The paper investigates how learned multimodal representations of high‑dimensional health data can fail to preserve causal information, even when trained to reconstruct the data. It establishes conditions under which a representation is sufficient for causal identification, shows that interventional mediation requires stricter preservation than counterfactual comparison, and proves that reconstruction‑based objectives cannot certify sufficiency. The authors decompose the error of cross‑fitted doubly robust estimators into a first‑order representation bias and a second‑order nuisance‑estimation remainder, demonstrating that coverage can deteriorate with larger samples if the representation is insufficient. They propose a practical workflow that combines modality‑specific variational autoencoders, cross‑fitted doubly robust estimation, locked phenotypes, selective pathway analysis, portable representations, and diagnostics for representation insufficiency, exposure leakage, and sensitivity, and validate it in simulations where a rare confounder’s omission leads to severely under‑covered indirect‑effect intervals.
By Cong Cao
CIDER-FM is a causal foundation model that combines finite observational data with surrogate-interventional datasets to predict target conditional interventional distributions more accurately than using observational data alone. It employs an intervention-aware representation and hierarchical three‑axis attention to integrate information across variables, samples, and experimental regimes. Experiments on synthetic graphs, simulated data, and real‑world Causal Chambers data show that incorporating experimental context improves CID prediction performance.
By Yuche Gao, Arik Reuter, Siyuan Guo, Anish Dhir, Bernhard Sch\"olkopf, Adrian Weller
arXiv:2604. 02458v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are increasingly used to simulate human responses and estimate treatment effect of interventions when real-world experiments are costly or infeasible.
By Zonghan Li, Feng Ji
The paper investigates how the granularity of interventions—single components versus full treatment bundles—affects counterfactual simulations using a clinical world model. By analyzing 945,707 patient-hours from MIMIC-IV, the authors show that interventions are naturally bundled, and editing a single component often represents an unreal scenario. Experiments with Clin‑JEPA on 1,019 ventilation onsets demonstrate that editing the complete bundle produces larger predicted state changes than editing individual settings, indicating that bundle-aware editing better captures treatment sensitivity.
By Fangzhou Wang, Yixuan Yang, Camilla Balzarotti, Rishikesan Kamaleswaran
arXiv:2606. 17165v1 Announce Type: cross Abstract: Organizations and researchers show increasing interest in using large language models (LLMs) in place of human participants in A/B tests, in the hope of experimenting faster and at lower cost.
By Joel Persson, M{\aa}rten Schultzberg, Sebastian Ankargren
arXiv:2610.41368v1 Announce Type: new
Abstract: Overly restrictive eligibility criteria contribute to widespread enrollment failure in clinical trials, yet published recommendations for how specific...
By Russell Ro, Reza Abbasi-Asl
ACTIVA is a transformer‑based conditional variational autoencoder designed to estimate full interventional distributions from observational data without requiring interventional data, a causal graph, or per‑task fitting. It learns a conditional latent prior that amortizes causal knowledge across diverse training tasks, enabling a single forward pass to handle new causal models drawn from the training distribution. Empirical results on synthetic and biologically realistic gene‑expression simulations show that ACTIVA outperforms correlational baselines, reduces spurious non‑descendant effects, and matches strong amortized baselines, while also validating its theoretical mixture‑of‑distributions property on observationally equivalent models.
By Andreas Sauter, Saber Salehkaleybar, Frank van Harmelen, Aske Plaat, Erman Acar
The study evaluates counterfactual bias in ten open‑source large language models (LLMs) for pediatric Emergency Severity Index (ESI) prediction. By creating paired clinical vignettes that differ only in demographic or socioeconomic variables, the authors measure shifts in acuity assignment, finding that counterfactual sensitivity varies widely across model families and sizes. A fine‑tuned Qwen2.5‑7B model exhibited the lowest sensitivity, while larger or medical‑domain models sometimes showed greater shifts, highlighting the need for fairness assessment before clinical deployment.
By Manar Aljohani, Brandon Ho, Kenneth McKinley, Dennis Ren, Xuan Wang
The paper introduces Bayesian Propensity Score Predictive Inference (PSPI), an interaction‑based outcome modeling framework that uses Bayesian additive regression trees and natural cubic splines of the sampling propensity score to extend randomized trial results to broader target populations. PSPI includes variants that apply splines to the treatment effect component alone or to both prognostic and treatment effect components, enabling separate regularization of selection‑related variation and explicit extrapolation beyond boundary knots. Simulation studies demonstrate that spline‑based PSPI models outperform competing methods, achieving lower root mean squared errors and near‑nominal coverage, especially under substantial covariate shift, and in the CODIACS Vanguard trial application PSPI estimates larger intervention benefits than the unadjusted trial estimate.
By Jungang Zou, Joseph E. Schwartz, Roderick J. A. Little, Nathalie Moise, Qixuan Chen
arXiv:2607. 14940v1 Announce Type: new Abstract: We study causal inference under outcome interference for sequential, observational settings.
By Phevos Paschalidis, Constantinos Daskalakis, Devavrat Shah
arXiv:2607. 25497v1 Announce Type: cross Abstract: Pathology foundation models are approaching clinical deployment, yet remain vulnerable to systematic non-biological variation across centres.
By Cl\'ement Grisi, Jeroen van der Laak, Geert Litjens
arXiv:2607. 28608v1 Announce Type: new Abstract: Clinical risk models routinely achieve strong aggregate performance while producing materially different error rates across patient subgroups.
By Sparsh Roy, Samuel Girmachew, Nishita Chavan