Jun Zhang

Liheng Bian, Haoze Song, Lintao Peng et al.

Benefiting from its single-photon sensitivity, single-photon avalanche diode (SPAD) array has been widely applied in various fields such as fluorescence lifetime imaging and quantum computing. However, large-scale high-fidelity single-photon imaging remains a big challenge, due to the complex hardware manufacture craft and heavy noise disturbance of SPAD arrays. In this work, we introduce deep learning into SPAD, enabling super-resolution single-photon imaging over an order of magnitude, with significant enhancement of bit depth and imaging quality. We first studied the complex photon flow model of SPAD electronics to accurately characterize multiple physical noise sources, and collected a real SPAD image dataset (64 $\times$ 32 pixels, 90 scenes, 10 different bit depth, 3 different illumination flux, 2790 images in total) to calibrate noise model parameters. With this real-world physical noise model, we for the first time synthesized a large-scale realistic single-photon image dataset (image pairs of 5 different resolutions with maximum megapixels, 17250 scenes, 10 different bit depth, 3 different illumination flux, 2.6 million images in total) for subsequent network training. To tackle the severe super-resolution challenge of SPAD inputs with low bit depth, low resolution, and heavy noise, we further built a deep transformer network with a content-adaptive self-attention mechanism and gated fusion modules, which can dig global contextual features to remove multi-source noise and extract full-frequency details. We applied the technique on a series of experiments including macroscopic and microscopic imaging, microfluidic inspection, and Fourier ptychography. The experiments validate the technique's state-of-the-art super-resolution SPAD imaging performance, with more than 5 dB superiority on PSNR compared to the existing methods.

Chenxin An, Fei Huang, Jun Zhang et al.

The ability of Large Language Models (LLMs) to process and generate coherent text is markedly weakened when the number of input tokens exceeds their pretraining length. Given the expensive overhead of finetuning large-scale models with longer sequences, we propose Dual Chunk Attention (DCA), which enables Llama2 70B to support context windows of more than 100k tokens without continual training. By decomposing the attention computation for long sequences into chunk-based modules, DCA manages to effectively capture the relative positional information of tokens within the same chunk (Intra-Chunk) and across distinct chunks (Inter-Chunk), as well as integrates seamlessly with Flash Attention. In addition to its impressive extrapolation capability, DCA achieves performance on practical long-context tasks that is comparable to or even better than that of finetuned models. When compared with proprietary models, our training-free 70B model attains 94% of the performance of gpt-3.5-16k, indicating it is a viable open-source alternative. All code and data used in this work are released at \url{https://github.com/HKUNLP/ChunkLlama}.

Jiajia Guo, Peiwen Jiang, Chao-Kai Wen et al.

Accurate channel state information (CSI) is critical to the performance of wireless communication systems, especially with the increasing scale and complexity introduced by 5G and future 6G technologies. While artificial intelligence (AI) offers a promising approach to CSI acquisition and utilization, existing methods largely depend on task-specific neural networks (NNs) that require expert-driven design and large training datasets, limiting their generalizability and practicality. To address these challenges, we propose LVM4CSI, a general and efficient framework that leverages the structural similarity between CSI and computer vision (CV) data to directly apply large vision models (LVMs) pre-trained on extensive CV datasets to wireless tasks without any fine-tuning, in contrast to large language model-based methods that generally necessitate fine-tuning. LVM4CSI maps CSI tasks to analogous CV tasks, transforms complex-valued CSI into visual formats compatible with LVMs, and integrates lightweight trainable layers to adapt extracted features to specific communication objectives. We validate LVM4CSI through three representative case studies, including channel estimation, human activity recognition, and user localization. Results demonstrate that LVM4CSI achieves comparable or superior performance to task-specific NNs, including an improvement exceeding 9.61 dB in channel estimation and approximately 40% reduction in localization error. Furthermore, it significantly reduces the number of trainable parameters and eliminates the need for task-specific NN design.

Bing Bai, Jianyao Huang, Guan-Ru Qiao et al.

Quantum random number generators (QRNGs) can produce true random numbers. Yet, the two most important QRNG parameters highly desired for practical applications, i.e., speed and size, have to be compromised during implementations. Here, we present the fastest and miniaturized QRNG with a record real-time output rate as high as 18.8 Gbps by combining a photonic integrated chip and the technology of optimized randomness extraction. We assemble the photonic integrated circuit designed for vacuum state QRNG implementation, InGaAs homodyne detector and high-bandwidth transimpedance amplifier into a single chip using hybrid packaging, which exhibits the excellent characteristics of integration and high-frequency response. With a sample rate of 2.5 GSa/s in a 10-bit analog-to-digital converter and subsequent paralleled postprocessing in a field programmable gate array, the QRNG outputs ultrafast random bitstreams via a fiber optic transceiver, whose real-time speed is validated in a personal computer.

Pengfei Gao, Hongyi Xie, Jun Zhang et al.

Power side-channel attacks, which can deduce secret data via statistical analysis, have become a serious threat. Masking is an effective countermeasure for reducing the statistical dependence between secret data and side-channel information. However, designing masking algorithms is an error-prone process. In this paper, we propose a hybrid approach combing type inference and model-counting to verify masked arithmetic programs against side-channel attacks. The type inference allows an efficient, lightweight procedure to determine most observable variables whereas model-counting accounts for completeness. In case that the program is not perfectly masked, we also provide a method to quantify the security level of the program. We implement our methods in a tool QMVerif and evaluate it on cryptographic benchmarks. The experimental results show the effectiveness and efficiency of our approach.

You-Qi Nie, Jian-Yu Guan, Hongyi Zhou et al.

The randomness from a quantum random number generator (QRNG) relies on the accurate characterization of its devices. However, device imperfections and inaccurate characterizations can result in wrong entropy estimation and bias in practice, which highly affects the genuine randomness generation and may even induce the disappearance of quantum randomness in an extreme case. Here we experimentally demonstrate a measurement-device-independent (MDI) QRNG based on time-bin encoding to achieve certified quantum randomness even when the measurement devices are uncharacterized and untrusted. The MDI-QRNG is randomly switched between the regular randomness generation mode and a test mode, in which four quantum states are randomly prepared to perform measurement tomography in real-time. With a clock rate of 25 MHz, the MDI-QRNG generates a final random bit rate of 5.7 Kbps. Such implementation with an all-fiber setup provides an approach to construct a fully-integrated MDI-QRNG with trusted but error-prone devices in practice.