Search Results (1,646)

Open-weight generative large language models (LLMs) can be freely downloaded and modified. Yet, little empirical evidence exists on how these models are systematically altered and redistributed. This study provides a large-scale empirical analysis of safety-modified open-weight LLMs, drawing on 8608 model repositories and evaluating 20 representative modified models on unsafe prompts designed to elicit, for example, election disinformation, criminal instruction, and regulatory evasion. This study demonstrates that modified models exhibit substantially higher compliance: while an average of unmodified models complied with only 19.2% of unsafe requests, modified variants complied at an average rate of 80.0%. Modification effectiveness was independent of model size, with smaller, 14-billion-parameter variants sometimes matching or exceeding the compliance levels of 70B parameter versions. The ecosystem is highly concentrated yet structurally decentralized; for example, the top 5% of providers account for over 60% of downloads and the top 20 for nearly 86%. Moreover, more than half of the identified models use GGUF packaging, optimized for consumer hardware, and 4-bit quantization methods proliferate widely, though full-precision and lossless 16-bit models remain the most downloaded. These findings demonstrate how locally deployable, modified LLMs represent a paradigm shift for Internet safety governance, calling for new regulatory approaches suited to decentralized AI. Full article

Integrated sensing and communication (ISAC) is a transformative technology for sixth-generation (6G) wireless networks. In this paper, we investigate end-to-end constellation mapping and demapping in ISAC systems, leveraging OFDM-based waveforms and an adaptive DNN architecture for pulse-based transmission. Specifically, we propose an end-to-end autoencoder framework that optimizes the constellation through adaptive symbol distribution shaping via deep learning, enhancing communication reliability with symbol mapping and boosting sensing capabilities with an improved peak-to-sidelobe ratio (PSLR). The autoencoder consists of an autoencoder mapper (AE-Mapper) and an autoencoder demapper (AE-Demapper), jointly trained using a composite loss function to optimize constellation points and achieve flexible performance balance in communication and sensing. Simulation results demonstrate that the proposed DNN-based end-to-end design achieves dynamic balance between PSLR of the autocorrelation function (ACF) and bit error rate (BER). Full article

The implementation of K–12 engineering education presents significant challenges for current teachers. To strengthen the instructional readiness of elementary school teachers in Taiwan for engineering education, this study developed a STEM-oriented engineering design activity. It draws on international K–12 engineering education models and incorporates the computational thinking competencies prioritized by Taiwan’s national technology curriculum. The activity involves employing the micro:bit microcontroller to design an educational buzz wire game machine themed on energy education, integrating the engineering design process with knowledge from science, mathematics, and technology. The activity was implemented in an elementary teacher education course through a one-group pretest–posttest quasi-experimental design. The results indicate that participants completed engaging buzz wire game machines that incorporated energy concepts. Significant improvements were observed in participants’ engineering concepts, engineering design self-efficacy, computational thinking, and programming self-efficacy, with a considerable effect size noted in programming self-efficacy (N = 27, Cohen’s d = 1.84, p < 0.05). Most participants expressed highly positive attitudes toward the activity. The findings suggest that the engineering design activity developed in this study shows promising evidence of enhancing professional growth across multiple dimensions of engineering education for pre-service elementary teachers. The study serves as a valuable reference for future curriculum design in teacher education programs. Full article

We present a large-scale neuromodulation system-on-chip (SoC) that integrates a 128-channel neural recording and 32-channel stimulation ASIC designed for a wide range of neuroscientific applications. Each recording channel achieves low-noise performance (~4 μV_rms) with a configurable bandwidth of 0.05 Hz–7.5 kHz and supports 16-bit digitization with scalable sampling rates up to 30 kS/s. To enhance signal quality, the ASIC includes an adjustable digital high-pass filter and a fast-settling function for rapid recovery from stimulation artifacts. SoC also incorporates on-chip electrode-impedance measurements as a built-in safety feature by reusing the recording channels. The stimulation subsystem generates current-controlled monopolar biphasic pulses with a high compliance voltage of ±6 V using standard low-voltage (1.8 V/3.3 V) CMOS devices. Each of the 32 stimulation channels provides arbitrary 9-bit programmable waveforms and dual current modes (4 μA/bit and 8 μA/bit), supporting both fine-resolution microstimulation and high-current applications such as spinal-cord and deep-brain stimulation. On-chip charge-balancing switches in each channel further ensure safe and reliable stimulation delivery. The SoC supports digital communication via a standard SPI with both 3.3 V CMOS and low-voltage differential signaling options and integrates all required analog references and low-dropout regulators. The prototype was fabricated in a standard 180 nm CMOS process, occupying 31.92 mm² (equivalently, 0.2 mm² per recording-and-stimulation channel), and was fully validated through benchtop measurements and in vitro experiments. Full article

Generative steganography (GS) generates stego-media via secret messages, but existing GS only targets single-type multimedia data with poor universality. The generator and extractor sizes are highly coupled with resolution. Message mapping converts secret messages and noise, yet current GS schemes based on it use gridded data, failing to generate diverse multimedia universally. Inspired by implicit neural representation (INR), we propose generative implicit steganography via message mapping (GIS). We designed single-bit and multi-bit message mapping schemes in function domains. The scheme’s function generator eliminates the coupling between model and gridded data sizes, enabling diverse multimedia generation and breaking resolution limits. A dedicated point cloud extractor is trained for adaptability. Through a literature review, this scheme is the first to perform message mapping in the functional domain. During the experiment, taking images as an example, methods such as PSNR, StegExpose, and neural pruning were used to demonstrate that the generated image quality is almost indistinguishable from the real image. At the same time, the generated image is robust. The accuracy of message extraction can reach 96.88% when the embedding capacity is 1 bpp, 89.84% when the embedding capacity is 2 bpp, and 82.21% when the pruning rate is 0.3. Full article

You Only Look Once (YOLO) is a convolutional neural network-based object detection algorithm widely used in real-time vision applications. However, its high computational demand leads to significant power consumption and cost when deployed in graphics processing units. Field-programmable gate arrays offer a low-power alternative. However, their efficient implementation requires architecture-level optimization tailored to limited device resources. This study presents an optimized YOLOv2 accelerator for the Zynq-7000 system-on-chip (SoC). The design employs 16-bit integer quantization, a filter reuse structure, an input feature map reuse scheme using a line buffer, and tiling parameter optimization for the convolution and max pooling layers to maximize resource efficiency. In addition, a stall-based control mechanism is introduced to prevent structural hazards in the pipeline. The proposed accelerator was implemented on the Zynq-7000 SoC board, and a system-level evaluation confirmed a negligible accuracy drop of only 0.2% compared with the 32-bit floating-point baseline. Compared with previous YOLO accelerators on the same SoC, the design achieved up to 26% and 15% reductions in flip-flop and digital signal processor usage, respectively. This result demonstrates feasible deployment on XC7Z020 with DSP 57.27% and FF 16.55% utilization. Full article

This paper presents a high-gain wide-band planar antenna with a Reconfigurable Intelligent Surface (RIS) for modern wireless communication applications. The antenna consists of two main parts, a basic antenna part with cross-line slots and two light-dependent resistor switches, and a second part based on the RIS layer for beam steering. The RIS is constructed from 5 × 5-unit cells with two sides, forming a square geometry. The antenna substrate is a dielectric layer of FR4 epoxy glass with a thickness of 1.6 mm. The RIS inclusions are designed and tested numerically to achieve the desired electromagnetic properties at the frequency band of interest. The fabricated prototype shows a wide band covering frequencies from 0.9 GHz to 3.5 GHz with S11 below −10 dB, achieving an antenna gain varying from 10.5 dBi up to 16.8 dBi. Experimental measurements show effective aperture usage in all configurations, and beam steering from +22° to −22° is accomplished without degrading side-lobe levels. The proposed antenna performance is tested against real-world measurements to evaluate channel performance in terms of bit error rate (BER) and channel capacity (CC). The proposed LDR-controlled design achieves compact beam steering with minimal insertion loss, unlike conventional RIS-assisted antennas that rely on PIN or varactor switches. Full article

The deterioration of rocks’ mechanical properties during the late stage of water injection development significantly reduces the rock-breaking efficiency of PDC bits. In this study, X-ray diffraction mineral composition analysis and triaxial compression mechanics tests were used to systematically characterize the weakening mechanism of water injection on reservoir rocks. Based on an analysis of mechanical experimental characteristics, this study proposes a multi-scale collaborative optimization method: establish a single tooth–rock interaction model at the micro-scale through finite element simulation to optimize geometric cutting parameters; at the macro scale, adopt a differential bit design scheme. By comparing and analyzing the rock-breaking energy consumption characteristics of four-blade and five-blade bits, the most efficient rock-breaking configuration can be optimized. Based on Fluent simulation on the flow field scale, the nozzle configuration can be optimized to improve the bottom hole flow field. The research results provide important theoretical guidance and technical support for the personalized design of drill bits in the later stage of water injection development. Full article

Scalar multiplication is the basis of the widespread elliptic curve public key cryptography. Standard scalar multiplication is vulnerable to side-channel attacks that are able to infer the secret bit values by observing the power or delay traces. This work utilizes the arithmetic properties of scalar multiplication to propose two scalar multiplication algorithms to insulate ECC implementations from side-channel attacks. The two proposed designs rely on randomly permuting the ordering and storage locations of the different scalar multiplication values $2^{i}G$ as well as the corresponding secret key bits $k_i$. Statistical analysis and Python 3.9.13implementations confirm the validity of the two algorithms. Numerical results confirm that both designs produce the same results as the standard right-to-left scalar multiplication algorithm. Welch’s t-test as well as numerical simulations confirm the immunity of our proposed protocols to side-channel attacks. Full article

This work extends our previous research on spectral clustering for mitigating nonlinear phase noise in optical communication systems by presenting the first complete FPGA implementation of the algorithm, including on-chip eigenvector computation with parallelization strategies. The implementation addresses the computational complexity challenges of spectral clustering through a heterogeneous CPU/FPGA co-design approach that partitions algorithmic stages between ARM processors and the FPGA fabric. While the achieved processing speeds of approximately 36 symbols per second do not yet meet the requirements for commercial optical transceivers, our hardware prototype demonstrates the feasibility and practical challenges of deploying advanced clustering algorithms on real-time hardware architectures. We detail the parallel Jacobi method for eigenvector computation, the Greedy K-means++ initialization strategy, and the comprehensive hardware mapping of all clustering stages. The system processes streaming m-QAM data through a windowed architecture and integrates a demapper to ensure label consistency, demonstrating improved bit error rate performance compared to K-means under severe phase noise conditions of −90 dBc/Hz at a 1 MHz offset. This implementation offers valuable insights into memory bandwidth limitations and resource utilization trade-offs, underscoring the crucial role of FPGAs as a bridge between algorithm development and high-speed optical system deployment. Full article

Underwater wireless optical communication (UWOC) is a key technology in ocean resource development, and its link stability is often limited by the difficulty of optical alignment in complex underwater environments. In response to this difficulty, this study has focused on improving the Real-Time Detection Transformer v2 (RT-DETRv2) model. We have improved the underwater light source detection model by collaboratively designing a lightweight backbone network and deformable convolution, constructing a cross-stage local attention mechanism to reduce the number of network parameters, and introducing geometrically adaptive convolution kernels that dynamically adjust the distribution of sampling points, enhance the representation of spot-deformation features, and improve positioning accuracy under optical interference. To verify the effectiveness of the model, we have constructed an underwater light-emitting diode (LED) light-spot detection dataset containing 11,390 images was constructed, covering a transmission distance of 15–40 m, a ±45° deflection angle, and three different light-intensity conditions (noon, evening, and late night). Experiments show that the improved model achieves an average precision at an intersection-over-union threshold of 0.50 (AP50) value of 97.4% on the test set, which is 12.7% higher than the benchmark model. The UWOC system built based on the improved model achieves zero-bit-error-rate communication within a distance of 30 m after assisted alignment (an initial lateral offset angle of 0°–60°), and the bit-error rate remains stable in the 10⁻⁷–10⁻⁶ range at a distance of 40 m, which is three orders of magnitude lower than the traditional Remotely Operated Vehicle (ROV) underwater optical communication system (a bit-error rate of 10⁻⁶–10⁻³), verifying the strong adaptability of the improved model to complex underwater environments. Full article

Elastic optical networks (EONs) must allocate resources dynamically to accommodate heterogeneous, high-bandwidth demands. However, the continuous setup and teardown of connections with different bit rates can fragment the spectrum and lead to blocking. The blocking predictors enable proactive defragmentation and resource reallocation within network controllers. In this paper, we propose two novel deep learning models (based on CNN–BiLSTM and CNN–LSTM) to predict blocking in EONs by combining spatial feature extraction from spectrum snapshots using 2D convolutional layers with temporal sequence modeling. This hybrid spatio-temporal design learns how local fragmentation patterns evolve over time, allowing it to detect impending blocking scenarios more accurately than conventional methods. We evaluate our model on the simulated NSFNET topology and compare it against multiple baselines, namely 1D CNN, 2D CNN, k-nearest neighbors (KNN), and support vector machines (SVMs). The results show that the proposed CNN–BiLSTM/LSTM models consistently achieve higher performance. The CNN–BiLSTM model achieved the highest accuracy in blocking prediction, while the CNN–LSTM model shows slightly lower accuracy; however, it has much lower complexity and a faster learning time. Full article

The paper presents experimental results for a modified pulsed plasma thruster (PPT) with solid propellant, using a coaxial anode–cathode design. Graphite from pencil leads served as propellant, and a tungsten trigger electrode was tested to reduce carbonization effects. Experiments were performed in a vacuum chamber at 0.001 Pa, employing diagnostics such as discharge current/voltage recording, power measurement, ballistic pendulum, time-of-flight (TOF) method, and a Faraday cup. Current and voltage waveforms matched an oscillatory RLC circuit with variable plasma channel resistance. Key discharge parameters were measured, including current pulse duration/amplitude and plasma channel formation/decay dynamics. Impulse bit values, obtained with a ballistic pendulum, reached up to 8.5 μN·s. Increasing trigger capacitor capacitance reduced thrust due to unstable “pre-plasma” formation and partial pre-discharge energy loss. Using TOF and Faraday cup diagnostics, plasma front velocity, ion current amplitude, current density, and ion concentration were determined. Tungsten electrodes produced lower charged particle concentrations than graphite but offered better adhesion resistance, minimal carbonization, and stable long-term performance. The findings support optimizing trigger electrode materials and PPT operating modes to extend lifetime and stabilize thrust output. Full article

This paper presents an approach to error correction in wireless communication systems, with a focus on the Bluetooth Low Energy standard. Our method uses the redundancy provided by the cyclic redundancy check and leaves the transmitter unchanged. The approach has two components: an error-detection algorithm that validates data packets and a neural network with reject option that classifies signals received from the channel and identifies bit errors for later correction. This design localizes and corrects errors and reduces transmission failures. Extensive simulations were conducted, and the results demonstrated promising performance. The method achieved correction rates of 94–98% for single-bit errors and 54–68% for double-bit errors, which reduced the need for packet retransmissions and lowered the risk of data loss. When applied to images, the approach enhanced visual quality compared with baseline methods. In particular, we observed improvements in visual quality for signal-to-noise ratios between 9 and 11 dB. In many cases, these enhancements were sufficient to restore the integrity of corrupted images. Full article

The completeness and integrity of multimodal medical data are critical determinants of surgical success and postoperative recovery. However, because of issues such as poor sensor contact, small vibrations, and device discrepancies during signal acquisition, there are frequent missing values in patients’ medical data. This issue is especially prominent in rare or complex cases, where the inherent complexity and sparsity of multimodal data limit dataset diversity and degrade predictive model performance. As a result, clinicians’ understanding of patient conditions is restricted, and the development of robust algorithms to predict preoperative, intraoperative, and postoperative disease progression is hindered. To address these challenges, we propose Med-Diffusion, a diffusion-based generative framework designed to enhance sensor data by imputing missing multimodal clinical data, including both categorical and numerical variables. The framework integrates one-hot encoding, simulated bit encoding, and feature tokenization to improve adaptability to heterogeneous data types, utilizing conditional diffusion modeling for accurate data completion. Med-Diffusion effectively learns the underlying distributions of multimodal datasets, synthesizing plausible data for incomplete records, and it mitigates the data sparsity caused by poor sensor contact, vibrations, and device discrepancies. Extensive experiments demonstrate that Med-Diffusion accurately reconstructs missing multimodal clinical information and significantly enhances the performance of downstream predictive models. Full article