arXiv:2505.17895v2 Announce Type: replace 
Abstract: The quality of foundation models depends heavily on their training data. Consequently, great efforts have been put into dataset curation. Yet most approaches rely on manual tuning of coarse-grained mixtures of large buckets of data, or filtering by hand-crafted heuristics. An approach that is ultimately more scalable (let alone more satisfying) is to \emph{learn} which data is actually valuable for training. This type of meta-learning could allow more sophisticated, fine-grained, and effective curation. Our proposed \emph{DataRater} is an instance of this idea. It estimates the value of training on any particular data point. This is done by meta-learning using `meta-gradients', with the objective of improving training efficiency on held out data. In extensive experiments across a range of model scales and datasets, we find that using our DataRater to filter data is highly effective, resulting in significantly improved compute efficiency.

تسلط دراسة حديثة الضوء على أهمية تنسيق مجموعات البيانات لجودة النماذج الأساسية، مقترحةً نهج التعلم الميتا الذي يقوم بأتمتة تحديد البيانات القيمة للتدريب. تعد هذه الطريقة بتحسين القابلية للتوسع والكفاءة في تدريب النماذج، مبتعدةً عن الأساليب التقليدية للتعديل اليدوي. مع تزايد الطلب على نماذج الذكاء الاصطناعي عالية الجودة، قد يكون لهذا النهج المبتكر تأثير كبير على المجال، مما يسهل على الباحثين والمطورين إنشاء أنظمة ذكاء اصطناعي أكثر فعالية.

Un estudio reciente destaca la importancia de la curación de conjuntos de datos para la calidad de los modelos fundamentales, proponiendo un enfoque de aprendizaje meta que automatiza la identificación de datos de entrenamiento valiosos. Este método promete mejorar la escalabilidad y la eficiencia en el entrenamiento de modelos, alejándose de los métodos tradicionales de ajuste manual. A medida que crece la demanda de modelos de IA de alta calidad, este enfoque innovador podría tener un impacto significativo en el campo, facilitando a investigadores y desarrolladores la creación de sistemas de IA más efectivos.

Une étude récente souligne l'importance de la curation des ensembles de données pour la qualité des modèles de base, en proposant une approche d'apprentissage méta qui automatise l'identification des données d'entraînement précieuses. Cette méthode promet d'améliorer l'évolutivité et l'efficacité de l'entraînement des modèles, s'éloignant des méthodes traditionnelles de réglage manuel. Alors que la demande de modèles d'IA de haute qualité augmente, cette approche innovante pourrait avoir un impact significatif sur le domaine, facilitant la création de systèmes d'IA plus efficaces pour les chercheurs et les développeurs.

A recent study highlights the importance of dataset curation for the quality of foundation models, proposing a meta-learning approach that automates the identification of valuable training data. This method promises to enhance scalability and efficiency in model training, moving away from traditional manual tuning methods. As the demand for high-quality AI models grows, this innovative approach could significantly impact the field, making it easier for researchers and developers to create more effective AI systems.

DataRater: Meta-Learned Dataset Curation

arXiv:2511.18677v1 Announce Type: new 
Abstract: Sketch based person re-identification aims to match hand-drawn sketches with RGB surveillance images, but remains challenging due to significant modality gaps and limited annotated data. To address this, we introduce KTCAA, a theoretically grounded framework for few-shot cross-modal generalization. Motivated by generalization theory, we identify two key factors influencing target domain risk: (1) domain discrepancy, which quantifies the alignment difficulty between source and target distributions; and (2) perturbation invariance, which evaluates the model's robustness to modality shifts. Based on these insights, we propose two components: (1) Alignment Augmentation (AA), which applies localized sketch-style transformations to simulate target distributions and facilitate progressive alignment; and (2) Knowledge Transfer Catalyst (KTC), which enhances invariance by introducing worst-case perturbations and enforcing consistency. These modules are jointly optimized under a meta-learning paradigm that transfers alignment knowledge from data-rich RGB domains to sketch-based scenarios. Experiments on multiple benchmarks demonstrate that KTCAA achieves state-of-the-art performance, particularly in data-scarce conditions.

تم تقديم إطار جديد يسمى KTCAA لإعادة تحديد الهوية للأشخاص من خلال الرسوم التخطيطية المتقاطعة مع عدد قليل من الأمثلة، بهدف سد الفجوة بين الرسوم التخطيطية المرسومة باليد وصور المراقبة RGB. يتناول هذا الإطار التحديات المتعلقة باختلاف المجال وثبات الاضطرابات، ويقترح مكونات مبتكرة مثل تعزيز المحاذاة ومحفز نقل المعرفة لتحسين متانة النموذج وقدرات المحاذاة.

Se ha introducido un nuevo marco llamado KTCAA para la reidentificación de personas a partir de bocetos cruzados con pocos ejemplos, con el objetivo de cerrar la brecha entre los bocetos dibujados a mano y las imágenes de vigilancia RGB. Este marco aborda los desafíos relacionados con la discrepancia de dominio y la invariancia a perturbaciones, proponiendo componentes innovadores como la Aumento de Alineación y el Catalizador de Transferencia de Conocimiento para mejorar la robustez y las capacidades de alineación del modelo.

Un nouveau cadre appelé KTCAA a été introduit pour la ré-identification de personnes à partir de croquis croisés avec peu d'exemples, visant à combler le fossé entre les croquis dessinés à la main et les images de surveillance RGB. Ce cadre aborde les défis liés à la différence de domaine et à l'invariance aux perturbations, en proposant des composants innovants tels que l'Augmentation d'Alignement et le Catalyseur de Transfert de Connaissances pour améliorer la robustesse et les capacités d'alignement du modèle.

A new framework called KTCAA has been introduced for few-shot cross-modal sketch person re-identification, aiming to bridge the gap between hand-drawn sketches and RGB surveillance images. This framework addresses challenges related to domain discrepancy and perturbation invariance, proposing innovative components like Alignment Augmentation and Knowledge Transfer Catalyst to enhance model robustness and alignment capabilities.

A Theory-Inspired Framework for Few-Shot Cross-Modal Sketch Person Re-Identification

arXiv:2505.11579v2 Announce Type: replace-cross 
Abstract: As AI systems evolve from static tools to dynamic agents, traditional categorical governance frameworks -- based on fixed risk tiers, levels of autonomy, or human oversight models -- are increasingly insufficient on their own. Systems built on foundation models, self-supervised learning, and multi-agent architectures increasingly blur the boundaries that categories were designed to police. In this Perspective, we make the case for dimensional governance: a framework that tracks how decision authority, process autonomy, and accountability (the 3As) distribute dynamically across human-AI relationships. A critical advantage of this approach is its ability to explicitly monitor system movement toward and across key governance thresholds, enabling preemptive adjustments before risks materialize. This dimensional approach provides the necessary foundation for more adaptive categorization, enabling thresholds and classifications that can evolve with emerging capabilities. While categories remain essential for decision-making, building them upon dimensional foundations allows for context-specific adaptability and stakeholder-responsive governance that static approaches cannot achieve. We outline key dimensions, critical trust thresholds, and practical examples illustrating where rigid categorical frameworks fail -- and where a dimensional mindset could offer a more resilient and future-proof path forward for both governance and innovation at the frontier of artificial intelligence.

تطور أنظمة الذكاء الاصطناعي من أدوات ثابتة إلى وكلاء ديناميكيين يتطلب تغييرًا في أطر الحوكمة، حيث أصبحت النماذج الفئوية التقليدية غير كافية بشكل متزايد. يركز إطار الحوكمة البُعدية المقترح على التوزيع الديناميكي لسلطة القرار، واستقلالية العمليات، والمساءلة في العلاقات بين الإنسان والذكاء الاصطناعي، بهدف معالجة المخاطر بشكل استباقي قبل أن تتجلى.

La evolución de los sistemas de IA de herramientas estáticas a agentes dinámicos requiere un cambio en los marcos de gobernanza, ya que los modelos categóricos tradicionales son cada vez más inadecuados. El marco de gobernanza dimensional propuesto se centra en la distribución dinámica de la autoridad de decisión, la autonomía del proceso y la responsabilidad en las relaciones humano-IA, con el objetivo de abordar proactivamente los riesgos antes de que se materialicen.

L'évolution des systèmes d'IA, passant d'outils statiques à des agents dynamiques, nécessite un changement dans les cadres de gouvernance, car les modèles catégoriels traditionnels deviennent de plus en plus inadéquats. Le cadre de gouvernance dimensionnelle proposé se concentre sur la distribution dynamique de l'autorité décisionnelle, de l'autonomie des processus et de la responsabilité dans les relations homme-IA, visant à anticiper les risques avant qu'ils ne se matérialisent.

The evolution of AI systems from static tools to dynamic agents necessitates a shift in governance frameworks, as traditional categorical models are increasingly inadequate. The proposed dimensional governance framework focuses on the dynamic distribution of decision authority, process autonomy, and accountability in human-AI relationships, aiming to preemptively address risks before they materialize.

DataRater: Meta-Learned Dataset Curation

Was this article worth reading? Share it

Augmeta

Artefacts.ai

Superset