arXiv:2512.01242v1 Announce Type: new 
Abstract: We study abstract visual composition, in which identity is primarily determined by the spatial configuration and relations among a small set of geometric primitives (e.g., parts, symmetry, topology). They are invariant primarily to texture and photorealistic detail. Composing such structures from fixed components under geometric constraints and vague goal specification (such as text) is non-trivial due to combinatorial placement choices, limited data, and discrete feasibility (overlap-free, allowable orientations), which create a sparse solution manifold ill-suited to purely statistical pixel-space generators. We propose a constraint-guided framework that combines explicit geometric reasoning with neural semantics. An AlphaGo-style search enforces feasibility, while a fine-tuned vision-language model scores semantic alignment as reward signals. Our algorithm uses a policy network as a heuristic in Monte-Carlo Tree Search and fine-tunes the network via search-generated plans. Inspired by the Generative Adversarial Network, we use the generated instances for adversarial reward refinement. Over time, the generation should approach the actual data more closely when the reward model cannot distinguish between generated instances and ground-truth. In the Tangram Assembly task, our approach yields higher validity and semantic fidelity than diffusion and auto-regressive baselines, especially as constraints tighten.

تم اقتراح إطار عمل جديد يسمى Generative Adversarial Gumbel MCTS لتوليد التراكيب البصرية المجردة، مع التركيز على التكوين المكاني للبدائل الهندسية. تدمج هذه الطريقة بين التفكير الهندسي والدلالات العصبية، مستخدمةً نهج بحث شجرة مونت كارلو لفرض القابلية، ونموذج رؤية-لغة لتقييم توافق الدلالات.

Se ha propuesto un nuevo marco llamado Generative Adversarial Gumbel MCTS para la generación de composiciones visuales abstractas, centrándose en la configuración espacial de primitivas geométricas. Este método integra el razonamiento geométrico con la semántica neuronal, utilizando un enfoque de búsqueda de árbol de Monte-Carlo para hacer cumplir la viabilidad y un modelo de visión-lenguaje para puntuar la alineación semántica.

Un nouveau cadre appelé Generative Adversarial Gumbel MCTS a été proposé pour générer des compositions visuelles abstraites, en se concentrant sur la configuration spatiale de primitives géométriques. Cette méthode intègre le raisonnement géométrique avec la sémantique neuronale, utilisant une approche de recherche d'arbre de Monte-Carlo pour faire respecter la faisabilité et un modèle de vision-langage pour évaluer l'alignement sémantique.

A new framework called Generative Adversarial Gumbel MCTS has been proposed for generating abstract visual compositions, focusing on the spatial configuration of geometric primitives. This method integrates geometric reasoning with neural semantics, utilizing a Monte-Carlo Tree Search approach to enforce feasibility and a vision-language model for scoring semantic alignment.

Generative Adversarial Gumbel MCTS for Abstract Visual Composition Generation

arXiv:2601.08260v1 Announce Type: new 
Abstract: Cardiac amyloidosis (CA) is a rare and underdiagnosed infiltrative cardiomyopathy, and available datasets for machine-learning models are typically small, imbalanced and heterogeneous. This paper presents a Generative Adversarial Network (GAN) and a graphical command-line interface for generating realistic synthetic electrocardiogram (ECG) beats to support early diagnosis and patient stratification in CA. The tool is designed for usability, allowing clinical researchers to train class-specific generators once and then interactively produce large volumes of labelled synthetic beats that preserve the distribution of minority classes.

تم تطوير أداة جديدة تستخدم شبكة خصومة توليدية (GAN) لتوليد نبضات تخطيط القلب الكهربائي (ECG) الاصطناعية، بهدف دعم البحث في مرض الأميلويد القلبي، وهو حالة نادرة وغالبًا ما يتم تشخيصها بشكل خاطئ. تتيح هذه الأداة للباحثين السريريين إنشاء كميات كبيرة من بيانات ECG الاصطناعية المسمى التي تعكس توزيع الفئات الأقل، مما يعزز الإمكانية للتشخيص المبكر وتصنيف المرضى.

Se ha desarrollado una nueva herramienta que utiliza una Red Generativa Antagónica (GAN) para generar latidos de electrocardiograma (ECG) sintéticos, destinada a apoyar la investigación en amiloidosis cardíaca, una condición rara y a menudo mal diagnosticada. Esta herramienta permite a los investigadores clínicos crear grandes volúmenes de datos de ECG sintéticos etiquetados que reflejan la distribución de clases minoritarias, mejorando así el potencial para un diagnóstico temprano y la estratificación de pacientes.

Un nouvel outil utilisant un réseau antagoniste génératif (GAN) a été développé pour générer des battements d'électrocardiogramme (ECG) synthétiques, spécifiquement destiné à aider la recherche sur l'amylose cardiaque, une condition rare et souvent mal diagnostiquée. Cet outil permet aux chercheurs cliniques de créer de grands volumes de données ECG synthétiques étiquetées qui reflètent la distribution des classes minoritaires, améliorant ainsi le potentiel de diagnostic précoce et de stratification des patients.

A new tool utilizing a Generative Adversarial Network (GAN) has been developed for generating synthetic electrocardiogram (ECG) beats specifically aimed at aiding research in cardiac amyloidosis, a rare and often misdiagnosed condition. This tool allows clinical researchers to create large volumes of labeled synthetic ECG data that reflect the distribution of minority classes, enhancing the potential for early diagnosis and patient stratification.

A Usable GAN-Based Tool for Synthetic ECG Generation in Cardiac Amyloidosis Research

One More Thing in AI – Your Shortcut to AI Mastery

Generative Adversarial Gumbel MCTS for Abstract Visual Composition Generation

Was this article worth reading? Share it

One More Thing in AI

Golan AI

Glima

4o Image Gen

Republiclabs.ai

RandomGenerator.AI

Ready to build your own newsroom?