arXiv:2510.21088v1 Announce Type: new 
Abstract: Molecular property prediction (MPP) is a cornerstone of drug discovery and materials science, yet conventional deep learning approaches depend on large labeled datasets that are often unavailable. Few-shot Molecular property prediction (FSMPP) addresses this scarcity by incorporating relational inductive bias through a context graph that links molecule nodes to property nodes, but such molecule-property graphs offer limited structural guidance. We propose a comprehensive solution: Motif Driven Global-Local Context Graph for few-shot molecular property prediction, which enriches contextual information at both the global and local levels. At the global level, chemically meaningful motif nodes representing shared substructures, such as rings or functional groups, are introduced to form a global tri-partite heterogeneous graph, yielding motif-molecule-property connections that capture long-range compositional patterns and enable knowledge transfer among molecules with common motifs. At the local level, we build a subgraph for each node in the molecule-property pair and encode them separately to concentrate the model's attention on the most informative neighboring molecules and motifs. Experiments on five standard FSMPP benchmarks demonstrate that our framework consistently outperforms state-of-the-art methods. These results underscore the effectiveness of integrating global motif knowledge with fine-grained local context to advance robust few-shot molecular property prediction.

تُحدث طريقة جديدة لتوقع الخصائص الجزيئية (MPP) ضجة في اكتشاف الأدوية وعلوم المواد. تُعرف هذه الطريقة باسم توقع الخصائص الجزيئية القليلة (FSMPP)، وتتناول تحدي نقص مجموعات البيانات المعلّمة من خلال استخدام رسم بياني سياقي يربط بين العقد الجزيئية وعقد الخصائص. تُحسن هذه التقنية المبتكرة من عملية التوقع، مما يوفر حلاً واعدًا لعائق كبير في هذا المجال. مع استمرار الباحثين في استكشاف إمكانياتها، قد يُحدث FSMPP ثورة في فهمنا وتطوير مواد وأدوية جديدة.

Un nuevo enfoque para la predicción de propiedades moleculares (MPP) está causando revuelo en el descubrimiento de fármacos y la ciencia de materiales. El método, conocido como predicción de propiedades moleculares de pocos disparos (FSMPP), aborda el desafío de los conjuntos de datos etiquetados limitados mediante el uso de un grafo de contexto que conecta nodos moleculares con nodos de propiedad. Esta técnica innovadora mejora el proceso de predicción, ofreciendo una solución prometedora a un obstáculo significativo en el campo. A medida que los investigadores continúan explorando su potencial, el FSMPP podría revolucionar nuestra comprensión y desarrollo de nuevos materiales y medicamentos.

Une nouvelle approche de la prédiction des propriétés moléculaires (MPP) fait sensation dans la découverte de médicaments et la science des matériaux. La méthode, connue sous le nom de prédiction des propriétés moléculaires à faible échantillon (FSMPP), s'attaque au défi des ensembles de données étiquetées limités en utilisant un graphe contextuel qui relie les nœuds moléculaires aux nœuds de propriété. Cette technique innovante améliore le processus de prédiction, offrant une solution prometteuse à un obstacle majeur dans le domaine. Alors que les chercheurs continuent d'explorer son potentiel, le FSMPP pourrait révolutionner notre compréhension et le développement de nouveaux matériaux et médicaments.

A new approach to molecular property prediction (MPP) is making waves in drug discovery and materials science. The method, known as few-shot molecular property prediction (FSMPP), tackles the challenge of limited labeled datasets by using a context graph that connects molecule nodes to property nodes. This innovative technique enhances the prediction process, offering a promising solution to a significant hurdle in the field. As researchers continue to explore its potential, FSMPP could revolutionize how we understand and develop new materials and drugs.

M-GLC: Motif-Driven Global-Local Context Graphs for Few-shot Molecular Property Prediction

arXiv:2410.13597v3 Announce Type: replace 
Abstract: Molecular optimization (MO) is a crucial stage in drug discovery in which task-oriented generated molecules are optimized to meet practical industrial requirements. Existing mainstream MO approaches primarily utilize external property predictors to guide iterative property optimization. However, learning all molecular samples in the vast chemical space is unrealistic for predictors. As a result, errors and noise are inevitably introduced during property prediction due to the nature of approximation. This leads to discrepancy accumulation, generalization reduction and suboptimal molecular candidates. In this paper, we propose a text-guided multi-property molecular optimization method utilizing transformer-based diffusion language model (TransDLM). TransDLM leverages standardized chemical nomenclature as semantic representations of molecules and implicitly embeds property requirements into textual descriptions, thereby mitigating error propagation during diffusion process. By fusing physically and chemically detailed textual semantics with specialized molecular representations, TransDLM effectively integrates diverse information sources to guide precise optimization, which enhances the model's ability to balance structural retention and property enhancement. Additionally, the success of a case study further demonstrates TransDLM's ability to solve practical problems. Experimentally, our approach surpasses state-of-the-art methods in maintaining molecular structural similarity and enhancing chemical properties on the benchmark dataset.

تم اقتراح طريقة جديدة لتحسين الجزيئات (MO) باستخدام نموذج لغة انتشار قائم على المحولات (TransDLM) لتحسين عمليات اكتشاف الأدوية. تتناول هذه الطريقة قيود تقنيات MO الحالية التي تعتمد على متنبئين خارجيين للخصائص، مما يؤدي غالبًا إلى إدخال أخطاء وضوضاء في تنبؤات الخصائص، مما يؤدي إلى مرشحين جزيئيين دون المستوى الأمثل.

Se ha propuesto un nuevo método para la optimización molecular (OM), utilizando un modelo de lenguaje de difusión basado en transformadores (TransDLM) para mejorar los procesos de descubrimiento de fármacos. Este enfoque aborda las limitaciones de las técnicas de OM existentes que dependen de predictores de propiedades externas, que a menudo introducen errores y ruido en las predicciones de propiedades, lo que lleva a candidatos moleculares subóptimos.

Une nouvelle méthode d'optimisation moléculaire (OM) a été proposée, utilisant un modèle de langage de diffusion basé sur des transformateurs (TransDLM) pour améliorer les processus de découverte de médicaments. Cette approche répond aux limites des techniques d'OM existantes qui s'appuient sur des prédicteurs de propriétés externes, souvent sources d'erreurs et de bruit dans les prédictions de propriétés, conduisant à des candidats moléculaires sous-optimaux.

A new method for molecular optimization (MO) has been proposed, utilizing a transformer-based diffusion language model (TransDLM) to enhance drug discovery processes. This approach addresses the limitations of existing MO techniques that rely on external property predictors, which often introduce errors and noise in property predictions, leading to suboptimal molecular candidates.

Text-guided multi-property molecular optimization with a diffusion language model

arXiv:2511.16541v2 Announce Type: replace 
Abstract: The rapid advancement of generative artificial intelligence has enabled the creation of synthetic images that are increasingly indistinguishable from authentic content, posing significant challenges for digital media integrity. This problem is compounded by the accelerated release cycle of novel generative models, which renders traditional detection approaches (reliant on periodic retraining) computationally infeasible and operationally impractical.
  This work proposes a novel two-stage detection framework designed to address the generalization challenge inherent in synthetic image detection. The first stage employs a vision deep learning model trained via supervised contrastive learning to extract discriminative embeddings from input imagery. Critically, this model was trained on a strategically partitioned subset of available generators, with specific architectures withheld from training to rigorously ablate cross-generator generalization capabilities. The second stage utilizes a k-nearest neighbors (k-NN) classifier operating on the learned embedding space, trained in a few-shot learning paradigm incorporating limited samples from previously unseen test generators.
  With merely 150 images per class in the few-shot learning regime, which are easily obtainable from current generation models, the proposed framework achieves an average detection accuracy of 91.3%, representing a 5.2 percentage point improvement over existing approaches . For the source attribution task, the proposed approach obtains improvements of of 14.70% and 4.27% in AUC and OSCR respectively on an open set classification context, marking a significant advancement toward robust, scalable forensic attribution systems capable of adapting to the evolving generative AI landscape without requiring exhaustive retraining protocols.

تم اقتراح إطار جديد للكشف على مرحلتين لمواجهة تحديات الكشف عن الصور التي تم إنشاؤها بواسطة الذكاء الاصطناعي، والتي أصبحت متشابهة بشكل متزايد مع المحتوى الحقيقي بسبب التقدم في الذكاء الاصطناعي التوليدي. تستخدم المرحلة الأولى نموذج تعلم عميق بصري تم تدريبه من خلال التعلم التبايني المراقب لاستخراج تمثيلات تمييزية من الصور، مما يعالج تحدي التعميم في الكشف عن الصور الاصطناعية.

Se ha propuesto un nuevo marco de detección en dos etapas para abordar los desafíos de la detección de imágenes generadas por IA, que se han vuelto cada vez más indistinguibles del contenido real debido a los avances en inteligencia artificial generativa. La primera etapa utiliza un modelo de aprendizaje profundo visual entrenado mediante aprendizaje contrastivo supervisado para extraer incrustaciones discriminativas de las imágenes, abordando el desafío de la generalización en la detección de imágenes sintéticas.

Un nouveau cadre de détection en deux étapes a été proposé pour relever les défis de la détection des images générées par l'IA, qui sont devenues de plus en plus indistinguables du contenu réel en raison des avancées de l'intelligence artificielle générative. La première étape utilise un modèle d'apprentissage profond visuel formé par apprentissage contrastif supervisé pour extraire des embeddings discriminants des images, abordant le défi de la généralisation dans la détection d'images synthétiques.

A novel two-stage detection framework has been proposed to tackle the challenges of detecting AI-generated images, which have become increasingly indistinguishable from real content due to advancements in generative artificial intelligence. The first stage utilizes a vision deep learning model trained through supervised contrastive learning to extract discriminative embeddings from images, addressing the generalization challenge in synthetic image detection.

M-GLC: Motif-Driven Global-Local Context Graphs for Few-shot Molecular Property Prediction

Was this article worth reading? Share it

Cogent

ZeroGPT.org

Langfuse