arXiv:2511.00925v1 Announce Type: new 
Abstract: The problem of zero-shot sketch-based image retrieval (ZS-SBIR) has achieved increasing attention due to its wide applications, e.g. e-commerce. Despite progress made in this field, previous works suffer from using imbalanced samples of modalities and inconsistent low-quality information during training, resulting in sub-optimal performance. Therefore, in this paper, we introduce an approach called Dynamic Multi-level Weighted Alignment Network for ZS-SBIR. It consists of three components: (i) a Uni-modal Feature Extraction Module that includes a CLIP text encoder and a ViT for extracting textual and visual tokens, (ii) a Cross-modal Multi-level Weighting Module that produces an alignment weight list by the local and global aggregation blocks to measure the aligning quality of sketch and image samples, (iii) a Weighted Quadruplet Loss Module aiming to improve the balance of domains in the triplet loss. Experiments on three benchmark datasets, i.e., Sketchy, TU-Berlin, and QuickDraw, show our method delivers superior performances over the state-of-the-art ZS-SBIR methods.

تم تقديم نهج جديد يسمى شبكة المحاذاة متعددة المستويات الديناميكية لمعالجة تحديات استرجاع الصور المستند إلى الرسومات في حالة عدم وجود بيانات (ZS-SBIR)، والذي يكتسب اهتمامًا بسبب تطبيقاته في التجارة الإلكترونية. تتناول هذه الطريقة المشكلات المتعلقة بنماذج العينات غير المتوازنة والمعلومات ذات الجودة المنخفضة التي أعاقت الجهود السابقة، مما يعد بتحسين الأداء في استرجاع الصور بناءً على الرسومات. هذه الخطوة مهمة لأنها قد تعزز تجارب المستخدمين في التسوق عبر الإنترنت وغيرها من المجالات التي تعتمد على البحث المرئي.

Se ha introducido un nuevo enfoque llamado Red de Alineación Ponderada Multinivel Dinámica para abordar los desafíos de la recuperación de imágenes basada en bocetos en cero disparos (ZS-SBIR), que está ganando atención debido a sus aplicaciones en el comercio electrónico. Este método aborda problemas relacionados con muestras de modalidades desequilibradas e información de baja calidad que han obstaculizado esfuerzos anteriores, prometiendo un mejor rendimiento en la recuperación de imágenes a partir de bocetos. Este avance es significativo ya que podría mejorar la experiencia del usuario en las compras en línea y otros campos que dependen de la búsqueda visual.

Une nouvelle approche appelée Réseau d'Alignement Pondéré Multi-niveaux Dynamique a été introduite pour relever les défis de la récupération d'images basée sur des croquis en zéro-shot (ZS-SBIR), qui gagne en popularité en raison de ses applications dans le commerce électronique. Cette méthode traite les problèmes liés aux modalités d'échantillons déséquilibrés et aux informations de faible qualité qui ont entravé les efforts précédents, promettant une meilleure performance dans la récupération d'images à partir de croquis. Cette avancée est significative car elle pourrait améliorer l'expérience utilisateur dans les achats en ligne et d'autres domaines reposant sur la recherche visuelle.

A new approach called Dynamic Multi-level Weighted Alignment Network has been introduced to tackle the challenges of zero-shot sketch-based image retrieval (ZS-SBIR), which is gaining traction due to its applications in e-commerce. This method addresses issues related to imbalanced sample modalities and low-quality information that have hindered previous efforts, promising improved performance in retrieving images based on sketches. This advancement is significant as it could enhance user experiences in online shopping and other fields relying on visual search.

Dynamic Multi-level Weighted Alignment Network for Zero-shot Sketch-based Image Retrieval

arXiv:2601.08621v1 Announce Type: new 
Abstract: Recent advances in search-augmented large reasoning models (LRMs) enable the retrieval of external knowledge to reduce hallucinations in multistep reasoning. However, their ability to operate on graph-structured data, prevalent in domains such as e-commerce, social networks, and scientific citations, remains underexplored. Unlike plain text corpora, graphs encode rich topological signals that connect related entities and can serve as valuable priors for retrieval, enabling more targeted search and improved reasoning efficiency. Yet, effectively leveraging such structure poses unique challenges, including the difficulty of generating graph-expressive queries and ensuring reliable retrieval that balances structural and semantic relevance. To address this gap, we introduce GraphSearch, the first framework that extends search-augmented reasoning to graph learning, enabling zero-shot graph learning without task-specific fine-tuning. GraphSearch combines a Graph-aware Query Planner, which disentangles search space (e.g., 1-hop, multi-hop, or global neighbors) from semantic queries, with a Graph-aware Retriever, which constructs candidate sets based on topology and ranks them using a hybrid scoring function. We further instantiate two traversal modes: GraphSearch-R, which recursively expands neighborhoods hop by hop, and GraphSearch-F, which flexibly retrieves across local and global neighborhoods without hop constraints. Extensive experiments across diverse benchmarks show that GraphSearch achieves competitive or even superior performance compared to supervised graph learning methods, setting state-of-the-art results in zero-shot node classification and link prediction. These findings position GraphSearch as a flexible and generalizable paradigm for agentic reasoning over graphs.

تم تقديم إطار عمل جديد يسمى GraphSearch، الذي يوسع التفكير المعزز بالبحث إلى تعلم الرسوم البيانية، مما يمكّن التعلم من الرسوم البيانية بدون ضبط محدد للمهمة. هذا التقدم يعالج التحديات المتعلقة بالتعامل مع البيانات الهيكلية في الرسوم البيانية، والتي تزداد شيوعًا في مجالات متنوعة مثل التجارة الإلكترونية والشبكات الاجتماعية.

Se ha introducido un nuevo marco llamado GraphSearch, que extiende el razonamiento aumentado por búsqueda al aprendizaje de grafos, permitiendo el aprendizaje de grafos sin ajuste específico a la tarea. Este avance aborda los desafíos de operar con datos estructurados en grafos, que son cada vez más prevalentes en diversos dominios como el comercio electrónico y las redes sociales.

Un nouveau cadre nommé GraphSearch a été introduit, étendant le raisonnement augmenté par la recherche à l'apprentissage des graphes, permettant un apprentissage de graphes sans coupure spécifique à la tâche. Cette avancée répond aux défis liés à l'exploitation des données structurées en graphes, de plus en plus présentes dans divers domaines tels que le commerce électronique et les réseaux sociaux.

A new framework named GraphSearch has been introduced, extending search-augmented reasoning to graph learning, enabling zero-shot graph learning without the need for task-specific fine-tuning. This advancement addresses the challenges of operating on graph-structured data, which is increasingly prevalent in various domains such as e-commerce and social networks.

GraphSearch: Agentic Search-Augmented Reasoning for Zero-Shot Graph Learning

One More Thing in AI – Your Shortcut to AI Mastery

Dynamic Multi-level Weighted Alignment Network for Zero-shot Sketch-based Image Retrieval

Was this article worth reading? Share it

One More Thing in AI

FaceSwapper AI Headshot Generator

Bulk Image Generation AI

Blunge

Lenso.ai

Zemith-3bda3b

Ready to build your own newsroom?