Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem

Turan Orujlu, Christian Gumbsch, Martin V. Butz, Charley M Wu
Proceedings of the Fifth Conference on Causal Learning and Reasoning, PMLR 323:1450-1480, 2026.

Abstract

Most neural models of causality assume static causal graphs, failing to capture the dynamic and sparse nature of physical interactions where causal relationships emerge and dissolve over time. We introduce the Causal Process Framework and its neural implementation, Causal Process Models (CPMs), for learning sparse, time-varying causal graphs from visual observations. Unlike traditional approaches that maintain dense connectivity, our model explicitly constructs causal edges only when objects actively interact, dramatically improving both interpretability and computational efficiency. We achieve this by casting dynamic interaction-graph construction for world modeling as a multi-agent reinforcement learning problem, where specialized agents sequentially decide which objects are causally connected at each timestep. Our key innovation is a structured representation that factorizes object and force vectors along three learned dimensions (mutability, causal relevance, and control relevance), enabling the automatic discovery of semantically meaningful encodings. We demonstrate that a CPM significantly outperforms dense graph baselines on physical prediction tasks, particularly for longer horizons and varying object counts.

Cite this Paper

BibTeX
@InProceedings{pmlr-v323-orujlu26a, title = {Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem}, author = {Orujlu, Turan and Gumbsch, Christian and Butz, Martin V. and Wu, Charley M}, booktitle = {Proceedings of the Fifth Conference on Causal Learning and Reasoning}, pages = {1450--1480}, year = {2026}, editor = {Mazaheri, Bijan and Hanson, Niels Richard}, volume = {323}, series = {Proceedings of Machine Learning Research}, month = {06--08 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v323/main/assets/orujlu26a/orujlu26a.pdf}, url = {https://proceedings.mlr.press/v323/orujlu26a.html}, abstract = {Most neural models of causality assume static causal graphs, failing to capture the dynamic and sparse nature of physical interactions where causal relationships emerge and dissolve over time. We introduce the Causal Process Framework and its neural implementation, Causal Process Models (CPMs), for learning sparse, time-varying causal graphs from visual observations. Unlike traditional approaches that maintain dense connectivity, our model explicitly constructs causal edges only when objects actively interact, dramatically improving both interpretability and computational efficiency. We achieve this by casting dynamic interaction-graph construction for world modeling as a multi-agent reinforcement learning problem, where specialized agents sequentially decide which objects are causally connected at each timestep. Our key innovation is a structured representation that factorizes object and force vectors along three learned dimensions (mutability, causal relevance, and control relevance), enabling the automatic discovery of semantically meaningful encodings. We demonstrate that a CPM significantly outperforms dense graph baselines on physical prediction tasks, particularly for longer horizons and varying object counts.} }
Endnote
%0 Conference Paper %T Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem %A Turan Orujlu %A Christian Gumbsch %A Martin V. Butz %A Charley M Wu %B Proceedings of the Fifth Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2026 %E Bijan Mazaheri %E Niels Richard Hanson %F pmlr-v323-orujlu26a %I PMLR %P 1450--1480 %U https://proceedings.mlr.press/v323/orujlu26a.html %V 323 %X Most neural models of causality assume static causal graphs, failing to capture the dynamic and sparse nature of physical interactions where causal relationships emerge and dissolve over time. We introduce the Causal Process Framework and its neural implementation, Causal Process Models (CPMs), for learning sparse, time-varying causal graphs from visual observations. Unlike traditional approaches that maintain dense connectivity, our model explicitly constructs causal edges only when objects actively interact, dramatically improving both interpretability and computational efficiency. We achieve this by casting dynamic interaction-graph construction for world modeling as a multi-agent reinforcement learning problem, where specialized agents sequentially decide which objects are causally connected at each timestep. Our key innovation is a structured representation that factorizes object and force vectors along three learned dimensions (mutability, causal relevance, and control relevance), enabling the automatic discovery of semantically meaningful encodings. We demonstrate that a CPM significantly outperforms dense graph baselines on physical prediction tasks, particularly for longer horizons and varying object counts.
APA
Orujlu, T., Gumbsch, C., Butz, M.V. & Wu, C.M.. (2026). Causal Process Models: Reframing Dynamic Causal Graph Discovery as a Reinforcement Learning Problem. Proceedings of the Fifth Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 323:1450-1480 Available from https://proceedings.mlr.press/v323/orujlu26a.html.

Related Material