arXiv:2606. 21806v2 Announce Type: replace Abstract: Deep generative models reproduce the observational distribution of their training data, inheriting any spurious associations it contains.
arXiv:2609.14124v1 Announce Type: cross Abstract: Medical image analysis is often hindered by biased datasets, which can lead to biased models and limited clinical applicability. A promising strategy...
arXiv:2608. 06427v1 Announce Type: new Abstract: Generative models can reproduce an observational distribution while encoding an incorrect causal structure.
arXiv:2603.20111v2 Announce Type: replace Abstract: The Joint-Embedding Predictive Architecture (JEPA) is often seen as a non-generative alternative to likelihood-based self-supervised learning, emph...
arXiv:2609.24879v1 Announce Type: new Abstract: Counterfactual image generation answers questions about how a subject would have looked under retrospective, hypothetical scenarios. Recent methods hav...
The paper introduces a control‑variable framework for deep neural networks to mitigate omitted variable bias, particularly shortcut learning where covariates like demographics influence predictions. It refits the final layer of a pre‑trained network using cross‑fitting with ridge penalisation, orthogonalises covariate effects, and marginalises predictions over covariate distributions to achieve unbiased, interpretable results. Experiments on simulated images and neuroimaging data show consistent estimation of true effects and performance close to models trained on unconfounded data.
The paper investigates why joint audio–video generators often learn to predict sound from visual appearance rather than from the underlying event, a problem termed the visual shortcut. By constructing a controlled causal model where audio is independent of video appearance, the authors show that common remedies such as shared latent spaces fail to prevent this shortcut. They propose that intervening on the nuisance appearance is necessary and sufficient for counterfactual invariance, and validate this approach across synthetic and real datasets, highlighting the remaining challenge of unknown nuisances.
The paper introduces a mixture‑learning framework for causal inference with unobserved confounding, treating latent confounders as sources of heterogeneity that create mixture structures in observed data. By assuming suitable structural and identifiability conditions, it shows that recovering the mixing distribution and component mechanisms allows estimation of interventional distributions and causal estimands. The authors illustrate the approach with Bernoulli mixture examples, extend it to high‑dimensional exponential‑family mixtures with dependent outcomes, and relate it to panel‑data settings, latent factor models, and synthetic interventions.
arXiv:2609.07874v1 Announce Type: new Abstract: Multimodal large language models often capture visual-linguistic correlations but struggle to predict how local visual interventions propagate and affe...
arXiv:2607. 01104v1 Announce Type: cross Abstract: In Large Language Model (LLM) training, data mixing plays a pivotal role in determining model performance.
arXiv:2401. 04890v2 Announce Type: replace-cross Abstract: This work introduces a novel principle for disentanglement we call mechanism sparsity regularization, which applies when the latent factors of interest depend sparsely on observed auxiliary variables and/or past latent factors.
arXiv:2608.29335v1 Announce Type: new Abstract: Latent generative models typically follow a two-stage pipeline, training a variational autoencoder for reconstruction and then a generative model on th...
arXiv:2607. 08254v1 Announce Type: new Abstract: Quantifying variability in a target population relative to a reference population is central to many scientific and clinical problems (e.