The rapid growth of digital communication has increased the demand for sticker retrieval systems that can match expressive stickers to users' communicative needs. In practice, however, sticker retrieval encounters significant out-of-distribution (OOD) challenges arising from unseen queries and stickers, driven by the diversity of user expression habits and sticker visual representations. These OOD issues often lead to irrelevant or inappropriate retrieval results, undermining the user experience. Drawing on symbolic interactionism in cognition, we propose XAlign-SR, a method that enhances OOD robustness by aligning abstract expressive intent between queries and stickers across modalities. To support this study, we construct OOD benchmarks from sticker datasets that simulate realistic query–sticker scenarios. Experiments demonstrate that our approach significantly outperforms state-of-the-art baselines.

Large language models (LLMs) have demonstrated significant potential in listwise document reranking. Due to their limited context length, LLM-based listwise reranking methods often rely on a sliding window strategy that only processes a small subset of documents at a time. While effective, this approach lacks interactions between documents, increases computational overhead, and results in significant API costs. It is crucial to develop long-context LLMs for enabling the full ranking of all documents in one pass.

The emergence of spatial transcriptomics has greatly advanced our understanding of disease mechanisms, identified novel therapeutic targets, and contributed to progress in personalized medicine. However, its high costs and technical complexity limit its widespread application. A promising alternative is to predict spatial gene expression from H&E-stained pathology images, yet existing methods have not fully exploited the hierarchical information from pathological images. We propose HFFST, a hierarchical feature fusion algorithm for predicting spatial gene expression from H&E-stained pathology images. Leveraging multi-level feature extraction and fusion from whole-slide images, our method employs a coarse-to-fine regression framework to predict spatial transcriptomic profiles. To validate the algorithm's performance, we conducted cross-validation on five public datasets and performed external validation using high-resolution Visium data from 10X Genomics. HFFST has shown promise in predicting spatial gene expression and identifying spatial regions, demonstrating certain advantages over state-of-the-art methods.

Knowledge-intensive language tasks (KILTs) typically require retrieving relevant documents from trustworthy corpora, e.g., Wikipedia, to produce specific answers. Very recently, a pre-trained generative retrieval model for KILTs, named CorpusBrain, was proposed and reached new state-of-the-art retrieval performance. However, most research on KILTs, including CorpusBrain, has predominantly focused on a static document collection, overlooking the dynamic nature of real-world scenarios, where new documents are continuously being incorporated into the source corpus. To address this gap, it is crucial to explore the capability of retrieval models to effectively handle the dynamic retrieval scenario inherent in KILTs. In this work, we first introduce the continual document learning (CDL) task for KILTs and build a novel benchmark dataset named KILT++ based on the original KILT dataset for evaluation. Then, we conduct a comprehensive study of the use of pre-trained CorpusBrain on KILT++. Unlike the promising results in the stationary scenario, CorpusBrain is prone to catastrophic forgetting in the dynamic scenario, hence hampering retrieval performance. To alleviate this issue, we propose CorpusBrain++, a continual generative pre-training framework that enhances the original model along two key dimensions: (i) We employ a backbone-adapter architecture: the dynamic adapter is learned for each downstream KILT task via task-specific pre-training objectives; the backbone parameters that are task-shared are kept unchanged to offer foundational retrieval capacity. (ii) We use an experience replay strategy based on exemplar documents that are similar to new documents, to prevent catastrophic forgetting of old documents. Empirical results demonstrate the effectiveness and efficiency of CorpusBrain++ in comparison to both traditional and generative information retrieval methods.

In the practice of clinical endoscopy, the precise estimation of the lesion size is quite significant for diagnosis. In this paper, we propose a three-dimensional (3D) measurement method for binocular endoscopes based on deep learning, which can overcome the poor robustness of the traditional binocular matching algorithm in texture-less areas. A simulated binocular image dataset is created from the target 3D data obtained by a 3D scanner and the binocular camera is simulated by 3D rendering software to train a disparity estimation model for 3D measurement. The experimental results demonstrate that, compared with the traditional binocular matching algorithm, the proposed method improves the accuracy and disparity map generation speed by 48.9% and 90.5%, respectively. This can provide more accurate and reliable lesion size and improve the efficiency of endoscopic diagnosis.