As digital transformation accelerates, data has become a core driver of technological innovation and economic growth. However, a key challenge in data utilization is the semantic misalignment between data supply and the demands of business scenarios. This misalignment significantly hinders efficient data flow and collaborative utilization. To address this issue, this article proposes a federated service solution integrating blockchain and decentralized autonomous organizations and operations (DAOs), large language models (LLMs) and scenarios engineering, federated learning and edge computing, as well as encryption technologies and privacy-computing. A five-layer federated service framework is introduced, consisting of the foundation layer, the data-scenario layer, the semantic coordination layer, the incentive-security layer, and the application layer, which is designed to ensure efficient and context-aware data supply–demand matching while preserving privacy and scalability. Moreover, the core mechanisms for semantic coordination are proposed, and a detailed solution process for resolving semantic misalignment with these mechanisms, as well as an illustrative example, is also presented. The proposed federated service framework offers an effective solution to semantic misalignment in supply–demand matching, fostering seamless data collaboration across diverse business scenarios. This work provides an intelligent service paradigm that leverages distributed data co-governance to address semantic challenges in the digital economy.

In real-world applications, detectors are expected to evolve and improve their perceptual abilities through incremental learning of the unknown. Open world object detection (OWOD) simulates this process by recognizing unknown objects and incrementally learning these unknown classes as labeled data becomes available. However, two issues exist in the training images of OWOD datasets: instances of both 1) unknown classes and 2) previously known classes are present. To solve them, this paper proposes Open-Set Incremental Object Detection (OSIOD), which is defined as two processes: learning from the known, and an infinite cyclical process consisting of detection of the known and the unknown, together with incremental learning of the unknown. We construct a benchmark dataset by filtering out images containing unknown and/or previously known instances, which can also be used to create a new incremental object detection scenario where images with only new classes are used for incremental learning. Additionally, we propose a baseline method for OSIOD by introducing label smoothing and finetuning for unknown class discovery and catastrophic forgetting, respectively. Comprehensive experiments provide insights into the baseline. We hope that our benchmark, baseline, and insights will promote research towards OSIOD.

Partially annotated images are easy to obtain in multi-label classification. However, unknown labels in partially annotated images exacerbate the positive-negative imbalance inherent in multi-label classification, which affects supervised learning of known labels. Most current methods require sufficient image annotations, and do not focus on the imbalance of the labels in the supervised training phase. In this paper, we propose saliency regularization (SR) for a novel self-training framework. In particular, we model saliency on the class-specific maps, and strengthen the saliency of object regions corresponding to the present labels. Besides, we introduce consistency regularization to mine unlabeled information to complement unknown labels with the help of SR. It is verified to alleviate the negative dominance caused by the imbalance, and achieve state-of-the-art performance on Pascal VOC 2007, MS-COCO, VG-200, and OpenImages V3.

Statistical analysis of alignments of large numbers of protein sequences has revealed "sectors" of collectively coevolving amino acids in several protein families. Here, we show that selection acting on any functional property of a protein, represented by an additive trait, can give rise to such a sector. As an illustration of a selected trait, we consider the elastic energy of an important conformational change within an elastic network model, and we show that selection acting on this energy leads to correlations among residues. For this concrete example and more generally, we demonstrate that the main signature of functional sectors lies in the small-eigenvalue modes of the covariance matrix of the selected sequences. However, secondary signatures of these functional sectors also exist in the extensively-studied large-eigenvalue modes. Our simple, general model leads us to propose a principled method to identify functional sectors, along with the magnitudes of mutational effects, from sequence data. We further demonstrate the robustness of these functional sectors to various forms of selection, and the robustness of our approach to the identification of multiple selected traits.