Embedding-as-a-Service (EaaS) is an effective and convenient deployment solution for addressing various NLP tasks. Nevertheless, recent research has shown that EaaS is vulnerable to model extraction attacks, which could lead to significant economic losses for model providers. For copyright protection, existing methods inject watermark embeddings into text embeddings and use them to detect copyright infringement. However, current watermarking methods often resist only a subset of attacks and fail to provide comprehensive protection. To this end, we present the region-triggered semantic watermarking framework called RegionMarker, which defines trigger regions within a low-dimensional space and injects watermarks into text embeddings associated with these regions. By utilizing a secret dimensionality reduction matrix to project onto this subspace and randomly selecting trigger regions, RegionMarker makes it difficult for watermark removal attacks to evade detection. Furthermore, by embedding watermarks across the entire trigger region and using the text embedding as the watermark, RegionMarker is resilient to both paraphrasing and dimension-perturbation attacks. Extensive experiments on various datasets show that RegionMarker is effective in resisting different attack methods, thereby protecting the copyright of EaaS.

Shufan Yang, Zifeng Cheng, Zhiwei Jiang, Qingfeng Qi, Yafeng Yin, Cong Wang, Ao Zhou, Qing Gu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recent advancements in near-infrared spectroscopy (NIRS) and associated optical techniques have significantly contributed to the development of wearable neuroimaging devices capable of capturing real-time neuronal activity with enhanced spatial and temporal resolution. However, despite these advancements, calibration remains a persistent challenge due to variability in NIRS sensor design, which can significantly degrade data quality. In this paper we introduce a universal data-driven calibration method designed to enhance the precision and reliability of NIRS sensor measurement. The proposed method integrates gradient descent–based optimisation with constraint-guided clustering to iteratively minimise calibration errors under realistic usage conditions. To evaluate its effectiveness, the algorithm was tested using an in-silico phantom constructed in MATLAB/NIRFAST, demonstrating notable improvements in signal clarity and haemoglobin concentration estimation. Additionally, the approach exhibits robustness to motion artefacts, thereby improving measurement fidelity. These contributions advance the reliability and accessibility of wearable brain imaging systems, enabling broader applications in both neuroscience research and clinical diagnostics.

Functional Near-Infrared Spectroscopy (fNIRS) is a neuroimaging method which can be implemented with a wearable form factor. However, the data of fNIRS can be affected by motion artifact, which is conventionally processed offline using MATLAB-based software package via a bulky PC. This study trains a Support Vector Machine (SVM) algorithm and proposes a hardware design approach based on an FPGA to achieve the first real-time fNIRS motion artifact detection. The SVM hardware architecture proposed here utilizes a partially sequential–partially parallel implementation of the classification algorithm where Support Vector channels are consolidated into a single oversampled channel. A high classification accuracy of 97.42%, low FPGA resource utilization of 38,354 look-up tables and 6024 flip-flops with 10.92 us latency is achieved, outperforming conventional CPU SVM methods. These results show that an FPGA-based fNIRS motion artifact detector can be exploited whilst meeting real-time and resource constraints that are crucial in high-performance reconfigurable hardware systems.

Due to iterative matrix multiplications or gradient computations, machine learning modules often require a large amount of processing power and memory. As a result, they are often not feasible for use in wearable devices, which have limited processing power and memory. In this study, we propose an ultralow-power and real-time machine learning-based motion artifact detection module for functional near-infrared spectroscopy (fNIRS) systems. We achieved a high classification accuracy of 97.42%, low field-programmable gate array (FPGA) resource utilization of 38354 lookup tables and 6024 flip-flops, as well as low power consumption of 0.021 W in dynamic power. These results outperform conventional CPU support vector machine (SVM) methods and other state-of-the-art SVM implementations. This study has demonstrated that an FPGA-based fNIRS motion artifact classifier can be exploited while meeting low power and resource constraints, which are crucial in embedded hardware systems while keeping high classification accuracy.

Multi-Level Intermediate Representation (MLIR) is gaining increasing attention in reconfigurable hardware communities due to its capability to represent various abstract levels for software compilers. This project aims to be the first to provide an end-to-end framework that leverages open-source, cross-platform compilation technology to generate MLIR from SYCL. Additionally, it aims to explore a lowering pipeline that converts MLIR to RTL using open-source hardware intermediate representation (IR) and compilers. Furthermore, it aims to couple the generated hardware module with the host CPU using vendor-specific crossbars. Our preliminary results demonstrated the feasibility of lowering customized MLIR to RTL, thus paving the way for an end-to-end compilation.

This demonstration showcases a novel FPGA development pipeline for developing a low-power and real-time motion artefact detection module for a wearable functional near-infrared spectroscopy (fNIRS) processing system. We provide a brief overview of the development design flow for our learning-based motion artefact detector in a heterogeneous platform, as well as the evaluation method for removing motion artefacts, which are unwanted signal variations that occur due to subject motion during data acquisition.

Radar is an extremely valuable sensing technology for detecting moving targets and measuring their range, velocity, and angular positions. When people are monitored at home, radar is more likely to be accepted by end-users, as they already use WiFi, is perceived as privacy-preserving compared to cameras, and does not require user compliance as wearable sensors do. Furthermore, it is not affected by lighting conditions nor requires artificial lights that could cause discomfort in the home environment. So, radar-based human activities classification in the context of assisted living can empower an aging society to live at home independently longer. However, challenges remain as to the formulation of the most effective algorithms for radar-based human activities classification and their validation. To promote the exploration and cross-evaluation of different algorithms, our dataset released in 2019 was used to benchmark various classification approaches. The challenge was open from February 2020 to December 2020. A total of 23 organizations worldwide, forming 12 teams from academia and industry, participated in the inaugural Radar Challenge, and submitted 188 valid entries to the challenge. This paper presents an overview and evaluation of the approaches used for all primary contributions in this inaugural challenge. The proposed algorithms are summarized, and the main parameters affecting their performances are analyzed.

Abstract Modern software systems are increasingly expected to show higher degrees of autonomy and self-management to cope with uncertain and diverse situations. As a consequence, autonomous systems can exhibit unexpected and surprising behaviours. This is exacerbated due to the ubiquity and complexity of Artificial Intelligence (AI)-based systems. This is the case of Reinforcement Learning (RL), where autonomous agents learn through trial-and-error how to find good solutions to a problem. Thus, the underlying decision-making criteria may become opaque to users that interact with the system and who may require explanations about the system’s reasoning. Available work for eXplainable Reinforcement Learning (XRL) offers different trade-offs: e.g. for runtime explanations, the approaches are model-specific or can only analyse results after-the-fact. Different from these approaches, this paper aims to provide an online model-agnostic approach for XRL towards trustworthy and understandable AI. We present ETeMoX, an architecture based on temporal models to keep track of the decision-making processes of RL systems. In cases where the resources are limited (e.g. storage capacity or time to response), the architecture also integrates complex event processing, an event-driven approach, for detecting matches to event patterns that need to be stored, instead of keeping the entire history. The approach is applied to a mobile communications case study that uses RL for its decision-making. In order to test the generalisability of our approach, three variants of the underlying RL algorithms are used: Q-Learning, SARSA and DQN. The encouraging results show that using the proposed configurable architecture, RL developers are able to obtain explanations about the evolution of a metric, relationships between metrics, and were able to track situations of interest happening over time windows.

Inflammation of the gastrointestinal (GI) tract accompanies several diseases, including Crohn's disease. Currently, video capsule endoscopy and deep bowel enteroscopy are the main means for direct visualisation of the bowel surface. However, the use of optical imaging limits visualisation to the luminal surface only, which makes early-stage diagnosis difficult. In this study, we propose a learning enabled microultrasound ( μ US) system that aims to classify inflamed and non-inflamed bowel tissues. μ US images of the caecum, small bowel and colon were obtained from mice treated with agents to induce inflammation. Those images were then used to train three deep learning networks and to provide a ground truth of inflammation status. The classification accuracy was evaluated using 10-fold evaluation and additional B-scan images. Our deep learning approach allowed robust differentiation between healthy tissue and tissue with early signs of inflammation that is not detectable by current endoscopic methods or by human inspection of the μ US images. The methods may be a foundation for future early GI disease diagnosis and enhanced management with computer-aided imaging.

This paper discusses the place of radar for assisted living in the context of IoT for Health and beyond. First, the context of assisted living and the urgency to address the problem is described. The second part gives a literature review of existing sensing modalities for assisted living and explains why radar is an upcoming preferred modality to address this issue. The third section presents developments in machine learning that helps improve performances in classification especially with deep learning with a reflection on lessons learned from it. The fourth section introduces recent published work from our research group in the area that shows promise with multimodal sensor fusion for classification and long short-term memory applied to early stages in the radar signal processing chain. Finally, we conclude with open challenges still to be addressed in the area and open to future research directions in animal welfare.

Face detection and recognition benchmarks have shifted toward more difficult environments. The challenge presented in this paper addresses the next step in the direction of automatic detection and identification of people from outdoor surveillance cameras. While face detection has shown remarkable success in images collected from the web, surveillance cameras include more diverse occlusions, poses, weather conditions and image blur. Although face verification or closed-set face identification have surpassed human capabilities on some datasets, open-set identification is much more complex as it needs to reject both unknown identities and false accepts from the face detector. We show that unconstrained face detection can approach high detection rates albeit with moderate false accept rates. By contrast, open-set face recognition is currently weak and requires much more attention.

SpiNNaker is a biologically-inspired massively-parallel computer designed to model up to a billion spiking neurons in real-time. A full-fledged implementation of a SpiNNaker system will comprise more than 105 integrated circuits (half of which are SDRAMs and half multi-core systems-on-chip). Given this scale, it is unavoidable that some components fail and, in consequence, fault-tolerance is a foundation of the system design. Although the target application can tolerate a certain, low level of failures, important efforts have been devoted to incorporate different techniques for fault tolerance. This paper is devoted to discussing how hardware and software mechanisms collaborate to make SpiNNaker operate properly even in the very likely scenario of component failures and how it can tolerate system-degradation levels well above those expected.

A Token-ManagedAdmission Control (TMAC) mechanism is introduced in order to provide efficient Quality-of-Service (QoS) support for different types of application on a best-effort Globally-Asynchronous Locally-Synchronous (GALS) interconnect fabric. The mechanism is applied at the ingress edges of the fabric using tokens to allocate dynamic network resources and prevent network congestion. The degree of fairness is controllable, in order to balance the desired throughput and data transfer resource allocation appropriately for a particular application. The simulation and analysis presented here shows efficient QoS provision. Our detailed implementation and analysis show that TMAC provides service guarantees on the network while using a modest physical area because of the simplicity of the control logic.

Given the limited current understanding of the neural model of computation, hardware neural network architectures that impose a specific relationship between physical connectivity and model topology are likely to be overly restrictive. Here we introduce, in the SpiNNaker chip, an alternative approach: a mappable virtual topology using an asynchronous network-on-chip (NoC) that decouples the “logical” connectivity map from the physical wiring. Borrowing the established digital RAM model for synapses, we develop a concurrent memory access channel optimised for neural processing that allows each processing node to perform its own synaptic updates as if the synapses were local to the node. The highly concurrent nature of interconnect access, however, requires careful design of intermediate buffering and arbitration. We show here how a locally buffered, one-transaction-per-node model with multiple synapse updates per transaction enables the local node to offload continuous burst traffic from the NoC, allowing for a hardware-efficient design that supports biologically realistic speeds. The design not only presents a flexible model for neural connectivity but also suggests an ideal form for general-purpose high-performance on-chip interconnect.