Eye-controlled interaction in VR offers intuitive, hands-free control, but reducing false triggers such as the “Midas touch” remains challenging. This study evaluates target-assisted fixation (TA-Fixation) under varying fields of view (FOVs) and target densities. Experiment 1 compares TA-Fixation with standard fixation and gaze gesture techniques, showing that TA-Fixation balances accuracy and efficiency. However, in smaller FOVs and higher target densities, both fixation and TA-Fixation exhibit reduced accuracy, highlighting the relative advantage than gaze gestures. Experiment 2 explores how to optimize TA-Fixation in high-density scenarios, focusing on the effects of the relative position and size of the assisted-trigger object (AO) on performance based on Fitts's Law. The results indicate that smaller sizes and greater distances improve accuracy but decrease efficiency, while smaller distances and medium sizes offer a better trade-off. This study provides both theoretical and practical support and optimization strategies for gaze input techniques.

Neural Networks (NNs) have achieved breakthroughs in computer vision and natural language processing. However, modern models are computationally expensive, with floating-point operations posing a major bottleneck. Floatingpoint approximation, such as Mitchell’s logarithm, enables floating-point multiplication using simpler integer additions, thereby improving hardware efficiency. However, its practical adoption is hindered by challenges such as precision degradation, efficient hardware integrations, and management of trade-offs between accuracy and resource efficiency. In this paper, we propose a hardware-efficient down-samplingbased compensation method to mitigate precision loss and a flexible bias mechanism to accommodate diverse data distributions in NN models. Building on this foundation, we design configurable systolic arrays optimized for NN accelerators. To further support practical adoption, we introduce April, a co-design framework that balances the accuracy and resource usage of generated synthesizable systolic arrays. Our FPGA-based evaluations demonstrate that April-generated systolic arrays reduce root mean square error (RMSE) by up to $\mathbf{9 6 \%}$ and achieve $\mathbf{3 4 \% -5 2 \%}$ area reduction even compared to INT8-based implementations while maintaining comparable or improved model accuracy. Our design is open-sourced at https://github.com/CLabGit/April

Large Language Models (LLMs) have become foundational to modern natural language processing, yet their immense computational and memory demands pose major obstacles for efficient inference.Transformer-based LLMs rely heavily on floating-point general matrix-matrix multiplication (FP-GEMM), which dominates both compute and bandwidth.In this paper, we introduce Ax-Core, a quantization-aware, approximate GEMM unit that combines weight-only quantization with floating-point multiplication approximation (FPMA) to deliver highly efficient and accurate LLM inference.Unlike traditional GEMM units, AxCore eliminates multipliers entirely, replacing them with low-bit integer additions in a novel systolic array.AxCore features several key innovations: (1) a mixedprecision FPMA-based processing element that supports direct computation on compressed weights and high-precision activations;(2) a lightweight accuracy preservation strategy, including subnormal number handling, error compensation, and format-aware quantization; and (3) a set of systolic array optimizations, including shared correction and normalization logic.Evaluations on opensource LLMs show that AxCore achieves up to 6.3×-12.5×higher compute density than conventional FP GEMM units.Compared to state-of-the-art INT4-based accelerators, FIGLUT and FIGNA, AxCore improves compute density by 53% and 70%, respectively, while also delivering lower perplexity.AxCore is opensourced at: https://github.com/CLab

Gaze interaction in virtual reality (VR) offers promising advantages in speed and hands-free operation. However, selecting a suitable selection mechanism in VR interfaces with different visual encodings remains a challenge. This research compares two visual input techniques and explores specific visual elements, including field of view (FOV), design style, and number of targets, with the goal of alleviating these problems. The pilot study identifies two search stages, and it also demonstrates a significant effect of FOV on search time. In the formal experiment, a gaze-based trigger experiment is employed to examine the selection performance of fixation and smooth pursuit. The results reveal that optimizing FOV and number of targets improves the triggering efficiency. Furthermore, fixation outperforms pursuit in terms of triggering time and accuracy under the same conditions, while cognitive load during triggering process remains relatively similar. These findings are expected to propose design recommendations to enhance fixation and pursuit within VR interfaces.

The graph-based multiview clustering has gained significant attention due to its effectiveness in representing complex relationships among multiview data for enhanced clustering. Among the previous graph-based methods, the multiview graph learning (or graph fusion) technique has rapidly emerged as a promising direction, which, however, still suffers from two critical limitations. First, most of previous methods adopt a single-level of graph fusion, which lack the ability to go from single-level graph fusion to multilevel (deep) graph fusion. Second, they generally focus on constructing an optimal unified graph but cannot fully investigate the correlations among multiple views. Therefore, it is difficult to establish a comprehensive and obvious graph structure. In light of this, this article presents a new multiview graph learning method called deep similarity graph fusion (DSGF) for the multiview clustering task, where three pathways are simultaneously leveraged to fuse multilevel similarity into a unified graph. Particularly, multilevel graph fusion is utilized to obtain a view-specific similarity graph for each view and then fuse these single-view graphs (via three levels of graph fusion) into a robust graph, which takes advantage of deeper consensus information between various similarity graphs and improves the quality of the learned graph for the final spectral clustering process. Extensive experiments are conducted on six real-world multiview datasets, which demonstrate the highly competitive clustering performance of DSGF in comparison with state-of-the-art methods.

Curation in museums involves not only presenting exhibits for visitors but also deeply shaping a systematic narrative experience through deliberate spatial layout design of the museum space. In contrast, the dynamic nature of virtual reality (VR) environments establishes virtual museums as a more potent space for both layout optimization and narrative construction, particularly when integrating visitors’ diverse preferences to optimize the virtual museum and convey narratives. Therefore, we first collaborated with experienced curators to conduct a formative study to understand the workflow of curation and summarize the museum narratives that weave exhibits, galleries, and museum architecture into a compelling story. We then proposed a museum spatial layout framework that clarified three narrative levels (exhibit level, gallery level, and architecture level) to support the controllable spatial layout of the museum’s elements. Based on that, we developed VirtuNarrator, a proof-of-concept prototype designed to assist visitors in choosing different narrative themes, filtering exhibits, creating and adjusting galleries, and freely connecting them. The evaluation results validated that visitors received a more systematic museum narrative experience and perceptions of multi-perspective narrative design in VirtuNarrator. We also provided insights into VR-based museum narrative enhancement beyond spatial layout design.

High-Order Graph (HOG) clustering has received much attention for its advantage of exploiting the rich intrinsic structure of data. However, the construction of HOG involves the generation of a large number of redundant walks, which dilutes the useful walks and thus leads to untrustworthy high-order similarity and, consequently, suboptimal clustering results may be obtained. We formalize this issue as the Weight Explosion (WE) problem. Furthermore, current works rarely focus on exploiting the correlation between multi-order graphs that can capture high-order relations at various levels. In this paper, we first analyze the pattern of redundant walks, also termed as noise, and subsequently propose a novel$h$-length Simple Path Search ($h$-SPS) algorithm to solve the WE problem.$h$-SPS aims to find valid walks to denoise HOG and thus avoids enumerating walks to report the similarity. Regarding the second problem, we propose a multi-order graphs fusion method, which adaptively integrates graphs of varying orders by solving a convex problem. This allows us to capture information across different order levels effectively. Extensive experiments on benchmark datasets demonstrate that our method11https://github.com/YonghaoChen511/DenoHOG can effectively solve the proposed WE problem, while also well exploiting the correlation of multi-order graphs.

Conventional hash-based retrieval method rely on the assumption that the query and database are of the identical domain. However, cross-domain problem often occurs in real-world applications, leading to the unsatisfactory performance of existing hashing methods. Recently, some researchers have put forward domain adaptation retrieval (DAR) under the perspective of domain adaptation (DA) and achieved promising results. But the following limitations still exist: 1) a single function is used to handle two challenges, i.e., domain adaptation and hashing, which is not flexible to explore enough underlying information for simultaneously accomplishing these challenges well; 2) non-dominant features in the sample are ignored; 3) the dissimilarity structure of dissimilar samples is not taken into account. To address the above problems, we propose a novel framework named two-step strategy (TSS) for domain adaptation retrieval, which advocates dividing DAR into two steps: DA step and hashing step. A DA function and a hash function are learned to handle the above two challenges, respectively, making the process more reasonable. Additionally, a discriminant semantic fusion loss is proposed to improve the discriminative ability among classes. Unlike other works that focus on discovering dominant features, we exploit the neglected non-dominant features and assign them attention with sinusoidal semantic embedding, actively creating a clear separation between classes. At last, we present an adaptive similarity preserving loss to preserve the similarity structure of the original data in all intra-domain and inter-domain hash codes. Extensive experiments on various datasets demonstrate that the proposed TSS achieves state-of-the-art performance.

Hashing-based techniques present appealing solutions for cross-modal retrieval due to its low storage requirements and excellent query efficiency. The majority of cross-modal hashing methods typically adopt equal-length encoding scheme to represent multimodal data and achieve cross-modal similarity search. However, such scheme can be regarded as a relatively strict limitation, because it sacrifices the flexible representation of multimodal data in reality and cannot always guarantee the optimal retrieval performance. To address the challenge, this paper focuses on encoding heterogeneous data with varying hash lengths. To achieve this purpose, we propose a flexible cross-modal hashing approach, named Two-stage Asymmetric Similarity Preserving Hashing, TASPH for short, which can be applied to both unequal-length and equal-length retrieval scenarios. Specifically, in the first stage, TASPH designs a novel discrete asymmetric strategy to learn the modality-specific hash codes with varying lengths, enabling a flexible representation of heterogeneous data. Simultaneously, TASPH utilizes two semantic transformation matrices to establish the semantic correlations between varying hash codes. Different from most of the existing approaches that employ relaxation solutions, TASPH satisfies the discrete constraints without any relaxation. In the second stage, the learned semantic transformation matrices are employed to alleviate cross-modal heterogeneity, which guarantees that TASPH can learn more powerful hash functions to improve the discriminative ability of hash codes. Abundant experiments conducted on three benchmark datasets demonstrate encouraging results compared with the state-of-the-art approaches under different retrieval scenarios.

Partial multi-label learning (PML) needs to address the problem of multi-label learning when the dataset contains redundant information. PML is more challenging compared to traditional multi-label learning, because PML needs not only to perform the multi classification task, but also to reduce the impact of noise information on the model. Existing PML methods suffer from the following problems. (1) Only single source of noise is considered. (2) Some methods ignore the label correlations. To solve the above problems, we proposes a new dual noise elimination and dynamic label correlation guided partial multi-label learning (PML-DNDC). Specifically, the hidden ground-truth label matrix is decomposed into two compressed matrices of instance and classifier, which are used to approximate the candidate label matrix, to eliminate the negative effects of label noise on the model. On one hand, the compressed instance matrix maintains local structural consistency with the original instances, eliminating noise in the feature. On the other hand, dynamic label correlation guidance is designed to help classifier training by dynamically exploring the potential label correlations, which encourages relevant labels to obtain similar classifiers. After extensive experiments and analyses, we conclude that the proposed PML-DNDC is superior to the state-of-the-art methods.

In the field of digital humanities, color research aims to discover explanations for painting history and color usage habits. However, researchers analyzing color relationships is challenging and time-consuming, as it requires color extraction and a detailed review of many painting images for reference and comparison of color relationships. In our work, we propose ColorNetVis, an interactive color network analysis tool that enables researchers to explore color relationships through color networks. The core of ColorNetVis is a bipartite network model that establishes a bipartite relationship between colors and Chinese painting within a scope based on color difference measurement. It constructs a one-mode color network through projection algorithms and similarity calculation methods to discover the relationship between colors. We propose a coordinated set of views to demonstrate the combination of determined color networks with painting types and real-world attributes. We use color space view, color attribute distribution view, and single color query components to assist researchers in conducting detailed color analysis and validation. Through case studies, researcher reviews, and user studies, we demonstrate that ColorNetVis can effectively help researchers discover knowledge of color relationships and potential color research directions.

Based on a log-structured merge (LSM) tree, the key-value (KV) storage system can provide high reading performance and optimize random writing performance. It is widely used in modern data storage systems like e-commerce, online analytics, and real-time communication. An LSM tree stores new KV data in the memory and flushes to disk in batches. To prevent data loss in memory if there is an unexpected crash, RocksDB appends updating data in the write-ahead log (WAL) before updating the memory. However, synchronous WAL significantly reduces writing performance. In this paper, we present a new WAL mechanism named MyWAL. It directly manages raw devices (or partitions) instead of saving data on a traditional file system. These can avoid useless metadata updating and write data sequentially on disks. Experimental results show that MyWAL can significantly improve the data writing performance of RocksDB compared to the traditional WAL for small KV data on solid-state disks (SSDs), as much as five to eight times faster. On non-volatile memory express soild-state drives (NVMe SSDs) and non-volatile memory (NVM), MyWAL can improve data writing performance by 10%–30%. Furthermore, the results of YCSB (Yahoo! Cloud Serving Benchmark) show that the latency decreased by 50% compared with SpanDB.

Formal approaches to software reuse rely heavily upon a specification matching criterion, where a search query using formal specifications is used to search a library of components indexed by specifications. In previous investigations, we addressed the use of formal methods and component libraries to support software reuse and construction of software based on component specifications. A difficulty for all formal approaches to software reuse is the creation of the formal indices. We have developed an approach to reverse engineering that is based on the use of formal methods to derive formal specifications of existing programs. In this paper, we present an approach for combining software reverse engineering and software reuse to support populating specification libraries for the purposes of software reuse. In addition, we discuss the results of our initial investigations into the use of tools to support an entire process of populating and using a specification library to construct a software application.

Using existing components to construct software systems has significant potential to improving software productivity and quality. A key problem in software component reuse is the selection of appropriate components for satisfying a given requirement. In this paper we define a component interface generality relation that provides a foundation for component selection. This generality relation, represented in terms of formal specifications, precisely captures the semantic obligations for an existing component to satisfy the requirements of a target system. The formal specifications facilitate the (semi-) automatic determination of the generality relation. We show how this generality relation has been used to determine the reusability of software components in a software architecture-based reuse and integration environment.

Over the past several years, a number of techniques have been developed to address various issues involving software reuse, such as component classification, retrieval, and integration. However, it is not adequate to only have reuse techniques that address reuse issues separately. Instead, a seamless integration of these reuse techniques is critical to achieve effective reuse. In this paper, we present an integrated approach to software reuse. Based on software architecting techniques and formal methods, this approach addresses various reuse issues in a systematic and (semi) automatic fashion. An architecture-based software reuse and integration environment that supports this approach is also described.