Recent advances in large language models (LLMs) and text-aware graph learning have increased interest in reasoning over textattributed graphs (TAGs).In many real-world settings, such graphs are inherently heterogeneous, with most existing benchmarks remaining largely homogeneous in structure.As a result, the lack of large-scale benchmarks for heterogeneous text-attributed graphs has hindered systematic evaluation and fair comparison of existing methods.In this work, we introduce CITE -Catalytic Information Textual Entities Graph, the first and largest heterogeneous text-attributed citation graph benchmark for catalytic materials.CITE contains over 438K nodes and 1.2M edges spanning four node types and four relation types, with rich node-level textual information.We establish standardized evaluation protocols for node classification and link prediction, and conduct ablation studies to assess the impact of graph heterogeneity and textual attributes.Using CITE, we benchmark four classes of learning paradigms, including homogeneous graph models, heterogeneous graph models, LLM-centric models, and LLM+Graph models.By providing a largescale heterogeneous text-attributed benchmark together with standardized evaluation protocols and comprehensive baselines, CITE enables systematic assessment across diverse modeling paradigms and offers new insights into textaware and LLM-enhanced graph learning.The dataset 1 , codebase and evaluation suite 2 are publicly available.

Terahertz (THz) communication is a key enabling technology for achieving high-capacity, long-distance inter-satellite and satellite-ground communications in the next-generation wireless systems. Photonic-assisted upconversion provides a cost-effective approach to realizing ultra-wideband THz communication systems. To explore the potential of quantum neural network for the photonic-assisted THz communication systems, this work is the first to design the hybrid quantum-classical neural network for quadrature amplitude modulation (QAM) symbol recovery over tens of Gbit/s THz kilometer-level wireless transmission link, which can significantly enhance the receiver performance and reduce the computational complexity. For the proof of concept, we create the 10 Gbaud sub-THz prototype to validate the proposed scheme over 4.6km wireless long-distance. The results demonstrate that the proposed scheme achieves a 0.7dB improvement in receiver sensitivity and a one-hundred-times reduction in real-valued multiplications per symbol (RMps) compared to the classical ones.

Retrieval Augmented Generation can effectively reduce hallucinations in LLMs during question-answering, with embedding models directly influencing its performance. While current embedding models improve encoding through large-scale training, they often overlook the potential or explicit multi-aspect information within the text. To address this, we propose the Multi-Aspect Extraction Model (MAEM), an eigen decomposition-based approach that extracts and integrates text aspects into a unified vector for enhanced retrieval, and introduce a regularization loss function to assist in training. We utilized LLMs to create the Policy-Corpus dataset and validated the model on both Policy-Corpus and FiQA. Incorporating MAEM and regularization into GTEbaseimproved NDCG@10 by 3.8 points on FiQA and 4.56 points on Policy-Corpus. Respectively, achieving results comparable to larger models using a smaller parameter model.

Research project conflict risk detection aims to distinguish potential conflicts from multiple aspects including personnel allocation, resource distribution, and content overlap, etc. Existing research on project conflict detection faces three critical challenges: the difficulty in capturing semantic information from unstructured project documentation, limited model expressiveness in processing heterogeneous project relationships, and challenges in modeling diverse conflict patterns across interconnected project elements. To address these challenges, we first construct a research project heterogeneous graph with three types of nodes (i.e., projects, personnel, and resources) and five types of relationships capturing various project interactions. Besides, we propose a novel Semantic-augmented Heterogeneous grAph neural network for project conflict risk DEtection (SHADE) framework, equipped with three specially designed modules: 1) a semantic-augmented module leveraging large language models to extract fine-grained content representations from project descriptions, 2) a multi-view risk detection module with adaptive fusion to capture conflicts from diverse perspectives, and 3) a self-supervised contrastive learning module to enhance the discriminative power between positive and negative patterns. Extensive experiments on the real-world research project heterogeneous graph demonstrate that our proposed framework SHADE significantly outperforms state-of-the-art methods.

Sensitive content detection in financial reimbursement documents is fundamental to corporate compliance management. Traditional keyword-based approaches suffer from high false-positive rates due to their inability to capture contextual semantics. In this paper, we propose TripletAudit, a context-aware framework for sensitive content detection that leverages triplet learning, incorporating a hierarchical architecture combining BERT for semantic embedding, Bi-Lstm for sequential modeling, and multi-head attention for context understanding. Our triplet learning mechanism effectively distinguishes between sensitive and non-sensitive contexts through anchor-positive-negative samples. Experiments on real-world reimbursement datasets demonstrate that TripletAudit achieves state-of-the-art performance with 94.03% accuracy and 93.24% F1-score, significantly outperforming baseline methods in financial compliance scenarios.

In recent years, single-label intent recognition has faced significant challenges in handling short and semantically sparse user inputs. Traditional methods often treat intent labels as flat categories, neglecting their inherent hierarchical relationships and limiting model performance. To address these issues, we propose HiIntent , a novel zero-shot intent detection framework that integrates hierarchical semantic modeling with a collaborative generation-discriminative mechanism. HiIntent first constructs a structured label hierarchy through a two-stage process: a large language model (LLM) generates semantic abstractions of intent labels, which are then evaluated and refined by a discriminative module to ensure coherence and correctness. This is followed by a similarity-driven convergence strategy that enhances intra-class consistency and inter-class separability using multi-metric similarity calculations. Finally, a contrastive prompt construction method leverages the learned label hierarchy to generate enriched semantic descriptions for each intent, improving representation learning and facilitating accurate classification even in zero-shot scenarios. Extensive experiments on both general-purpose (CLINC-150) and domain-specific (RFMR) datasets demonstrate that HiIntent consistently outperforms existing approaches across multiple evaluation metrics. Ablation studies and hyperparameter analyses further validate the effectiveness of each component in the proposed framework.

Accurately predicting the future trajectories of traffic agents in real-world scenarios is critical for advancing intelligent cyber–physical systems (CPS), such as autonomous driving systems and smart cities. A fundamental challenge lies in mining the evolving interaction patterns among multiple agents from their past trajectories, as traffic scenarios often exhibit complex interactions and continuously evolve along the timeline. However, existing methods fail to fully exploit the temporality inherent in sequential interactions. In the process of modeling interactions, they lack a comprehensive understanding of static interactions that occur at constant timestamps and the evolving patterns of interactions across timestamps. To tackle these challenges, we propose E volutio n al T emporal- A ware I nteraction L earning ( EnTAIL ), a novel temporal-aware interaction learning framework to model and reason the interactions among agents. EnTAIL captures both static interaction patterns at individual timestamps and temporal-aware interaction patterns across timestamps through a unified framework. Specifically, we introduce a trainable constant time encoding to integrate with the interaction modeling in each individual timestamp, which aims to capture the static interaction information. We propose a dynamic evolution encoder to model temporal-aware interaction features, enabling learning both short-term and long-term interactions within multiscaled observation windows. Besides, EnTAIL also considers the temporal feature in the prediction stage and models the long-range interactions ignored during the encoding phase. Extensive experiments conducted on the challenging real-world Argoverse dataset demonstrate that our proposed model achieves substantial performance improvement and outperforms the baseline methods up to 2.5% in minimum Average Displacement Error (minADE) and 1.2% in minimum Final Displacement Error (minFDE).

Accurate and continuous prediction of vehicle trajectories is crucial for the secure deployment of intelligent transportation systems (ITS). Many previous approaches ignore the unique driving behaviors of individuals and have difficulty learning sequential dependencies effectively. However, the driver’s inherent driving pattern will directly impact the subsequent decision. Therefore, we propose a novel framework called Latent Individual Driving Pattern (LIDP), which models historical trajectories emphasizing individual driving patterns. Specifically, we introduce a novel contrastive loss to learn the individual driving pattern. Subsequently, we utilize a representation-augmented module to enhance the expressive capacity of each driving representation. We evaluate the proposed method on two real-world freeway trajectory datasets: US-101 and I-80 in the NGSIM highway dataset, and it achieves promising results, outperforming the recent vehicle trajectory prediction methods.

An efficient and accurate understanding of the intent of code is an indispensable skill in computer technology, especially in collaborative engineering and experimental reproduction. AI-assisted automated code comment generator, with the goal of generating programmer-readable explanations, has been an emerging hot topic for software project comprehension. Despite promising performances, three critical issues emerged: 1) The summary comment is limited in understanding the fine-grain details of the code. 2) key-word level guidance in the model should be included for better comments generation. 3) performance of the generative model may be dampened by noises in the manual annotation. In response, we propose a novel fine-grain comment generation, a scenario of generating the statement-level comment with the assistance of method-level comment. We also propose KFCC, a differentiation-aware and keyword-guided fine-grain comment generation model. Specifically, the proposed KFCC model generates the statement-level comments by incorporating the key information extracted by the keyword extractor in a gate fusion way. To enhance the effectiveness and robustness of the proposed KFCC model, we propose a differentiation-aware enhancing encoder comprehension, letting the model distinguish significant knowledge via contrastive learning. Extensive experiments conducted on open-source projects demonstrate that the KFCC model achieves outstanding performance in six programming languages (including Ruby, Python, JavaScript, Java, etc.) on the CodeSearchNet benchmark.

Graph anomaly detection has attracted considerable interest due to the wide use of graph structure data. Several GNN-based anomaly detection methods discover anomalies through the powerful node representation ability of GNNs. However, real-world graphs are typically rarely labeled, which leads these deep learning methods to face the challenge of under-fitting. A fundamental question here is: can anomalies in graphs be detected with few annotations? In this paper, we propose a novel semi-supervised anomaly detection method in graphs based on Consistency Exploitation (ConsE). First, ConsE adopts a consistency-based neighbor sampler, which ensures the consistency of a central node and its neighbors on attributes and categories during the aggregation process through attribute similarity and soft pseudo-labels. Afterward, ConsE encourages the consistency of node representations generated by a dedicated neighbor sampler and a generic neighbor sampler to improve its robustness in complex neighborhoods. Experimental results on real-world datasets demonstrate that our model significantly outperforms several state-of-the-art baseline methods.

Pedestrian trajectory prediction is a challenging task due to its inherent uncertainty and multi-modal nature of human intention. There are multiple possible trajectories based on the same historical locations. Our key insight is that humans are intention-driven, the future trajectories can be effectively captured by a set of intention proposals. This leads to our Hybrid Intention-proposal trajectory prediction (HyIntent) framework. HyIntent is a hybrid network architecture that acts solely on historically observed locations. We use LSTM as our recurrent backbone to memorize the sequential feature and a transformer decoder to capture the long-term dependency with intention. Given some orthogonal learned intention proposals, HyIntent reasons about the relations of intention proposals and the historical observations to tailor generate the multiple predictions in parallel. HyIntent demonstrates improved performance on public datasets (i.e. ETH/UCY, SDD) and outperforms state-of-the-art while minimizing the complexity.

The text summarization task aims to generate succinct sentences that summarise what an article tries to express. Based on pretrained language models, combining extractive and abstractive summarization approaches has been widely adopted in text summarization tasks. It has been proven to be effective in many existing pieces of research using extract‐then‐abstract algorithms. However, this method suffers from semantic information loss throughout the extraction process, resulting in incomprehensive sentences being generated during the abstract phase. Besides, current research on text summarization emphasizes only word‐level comprehension while paying little attention to understanding the level of the sentence. To tackle this problem, in this paper, we propose the SentMask component. Taking into account that the semantics of sentences that are filtered out during the extraction process is also worth considering, the paper designs a sentence‐aware mask attention mechanism in the process of generating a text summary. By applying the extractive approach, the paper first selects the most essential sentences to construct the initial summary phrases. This information leads the model to modify the weights of the attention mechanism, which provides supervision for the generative model to ensure that it focuses on the sentences that convey important semantics while not ignoring others. The final summary is constructed based on the key information provided. The experimental results demonstrate that our model achieves higher ROUGE and BLEU scores compared to other baseline models on two benchmark datasets.

In the era of big data, with the rapid increase in the number of texts, we have found ourselves submerged in the sea of texts. With the advent of the text summarization methodology, the time-consuming and energy-draining reading process of manual operations can be avoided by condensing and communicating the primary concept to the people. When generating sentences, the generator should comply with grammatical rules, such as lexical constraints. Leveraging lexical constraints in encoder-decoder models for the text summarization task has been extensively analyzed. Previous research add lexical constraints by enriching word embeddings in the encoding process. The model could be fed with more auxiliary knowledge. However, such lexical constraints cannot control the decoder, resulting in generating ungrammatical phrases. To address this issue, this paper proposes a grammar-aware text summarization method by incorporating Part-of-Speech (POS) constraints to guide the decoding process, along with auxiliary embeddings including POS, Lemmatization and Named Entity Recognition features in encoder. Detailed, the proposed POS-constrained AuxPOS model controls the generated word distribution during decoding according to the exact POS tag. Experimental results demonstrate that the proposed AuxPOS model can generate summaries that are more in concert with grammatical logic with high quality on the CNN/DailyMail, XSum and AESLC summarization datasets.

Carrier phase recovery (CPR) determines the accuracy of the receiver in modern coherent optical communication. The accurate estimation and tracking of carriers are particularly vital with the increase of throughput for long-distance transmit. It is a challenge to implement a real-time system because the computational complexity increases with fractional bits. Moreover, the conversion between polar coordinates and Cartesian coordinates introduces a high latency. In this paper, we present an FPGA implementation of low latency Viterbi-Viterbi 4thPower Estimation (VV4E) based CPR, which mainly performs the computation in Cartesian coordinates and implements the trigonometric function with a look-up table (LUT). Evaluations on Xilinx ZCU102 show that at a frequency of 370MHz, it introduces a 22-cycle latency to handle the 29.6 GBd QPSK signals, which is the minimum value to our knowledge.

We have experimentally demonstrated terahertz (THz) wireless transmission of 27.9 Gbit/s orthogonal frequency division multiplexing (OFDM) signal at 335 GHz over 10 km fiber and 400 m wireless link, employing advanced digital signal processing (DSP) algorithms. As we know, based on the photonics-aided scheme, THz-wireless OFDM signal transmission is achieved for the first time over record-breaking 400 m wireless distance.

We realized the bi-directional transmission of millimeter-wave (MMW) 4096-ary quadrature amplitude modulation (4096QAM) orthogonal frequency division multiplexing (OFDM) signal with frequencies of 83.5 GHz and 73.3 GHz at E-band over 2 m wireless distance with a net transmission rate of 91.46 Gbit/s. Meanwhile, we also achieved the transmission of the MMW 4096QAM OFDM signal at 117 GHz in D-band with a net transmission rate of 57.21 Gbit/s over 13.42 m wireless distance. With the help of probabilistic shaping (PS) technique and Volterra nonlinearity compensation (VNC) technology, both of the transmission systems can satisfy the 0.8 normalized generalized mutual information (NGMI) threshold with 25% soft-decision forward-error-correction (SD-FEC) overhead. In addition, the performances of different modulation formats are also be compared.

Applying for research funding projects is becoming one of the most important ways for scientists to carry on the research. How to find an appropriate collaborator/applicant is a major concern for scientists. Social networks provide one means of visualizing existing and potential collaborations. In this paper, we study the funding collaborators recommendation problems. We solve the problem by starting with analyzing the researchers' motivations for finding collaboration, which are (i) to form a competitive team (ii) to expand cooperation circle, which little work noticed. We model the funding relation as a complex network called co-applicant network. Based on that, we propose a utility function to take all the aspects of recommendation into account. And we propose a novel recommendation algorithm by modeling the utility function based on the group relations in the co-applicant network. We experiment our approaches on National Science Foundation of China (NSFC) funding projects and achieve effective results.

The key current challenges for the industrial application of all optical switching networks are energy consumption, transmission rate, spectrum efficiency, and switching throughput. The energy consumption problem is mainly researched in this paper. From the perspective of components and modules, node equipment, and network levels, different enabling technologies are proposed to overcome this problem, which are also evaluated through different experimental demonstrations. First, high-sampling-rate digital-to-analog converters (DACs) and WSS-based ROADM modules are demonstrated as components and modules for energy-efficient all optical switching networks. Then, an all optical transport network test-bed consisting of 10 Pbit/s level all optical switching nodes based on multi-level and multi-planar switching architecture is experimentally demonstrated for the first time, which can reduce power consumption by 43%. A control architecture for energy-efficient all optical switching networks is built with OpenFlow based software defined networking (SDN), and experimental results are given to verify the performance of this control architecture. Finally, we describe an All Optical Networks Innovation (AONI) project in China, which aims to explore transmission, switching, and networking technologies in all optical switching networks, and then two application scenarios are forecast based on the technical breakthroughs of this project.

In this paper, we carry out a knowledge-based scholar network practice towards Research community, named Research Social Networking, shortly DSN, by setting up a large knowledge base of scientists. We discuss key technologies in the paper, including scholar disambiguation and relationship extraction with the better performance evaluation than traditional methods. The DSN system has been implemented and integrated with Duckling cloud service, known as Research Online, with more than 60 thousand scientists and 100 thousand papers.

An energy efficient and transparent platform for optical wireless networks is proposed based on a novel reversely modulated optical single sideband scheme (RM-OSSB). This scheme is based on a parallel Mach-Zehnder modulator (P-MZM). The bandwidth limitation from electrical and optical components is overcome by down-conversion without electrical mixer, data format compatibility, and dispersion-free. It makes such scheme suitable for a universal platform. Moreover, the modulation power efficiency (MPE) of RM-OSSB can be improved. A theoretical model of RM-OSSB is built up for performance analysis. Both numerical simulations and experimental results match well with our theoretical model. Based on RM-OSSB scheme, a 58GHz full duplex RoF system is experimentally demonstrated. The 2.9Gb/s on-off keying (OOK) signals carried by 58GHz mm-wave is successfully delivered over 50km SMF with negligible power penalty.

Hot topics are usually those breaking news discussed most at online forums, especially microblogging systems, such as twitter, which helps to learn user concentration and public opinion. This paper focuses on the problem of predicting emerging hot topics. Previous prediction models usually focus on building the content profile to discover the hot topics, they may neglect the social network function or overlook the keyword feature of the post. In this paper, we address this problem by introducing a combined model using the content and the connection information. We define the concept of topic hotness, introduce the algorithm calculating the hotness with content based hotness and connection based hotness, and finally we predict those emerging hot topics by the hotness evolution model.

Top-k approximate querying on string collections is an important data analysis tool for many applications, and it has been exhaustively studied. However, the scale of the problem has increased dramatically because of the prevalence of the Web. In this paper, we aim to explore the efficient top-k similar string matching problem. Several efficient strategies are introduced, such as length aware and adaptive q-gram selection. We present a general q-gram based framework and propose two efficient algorithms based on the strategies introduced. Our techniques are experimentally evaluated on three real data sets and show a superior performance.

This paper presents an on-chip jitter measurement technique based on the Vernier ring time-do-digital converter (VRTDC). Vernier delay line is an attractive structure for the implementation of high performance TDC due to its sub-gate-delay resolution and cancellation of the first order process, voltage and temperature (PVT) variations. In order to improve the detectable range, area cost and power consumption of the conventional Vernier delay line TDC, the Vernier ring structure is developed to place two delay lines and comparator chains in ring format for the reuse of hardware, which enables the VR-TDC to achieve a fine resolution without sacrificing detectable range. The build-in coarse and fine interpolations reduce the power and area. This on-chip jitter measurement scheme can measure a large jitter with a fine resolution smaller than 8ps. An exemplary jitter test is given in this paper to demonstrate the capability of the proposed jitter measurement scheme.

BACKGROUND: Protein-DNA interaction constitutes a basic mechanism for the genetic regulation of target gene expression. Deciphering this mechanism has been a daunting task due to the difficulty in characterizing protein-bound DNA on a large scale. A powerful technique has recently emerged that couples chromatin immunoprecipitation (ChIP) with next-generation sequencing, (ChIP-Seq). This technique provides a direct survey of the cistrom of transcription factors and other chromatin-associated proteins. In order to realize the full potential of this technique, increasingly sophisticated statistical algorithms have been developed to analyze the massive amount of data generated by this method. RESULTS: Here we introduce HPeak, a Hidden Markov model (HMM)-based Peak-finding algorithm for analyzing ChIP-Seq data to identify protein-interacting genomic regions. In contrast to the majority of available ChIP-Seq analysis software packages, HPeak is a model-based approach allowing for rigorous statistical inference. This approach enables HPeak to accurately infer genomic regions enriched with sequence reads by assuming realistic probability distributions, in conjunction with a novel weighting scheme on the sequencing read coverage. CONCLUSIONS: Using biologically relevant data collections, we found that HPeak showed a higher prevalence of the expected transcription factor binding motifs in ChIP-enriched sequences relative to the control sequences when compared to other currently available ChIP-Seq analysis approaches. Additionally, in comparison to the ChIP-chip assay, ChIP-Seq provides higher resolution along with improved sensitivity and specificity of binding site detection. Additional file and the HPeak program are freely available at http://www.sph.umich.edu/csg/qin/HPeak.

This paper describes a kernel based Web Services (abbrevi-ated as service) matching mechanism for service discoveryand integration. The matching mechanism tries to exploitthe latent semantics by the structure of services. Using textual similarity and n-spectrum kernel values as features of low-level and mid-level, we build up a model to estimate thefunctional similarity between services, whose parameters arelearned by a Ranking-SVM. The experiment results showedthat several metrics for the retrieval of services have beenimproved by our approach.

Centralized UDDI currently becomes harder to catch up with the need of Internet-scale service publishing and discovery. Peer-to-peer technology with its advantages of scalability, fault-tolerance and dynamics is suitable for loose-coupled Web services' distribution in nature and prevailing nowadays. In this paper, we present a distributed, collaborative peer-to-peer service overlay suitable for service publishing and discovery based on skip graph which could be organized by planar model or tree model according to the scale of services. We propose a prefix-ordered clustering mechanism that congregates the keys within a fixed prefix into the same bucket and make the smallest one appeared as the key in the service overlay which would support prefix, range query and semantic matching on Web services. We then analyze the performance of our proposed with simulation that validates the effectiveness of the service overlay

The optical packet-switching network is considered to be one of the most promising solutions for end-to-end delivery of high-bitrate data, video, and voice signals across optical networks of the future. Optical label switching (OLS) technology incurs simpler extraction and processing of the labels so that the optical packets can be routed with low latency to the destinations. We have developed several key enabling technologies for integrated optical networks, including optical label generation, label swapping, optical buffering, clock recovery, and wavelength conversion. We have designed and experimentally demonstrated these enabling techniques that can provide efficient broadband services in future optical networks.