mlsanityData to 14 Oct 2026

Since Faithfulness Fails: The Performance Limits of Neural Causal Discovery

arXiv 2502.16056 ↗First posted 22 Feb 2025cs.LG stat.MLcited by 4

Not on the map. Why?

Abstract (arXiv)

Neural causal discovery methods have recently improved in terms of scalability and computational efficiency. However, our systematic evaluation highlights significant room for improvement in their accuracy when uncovering causal structures. We identify a fundamental limitation: neural networks cannot reliably distinguish between existing and non-existing causal relationships in the finite sample regime. Our experiments reveal that neural networks, as used in contemporary causal discovery approaches, lack the precision needed to recover ground-truth graphs, even for small graphs and relatively large sample sizes. Furthermore, we identify the faithfulness property as a critical bottleneck: (i) it is likely to be violated across any reasonable dataset size range, and (ii) its violation directly undermines the performance of neural discovery methods. These findings lead us to conclude that progress within the current paradigm is fundamentally constrained, necessitating a paradigm shift in this domain.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 17 of its 28 references that are in the map. First posted on arXiv 22 Feb 2025.

Builds on

17 of 28 references.

References · 17

  1. 2018DAGs with NO TEARS: Continuous Optimization for Structure Learningcited by 752
  2. 2015Causal inference using invariant prediction: identification and confidence intervalscited by 536
  3. 2019Gradient-Based Neural DAG Learningcited by 227
  4. 2020Differentiable Causal Discovery from Interventional Datacited by 218
  5. 2019Causality matters in medical imagingcited by 193
  6. 2022DAGMA: Learning DAGs via M-matrices and a Log-Determinant Acyclicity Characterizationcited by 140
  7. 2022Score matching enables causal discovery of nonlinear additive noise modelscited by 111
  8. 2021DiBS: Differentiable Bayesian Structure Learningcited by 106
  9. 2021Efficient Neural Causal Discovery without Acyclicity Constraintscited by 82
  10. 2021DAGs with No Curl: An Efficient DAG Structure Learning Approachcited by 53
  11. 2022Large-Scale Differentiable Causal Discovery of Factor Graphscited by 47
  12. 2019Scaling structural learning with NO-BEARS to infer causal transcriptome networkscited by 39
5 more
  1. 2023BayesDAG: Gradient-Based Posterior Inference for Causal Discoverycited by 33
  2. 2023Stable Differentiable Causal Discoverycited by 31
  3. 2022Reliable Causal Discovery with Improved Exact Search and Weaker Assumptionscited by 19
  4. 2024Are Bayesian networks typically faithful?cited by 6
  5. 2022Learning general conditional independence structures via the neighbourhood latticecited by 4

Built on by

Cited by year

Citing papers per year: 2025: 4.

4 papers.

  1. 2025CausalARC: Abstract Reasoning with Causal World Modelscited by 4
  2. 2025Causal-Adapter: Taming Text-to-Image Diffusion for Faithful Counterfactual Generationcited by 4
  3. 2025Heterogeneous Causal Discovery of Repeated Undesirable Health Outcomescited by 1
  4. 2025Amortized Causal Discovery with Prior-Fitted Networkscited by 1