aiqbench analysis
The paper extends identifiability theory for causal effects in linear stochastic differential equations, covering systems with reciprocal (2-cycle) interactions and partially observed latent variables. For readers working on causal discovery and inference, the practical value is in the graphical conditions that determine when causal effects can be recovered from stationary covariance alone, and in the instrumental-variable approach for the latent-variable case. These are theoretical results, so their impact depends on whether the assumptions, such as linearity and genericity, hold in the applied systems of interest.
Identifiability from Stationary Covariance in Linear SDEs with Reciprocal Interactions and Latent Variables
arXiv:2610.31310v1 Announce Type: new Abstract: Continuous-time stochastic differential equations (SDEs) provide a natural framework for modeling interacting dynamical systems. A central causal question is whether causal effects among observed variables can be identified from observational data drawn from the stationary distribution, given the underlying causal graph. For linear SDEs, existing identifiability results are largely restricted to fully observed systems whose causal graphs contain no directed 2-cycles. We first establish new graphical conditions for the generic identifiability of causal effects in systems that may contain directed 2-cycles, thereby allowing for reciprocal interactions. Next, we address the previously unexplored setting of latent variables by developing an instrumental-variable framework for identifying a target direct causal effect under partial observability.
Abstract:Continuous-time stochastic differential equations (SDEs) provide a natural framework for modeling interacting dynamical systems. A central causal question is whether causal effects among observed variables can be identified from observational data drawn from the stationary distribution, given the underlying causal graph. For linear SDEs, existing identifiability results are largely restricted to fully observed systems whose causal graphs contain no directed 2-cycles. We first establish new graphical conditions for the generic identifiability of causal effects in systems that may contain directed 2-cycles, thereby allowing for reciprocal interactions. Next, we address the previously unexplored setting of latent variables by developing an instrumental-variable framework for identifying a target direct causal effect under partial observability.
Machine Learning (cs.LG)
arXiv:2610.31310 [cs.LG]
(or arXiv:2610.31310v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.31310 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 205 words
Related coverage
CausalBind: Causal Modeling and Learning for Protein-Molecule Virtual Screening
arXiv · Oct 7, 2026
When Are Neural Interaction Discoveries Real? Identifiability, Recoverability, and a Pre-Fit Diagnostic
arXiv · Jun 9, 2026
Eight-qubit toric code tested on a superconducting device with AI assistance
Quantum Zeitgeist · Nov 4, 2026
A Trust-Network-Based Federated Learning Framework for Multi-Center Aging Clock Prediction
arXiv · Sep 11, 2026