DeconDTN-Toolkit simulates provenance shifts to expose ERM vulnerabilities and provides tools plus a robust OOD indicator for mitigating confounding by data provenance.
Towards Causal Representation Learning
6 Pith papers cite this work, alongside 76 external citations. Polarity classification is still indexing.
6
Pith papers citing it
76
external citations · Pith
abstract
The two fields of machine learning and graphical causality arose and developed separately. However, there is now cross-pollination and increasing interest in both fields to benefit from the advances of the other. In the present paper, we review fundamental concepts of causal inference and relate them to crucial open problems of machine learning, including transfer and generalization, thereby assaying how causality can contribute to modern machine learning research. This also applies in the opposite direction: we note that most work in causality starts from the premise that the causal variables are given. A central problem for AI and causality is, thus, causal representation learning, the discovery of high-level causal variables from low-level observations. Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.
citation-role summary
background 2
citation-polarity summary
years
2026 6roles
background 2polarities
background 2representative citing papers
Robots discover causal tool features through VLM suggestions and physics-based counterfactual perturbations in simulation, then transfer manipulation skills via conditioned keypoint matching.
Causal mediation analysis shows harmful LLM outputs arise in late layers from MLP failures and gating neurons, with early layers handling harm context detection and signal propagation.
Integrates partial ODE physics into SDE-based causal discovery via drift-diffusion separation, with sparsity-inducing quasi-likelihood estimation, recovery guarantees for stable/unstable systems, and robustness analysis to model misspecification.
In binary logistic temporal-graph models, higher Fisher information for parameter recovery coincides with higher irreducible predictive entropy, so the easiest-to-estimate regimes are the hardest to predict.
The paper introduces 'Platonic Projection Structures,' a reformulation of standard PSD operator theory applied to representation learning, with experiments that verify definitions rather than test predictions.
citing papers explorer
-
DeconDTN-Toolkit: A Library for Evaluation and Enhancement of Robustness to Provenance Shift
DeconDTN-Toolkit simulates provenance shifts to expose ERM vulnerabilities and provides tools plus a robust OOD indicator for mitigating confounding by data provenance.
-
Creative Robot Tool Use by Counterfactual Reasoning
Robots discover causal tool features through VLM suggestions and physics-based counterfactual perturbations in simulation, then transfer manipulation skills via conditioned keypoint matching.
-
Why Do Large Language Models Generate Harmful Content?
Causal mediation analysis shows harmful LLM outputs arise in late layers from MLP failures and gating neurons, with early layers handling harm context detection and signal propagation.
-
Causal Discovery from Heteroscedastic Stochastic Dynamical Systems under Imperfect Physical Models
Integrates partial ODE physics into SDE-based causal discovery via drift-diffusion separation, with sparsity-inducing quasi-likelihood estimation, recovery guarantees for stable/unstable systems, and robustness analysis to model misspecification.
-
Estimation-Prediction Tradeoff in Causal Probabilistic Temporal Graphs
In binary logistic temporal-graph models, higher Fisher information for parameter recovery coincides with higher irreducible predictive entropy, so the easiest-to-estimate regimes are the hardest to predict.
-
Platonic Projection Structures: Operator-Induced Observability in Representation Learning
The paper introduces 'Platonic Projection Structures,' a reformulation of standard PSD operator theory applied to representation learning, with experiments that verify definitions rather than test predictions.