Current image-text retrieval methods have demonstrated impressive performance in recent years. However, they still face two problems: the inter-modal matching missing problem and intra-modal semantic loss problem. These problems can significantly affect accuracy of retrieval. To address these challenges, we propose a novel method called Cross-modal Uni-modal Soft-label Alignment (CUSA). Our leverages power uni-modal pre-trained models to provide soft-label supervision signals for model....

Weakly supervised phrase grounding tasks aim to learn alignments between phrases and regions with coarse image-caption match information. One branch of previous methods established pseudo-label relationships based on the Expectation-Maximization (EM) algorithm combined contrastive learning. However, adopting a simplified batch-level local update (partial) pseudo-labels in E-step is sub-optimal, while extending it global requires inefficiently numerous computations. In addition, their failure...

Current methods for Knowledge-Based Question Answering (KBQA) usually rely on complex training techniques and model frameworks, leading to many limitations in practical applications. Recently, the emergence of In-Context Learning (ICL) capabilities Large Language Models (LLMs) provides a simple training-free semantic parsing paradigm KBQA: Given small number questions their labeled logical forms as demo examples, LLMs can understand task intent generate logic form new question. However,...

According to the proposed definition and classification of power system stability addressed by IEEE CIGRE Task Force, voltage refers maintaining steady magnitudes at all buses in a when is subjected disturbance from given operating condition (OC). Cascading outage due collapse probable consequence during insecure situations. In this regard, fast responding reliable security assessment (VSA) effective indispensable for survive conceivable contingencies. This paper aims establishing an online...

Abstract Background Generic drugs have been seen as a potentially powerful way to alleviate the financial burden on patients and health care systems. Two strategies for achieving rational prices of generic are tiered pricing framework pooled purchasing power. We compare pan-Canadian Tiered Pricing Framework (TPF) Chinese National Volume-Based Procurement (NVBP) comparators explore similarities differences between two mechanisms summarise lessons other jurisdictions. Methods This comparative...

Sentence Representation Learning (SRL) is a fundamental task in Natural Language Processing (NLP), with the Contrastive of Embeddings (CSE) being mainstream technique due to its superior performance. An intriguing phenomenon CSE significant performance gap between supervised and unsupervised methods, their only difference lying training data. Previous works attribute this differences two representation properties (alignment uniformity). However, since alignment uniformity measure results,...

With the push for higher efficiency and reliability, an increasing number of intelligent electronic devices (IEDs) associated information communication technology (ICT) are integrated into Internet Things (IoT)-enabled smart grid. These advanced technologies IEDs also bring potential vulnerabilities to cyber–physical State estimation, as a primary step system monitoring situational awareness, is target attackers. A other grid applications, such voltage stability assessment contingency...

Dynamic object roles are widely considered to be helpful conceptual modeling of application. This paper presents an approach role implementation based on mediator pattern, which is used behavior extension in class-based language. We present a prototypical the by extending Java language, called Rava. In approach, management (generation object, dynamic binding core and etc.) implemented reduces complexity client program. At same time, relationship between also saved coupling improving their...

Phasor measurement unit (PMU) testing is desired to ensure the continued power system monitoring and maneuvering precision of synchrophasor-based control. IEEE Std C37.118 synchrophasor test suite specification (TSS) clarify different requirements for both steady-state dynamic scenarios with varied error ranges. Existing PMU requires taking PMUs out service in lab typically manually. This article presents an architecture remote utilizing performance analyzer (PPA) suite, which can...

The continuing success of synchrophasors has ushered in new subdomains power system applications for real-time situational awareness, online decision support, and robust control. In this paper, an adaptive decision-tree-based systematic method open-loop regional voltage control is developed. This approach employs security assessment to generate secure insecure operating conditions tree learning. Parallel trees corresponding a set combinations are trained. To guarantee the robustness its...

This paper introduces a resiliency quantification approach for transmission power system, that incorporates decentralized Remedial Action Scheme (RAS). RAS is designed to alleviate line overloads under high wind conditions. A metric computed quantify the ability of system supply critical loads in presence RAS. First, de-centralized curtail excess optimally, order limit flow on lines. Next, effect operation quantified using proposed metric. The based configuration and real-time measurements...

Enhanced automation and control requires advanced sensors in the smart electric power grid. Phasor Measurement Unit (PMU) is one of with fastest streaming data enabling algorithms After PMUs are initially tested, installed, brought into service field, there no established procedure for in-field calibration testing following IEEE Synchrophasor Test Suite Specification (TSS). Hence, remote PMU necessary to guarantee that performs accurately when needed not affected any software/hardware error...

The power grid Is becoming Increasingly complex with multi-domain and multi-physics interaction given enhanced automation, increasing DERs, active distribution system push for resiliency. centralized control such a will be slow, non-scalable prone to failures. local controllers non-optimal, hard coded not fault-tolerant. preferred architecture distributed as it is relatively fast, scalable robust. supports the monitoring resiliency reliability, but need tested validated before field...

In recent years, the cyber and physical extreme events have increased impacted power system operations. Although there are multiple work reported for improving resiliency of grid systems, a limited number management tools available to operators. Addressing data quality issue is critical before feeding measurements situational awareness decision-making using tools. this work, we describe an automated ML-based measurement anomaly mitigation technique that uses regression, clustering, deep...

The Composed Image Retrieval (CIR) task aims to retrieve target images using a composed query consisting of reference image and modified text. Advanced methods often utilize contrastive learning as the optimization objective, which benefits from adequate positive negative examples. However, triplet for CIR incurs high manual annotation costs, resulting in limited Furthermore, existing commonly use in-batch sampling, reduces number available model. To address problem lack positives, we...

Current methods for Knowledge-Based Question Answering (KBQA) usually rely on complex training techniques and model frameworks, leading to many limitations in practical applications. Recently, the emergence of In-Context Learning (ICL) capabilities Large Language Models (LLMs) provides a simple training-free semantic parsing paradigm KBQA: Given small number questions their labeled logical forms as demo examples, LLMs can understand task intent generate logic form new question. However,...

Background The Chinese government has ended the “dynamic zero-COVID” policy, and residents are now living together with SARS-CoV-2 virus. Only a limited number of studies have investigated specific content structure COVID-19-related risk perceptions, as well their underlying determinants. This study measured residents’ perception COVID-19 analyzed predictors RP. Methods We conducted comprehensive questionnaire-based survey among mostly in Beijing, using specially designed scale consisting 11...

Sentence Representation Learning (SRL) is a crucial task in Natural Language Processing (NLP), where contrastive Self-Supervised (SSL) currently mainstream approach. However, the reasons behind its remarkable effectiveness remain unclear. Specifically, other research fields, SSL shares similarities both theory and practical performance with non-contrastive (e.g., alignment & uniformity, Barlow Twins, VICReg). SRL, outperforms significantly. Therefore, two questions arise: First, what...

Current image-text retrieval methods have demonstrated impressive performance in recent years. However, they still face two problems: the inter-modal matching missing problem and intra-modal semantic loss problem. These problems can significantly affect accuracy of retrieval. To address these challenges, we propose a novel method called Cross-modal Uni-modal Soft-label Alignment (CUSA). Our leverages power uni-modal pre-trained models to provide soft-label supervision signals for model....

Text embeddings from large language models (LLMs) have achieved excellent results in tasks such as information retrieval, semantic textual similarity, etc. In this work, we show an interesting finding: when feeding a text into the embedding LLMs, obtained will be able to aligned with key tokens input text. We first fully analyze phenomenon on eight LLMs and that is universal not affected by model architecture, training strategy, method. With deeper analysis, then find main change space...

Recent Text-to-SQL methods leverage large language models (LLMs) by incorporating feedback from the database management system. While these effectively address execution errors in SQL queries, they struggle with mismatches -- that do not trigger exceptions. Database include issues such as condition and stricter constraint mismatches, both of which are more prevalent real-world scenarios. To challenges, we propose a tool-assisted agent framework for inspection refinement, equipping LLM-based...