gradient-Based Posterior Inference for Causal Discovery. In: 36 (2023). Ed. by A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, pp. 1738–1763 (cit. on p. 15)

    2023

  41. [41]

    DAGMA: Learning DAGs via M- matrices and a Log-Determinant Acyclicity Characterization

    Kevin Bello, Bryon Aragam, and Pradeep Ravikumar. DAGMA: Learning DAGs via M- matrices and a Log-Determinant Acyclicity Characterization. In:Advances in Neural Informa- tion Processing Systems35 (2022), pp. 8226–8239 (cit. on pp. 15, 16)

    2022

  42. [42]

    Differentiable Causal Discovery from Interventional Data

    Philippe Brouillard, S ´ebastien Lachapelle, Alexandre Lacoste, Simon Lacoste-Julien, and Alexandre Drouin. Differentiable Causal Discovery from Interventional Data. In:Advances in Neural Information Processing Systems. Ed. by H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin. V ol. 33. Curran Associates, Inc., 2020, pp. 21865–21877 (cit. on p. 15)

    2020

  43. [43]

    CAM: Causal Additive Models, High- Dimensional Order Search and Penalized Regression

    Peter B ¨uhlmann, Jonas Peters, and Jan Ernest. CAM: Causal Additive Models, High- Dimensional Order Search and Penalized Regression. In:The Annals of Statistics42.6 (2014), pp. 2526–2556 (cit. on p. 15)

    2014

  44. [44]

    The igraph software package for complex network research

    G´abor Cs´ardi and Tam´as Nepusz. The igraph software package for complex network research. In:InterJournal, Complex Systems(2006), p. 1695 (cit. on p. 16)

    2006

  45. [45]

    Bayesian structure learning with generative flow networks

    Tristan Deleu, Ant´onio G´ois, Chris Emezue, Mansi Rankawat, Simon Lacoste-Julien, Stefan Bauer, and Yoshua Bengio. Bayesian structure learning with generative flow networks. In: Uncertainty in Artificial Intelligence. PMLR. 2022, pp. 518–528 (cit. on p. 15). 17

    2022

  46. [46]

    A Meta-Learning Approach to Bayesian Causal Discovery

    Anish Dhir, Matthew Ashman, James Requeima, and Mark van der Wilk. A Meta-Learning Approach to Bayesian Causal Discovery. In:The Thirteenth International Conference on Learning Representations. 2025 (cit. on p. 15)

    2025

  47. [47]

    Hagberg, Daniel A

    Aric A. Hagberg, Daniel A. Schult, and Pieter J. Swart. Exploring Network Structure, Dynam- ics, and Function using NetworkX. In:Proceedings of the 7th Python in Science Conference. Ed. by Ga¨el Varoquaux, Travis Vaught, and Jarrod Millman. Pasadena, CA USA, Aug. 2008, pp. 11–15 (cit. on p. 16)

    2008

  48. [48]

    Harris, K

    Charles R. Harris, K. Jarrod Millman, St´efan J. van der Walt, Ralf Gommers, Pauli Virtanen, David Cournapeau, Eric Wieser, Julian Taylor, Sebastian Berg, Nathaniel J. Smith, Robert Kern, Matti Picus, Stephan Hoyer, Marten H. van Kerkwijk, Matthew Brett, Allan Haldane, Jaime Fern´andez del R´ıo, Mark Wiebe, Pearu Peterson, Pierre G´erard-Marchant, Kevin S...

    2020

  49. [49]

    Adjustment Identification Dis- tance: A gadjid for Causal Structure Learning

    Leonard Henckel, Theo W¨urtzen, and Sebastian Weichwald. Adjustment Identification Dis- tance: A gadjid for Causal Structure Learning. In:The 40th Conference on Uncertainty in Artificial Intelligence. 2024 (cit. on p. 16)

    2024

  50. [50]

    Unitless Unrestricted Markov- Consistent SCM Generation: Better Benchmark Datasets for Causal Discovery

    Rebecca J Herman, Jonas Wahl, Urmi Ninad, and Jakob Runge. Unitless Unrestricted Markov- Consistent SCM Generation: Better Benchmark Datasets for Causal Discovery. In:Causal Learning and Reasoning. PMLR. 2025, pp. 1506–1531 (cit. on p. 16)

    2025

  51. [51]

    Generalized score functions for causal discovery

    Biwei Huang, Kun Zhang, Yizhu Lin, Bernhard Sch¨olkopf, and Clark Glymour. Generalized score functions for causal discovery. In:Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining. 2018, pp. 1551–1560 (cit. on p. 15)

    2018

  52. [52]

    J. D. Hunter. Matplotlib: A 2D graphics environment. In:Computing in Science & Engineering 9.3 (2007), pp. 90–95 (cit. on p. 16)

    2007

  53. [53]

    Constraint-based Causal Discovery: Conflict Resolution with Answer Set Programming

    Antti Hyttinen, Frederick Eberhardt, and Matti J¨arvisalo. Constraint-based Causal Discovery: Conflict Resolution with Answer Set Programming. In:UAI. 2014, pp. 340–349 (cit. on p. 15)

    2014

  54. [54]

    Estimating High-Dimensional Directed Acyclic Graphs with the PC-Algorithm

    Markus Kalisch and Peter B¨uhlman. Estimating High-Dimensional Directed Acyclic Graphs with the PC-Algorithm. In:Journal of Machine Learning Research8.3 (2007) (cit. on p. 15)

    2007

  55. [55]

    Large-Scale Targeted Cause Discovery via Learning from Simulated Data

    Jang-Hyun Kim, Claudia Skok Gibbs, Sangdoo Yun, Hyun Oh Song, and Kyunghyun Cho. Large-Scale Targeted Cause Discovery via Learning from Simulated Data. In:Transactions on Machine Learning Research(2025) (cit. on p. 15)

    2025

  56. [56]

    Gradient- based neural dag learning

    S´ebastien Lachapelle, Philippe Brouillard, Tristan Deleu, and Simon Lacoste-Julien. Gradient- based neural dag learning. In:International Conference on Learning Representations. 2020 (cit. on p. 15)

    2020

  57. [57]

    DiBS: Differentiable Bayesian Structure Learning

    Lars Lorch, Jonas Rothfuss, Bernhard Sch¨olkopf, and Andreas Krause. DiBS: Differentiable Bayesian Structure Learning. In:Advances in Neural Information Processing Systems34 (2021), pp. 24111–24123 (cit. on p. 15)

    2021

  58. [58]

    Sara Magliacane, Tom Claassen, and Joris M. Mooij. Ancestral causal inference. In:Advances in Neural Information Processing Systems29 (2016) (cit. on p. 15)

    2016

  59. [59]

    Causal inference and causal explanation with background knowledge

    Christopher Meek. Causal inference and causal explanation with background knowledge. In: Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence. UAI’95. Morgan Kaufmann Publishers Inc., 1995, pp. 403–410 (cit. on p. 14)

    1995

  60. [60]

    Scalable Causal Discovery with Score Matching

    Francesco Montagna, Nicoletta Noceti, Lorenzo Rosasco, Kun Zhang, and Francesco Locatello. Scalable Causal Discovery with Score Matching. In:Conference on Causal Learning and Reasoning. PMLR. 2023, pp. 752–771 (cit. on p. 15)

    2023

  61. [61]

    On the role of sparsity and dag constraints for learning linear dags

    Ignavier Ng, Amir Emad Ghassami, and Kun Zhang. On the role of sparsity and dag constraints for learning linear dags. In:Advances in Neural Information Processing Systems. V ol. 33. 2020 (cit. on p. 15)

    2020

  62. [62]

    Structure learning with continuous optimization: A sober look and beyond

    Ignavier Ng, Biwei Huang, and Kun Zhang. Structure learning with continuous optimization: A sober look and beyond. In:Causal Learning and Reasoning. PMLR. 2024, pp. 71–105 (cit. on p. 15)

    2024

  63. [63]

    Stan- dardizing Structural Causal Models

    Weronika Ormaniec, Scott Sussex, Lars Lorch, Bernhard Sch¨olkopf, and Andreas Krause. Stan- dardizing Structural Causal Models. In:International Conference on Learning Representations. 2025 (cit. on p. 15). 18

    2025

  64. [64]

    Version latest

    pandas Development Team.pandas-dev/pandas: Pandas. Version latest. Feb. 2020 (cit. on p. 16)

    2020

  65. [65]

    Structural Intervention Distance for Evaluating Causal Graphs

    Jonas Peters and Peter B ¨uhlmann. Structural Intervention Distance for Evaluating Causal Graphs. In:Neural computation27.3 (2015), pp. 771–799 (cit. on p. 15)

    2015

  66. [66]

    Mooij, Dominik Janzing, and Bernhard Sch¨olkopf

    Jonas Peters, Joris M. Mooij, Dominik Janzing, and Bernhard Sch¨olkopf. Causal discovery with continuous additive noise models. In: (2014) (cit. on p. 15)

    2014

  67. [67]

    Reisach, Christof Seiler, and Sebastian Weichwald

    Alexander G. Reisach, Christof Seiler, and Sebastian Weichwald. Beware of the Simulated DAG! Causal Discovery Benchmarks May Be Easy to Game. In:Advances in Neural Informa- tion Processing Systems (NeurIPS). V ol. 34. 2021 (cit. on p. 17)

    2021

  68. [68]

    Reisach, Myriam Tami, Christof Seiler, Antoine Chambaz, and Sebastian Weichwald

    Alexander G. Reisach, Myriam Tami, Christof Seiler, Antoine Chambaz, and Sebastian Weichwald. A Scale-Invariant Sorting Criterion to Find a Causal Order in Additive Noise Models. In:Advances in Neural Information Processing Systems (NeurIPS). V ol. 36. 2023 (cit. on p. 16)

    2023

  69. [69]

    Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models

    Paul Rolland, V olkan Cevher, Matth ¨aus Kleindessner, Chris Russell, Dominik Janzing, Bern- hard Sch ¨olkopf, and Francesco Locatello. Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models. In:International Conference on Machine Learning. PMLR, 2022, pp. 18741–18753 (cit. on p. 15)

    2022

  70. [70]

    Hoyer, and Kenneth Bollen

    Shohei Shimizu, Takanori Inazumi, Yasuhiro Sogawa, Aapo Hyv¨arinen, Yoshinobu Kawahara, Takashi Washio, Patrik O. Hoyer, and Kenneth Bollen. DirectLiNGAM: A direct method for learning a linear non-Gaussian structural equation model. In:Journal of Machine Learning Research12 (2011), pp. 1225–1248 (cit. on p. 15)

    2011

  71. [71]

    Learning Bayesian networks with discrete variables from data

    Peter Spirtes and Christopher Meek. Learning Bayesian networks with discrete variables from data. In:KDD. V ol. 1. 1995, pp. 294–299 (cit. on p. 15)

    1995

  72. [72]

    Towards Scalable Bayesian Learning of Causal DAGs

    Jussi Viinikka, Antti Hyttinen, Johan Pensar, and Mikko Koivisto. Towards Scalable Bayesian Learning of Causal DAGs. In:Advances in Neural Information Processing Systems33 (2020), pp. 6584–6594 (cit. on p. 15)

    2020

  73. [73]

    Michael L. Waskom. seaborn: statistical data visualization. In:Journal of Open Source Software 6.60 (2021), p. 3021 (cit. on p. 16)

    2021

  74. [74]

    Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning

    Marcel Wien¨obst, Leonard Henckel, and Sebastian Weichwald. Embracing Discrete Search: A Reasonable Approach to Causal Structure Learning. In:International Conference on Learning Representations. 2025 (cit. on pp. 15, 16)

    2025

  75. [75]

    DAG-GNN: DAG Structure Learning with Graph Neural Networks

    Yue Yu, Jie Chen, Tian Gao, and Mo Yu. DAG-GNN: DAG Structure Learning with Graph Neural Networks. In:International Conference on Machine Learning. PMLR, 2019, pp. 7154– 7163 (cit. on p. 15)

    2019

  76. [76]

    DAGs with NO TEARS: Continuous Optimization for Structure Learning

    Xun Zheng, Bryon Aragam, Pradeep K Ravikumar, and Eric P Xing. DAGs with NO TEARS: Continuous Optimization for Structure Learning. In:Advances in Neural Information Process- ing Systems. V ol. 32. 2018, pp. 9472–9483 (cit. on pp. 15, 16). 19

    2018