arXiv:2510.21204v1 Announce Type: new 
Abstract: Since the seminal work of TabPFN, research on tabular foundation models (TFMs) based on in-context learning (ICL) has challenged long-standing paradigms in machine learning. Without seeing any real-world data, models pretrained on purely synthetic datasets generalize remarkably well across diverse datasets, often using only a moderate number of in-context examples. This shifts the focus in tabular machine learning from model architecture design to the design of synthetic datasets, or, more precisely, to the prior distributions that generate them. Yet the guiding principles for prior design remain poorly understood. This work marks the first attempt to address the gap. We systematically investigate and identify key properties of synthetic priors that allow pretrained TFMs to generalize well. Based on these insights, we introduce Mitra, a TFM trained on a curated mixture of synthetic priors selected for their diversity, distinctiveness, and performance on real-world tabular data. Mitra consistently outperforms state-of-the-art TFMs, such as TabPFNv2 and TabICL, across both classification and regression benchmarks, with better sample efficiency.

تقدم دراسة حديثة مفهوم Mitra، وهو نهج جديد لتعزيز نماذج الأساس الجدولية (TFMs) باستخدام أولويات اصطناعية مختلطة. تستند هذه الأبحاث إلى العمل الرائد لـ TabPFN، حيث تُظهر أن النماذج المدربة على مجموعات بيانات اصطناعية يمكن أن تعمم بفعالية على السيناريوهات الواقعية دون التعرض المسبق للبيانات الحقيقية. هذا التطور مهم لأنه يغير التركيز في تعلم الآلة الجدولي من تحسين هياكل النماذج إلى استغلال تقنيات تدريب مبتكرة، مما قد يؤدي إلى تطبيقات تعلم آلة أكثر كفاءة.

Un estudio reciente presenta Mitra, un nuevo enfoque para mejorar los modelos de fundación tabular (TFM) utilizando priors sintéticos mixtos. Esta investigación se basa en el trabajo pionero de TabPFN, demostrando que los modelos entrenados con conjuntos de datos sintéticos pueden generalizar de manera efectiva a escenarios del mundo real sin haber estado expuestos previamente a datos reales. Este desarrollo es significativo ya que cambia el enfoque en el aprendizaje automático tabular de la mejora de arquitecturas de modelos a la utilización de técnicas de entrenamiento innovadoras, lo que podría llevar a aplicaciones de aprendizaje automático más eficientes.

Une étude récente présente Mitra, une nouvelle approche pour améliorer les modèles de fondation tabulaire (TFM) en utilisant des priors synthétiques mixtes. Cette recherche s'appuie sur le travail révolutionnaire de TabPFN, démontrant que les modèles formés sur des ensembles de données synthétiques peuvent se généraliser efficacement à des scénarios réels sans exposition préalable à des données réelles. Ce développement est significatif car il déplace l'accent dans l'apprentissage automatique tabulaire de l'amélioration des architectures de modèles vers l'exploitation de techniques d'entraînement innovantes, ce qui pourrait conduire à des applications d'apprentissage automatique plus efficaces.

A recent study introduces Mitra, a new approach to enhancing tabular foundation models (TFMs) using mixed synthetic priors. This research builds on the groundbreaking work of TabPFN, demonstrating that models trained on synthetic datasets can effectively generalize to real-world scenarios without prior exposure to actual data. This development is significant as it shifts the focus in tabular machine learning from merely improving model architectures to leveraging innovative training techniques, potentially leading to more efficient and effective machine learning applications.

Mitra: Mixed Synthetic Priors for Enhancing Tabular Foundation Models

arXiv:2512.06154v1 Announce Type: new 
Abstract: Learning invariant graph representations for out-of-distribution (OOD) generalization remains challenging because the learned representations often retain spurious components. To address this challenge, this work introduces a new tool from information theory called Partial Information Decomposition (PID) that goes beyond classical information-theoretic measures. We identify limitations in existing approaches for invariant representation learning that solely rely on classical information-theoretic measures, motivating the need to precisely focus on redundant information about the target $Y$ shared between spurious subgraphs $G_s$ and invariant subgraphs $G_c$ obtained via PID. Next, we propose a new multi-level optimization framework that we call -- Redundancy-guided Invariant Graph learning (RIG) -- that maximizes redundant information while isolating spurious and causal subgraphs, enabling OOD generalization under diverse distribution shifts. Our approach relies on alternating between estimating a lower bound of redundant information (which itself requires an optimization) and maximizing it along with additional objectives. Experiments on both synthetic and real-world graph datasets demonstrate the generalization capabilities of our proposed RIG framework.

تقدم دراسة جديدة إطارًا يسمى التعلم البياني invariant-guided بواسطة المعلومات الزائدة (RIG)، والذي يستخدم تحليل المعلومات الجزئية (PID) لتحسين التعميم خارج التوزيع (OOD) في تعلم تمثيلات الرسوم البيانية. تهدف هذه الطريقة إلى التخفيف من احتفاظ التمثيلات المتعلمة بالمكونات الزائفة من خلال تعظيم المعلومات الزائدة مع عزل المجموعات السببية.

Un nuevo estudio presenta un marco denominado aprendizaje gráfico invariante guiado por la redundancia (RIG), que utiliza la descomposición de información parcial (PID) para mejorar la generalización fuera de distribución (OOD) en el aprendizaje de representaciones gráficas. Este enfoque busca mitigar la retención de componentes espurios en las representaciones aprendidas al maximizar la información redundante mientras aísla los subgrupos causales.

Une nouvelle étude présente un cadre appelé apprentissage graphique invariant guidé par la redondance (RIG), qui utilise la décomposition d'information partielle (PID) pour améliorer la généralisation hors distribution (OOD) dans l'apprentissage des représentations graphiques. Cette approche vise à atténuer la rétention de composants fallacieux dans les représentations apprises en maximisant l'information redondante tout en isolant les sous-graphes causaux.

A new study introduces a framework called Redundancy-guided Invariant Graph learning (RIG), which utilizes Partial Information Decomposition (PID) to enhance out-of-distribution (OOD) generalization in graph representation learning. This approach aims to mitigate the retention of spurious components in learned representations by maximizing redundant information while isolating causal subgraphs.

Mitra: Mixed Synthetic Priors for Enhancing Tabular Foundation Models

Was this article worth reading? Share it

LucidQuery AI

One More Thing in AI

Zemith-3bda3b