WSVD: Weighted Low-Rank Approximation for Fast and Efficient Execution of Low-Precision Vision-Language Models

Singular Value Decomposition (SVD) has become an important technique for reducing the computational burden of Vision Language Models (VLMs), which play a central role in tasks such as image captioning and visual question answering. Although multiple prior works have proposed efficient SVD variants to enable low-rank operations, we find that in practice it remains difficult to achieve substantial latency reduction during model execution. To address this limitation, we introduce a new computational pattern and apply SVD at a finer granularity, enabling real and measurable improvements in execution latency. Furthermore, recognizing that weight elements differ in their relative importance, we adaptively allocate relative importance to each element during SVD process to better preserve accuracy, then extend this framework with quantization applied to both weights and activations, resulting in a highly efficient VLM. Collectively, we introduce~\textit{Weighted SVD} (WSVD), which outperforms other approaches by achieving over $1.8\times$ decoding speedup while preserving accuracy. We open source our code at: \href{https://github.com/SAI-Lab-NYU/WSVD}{\texttt{https://github.com/SAI-Lab-NYU/WSVD}

Human Respiration Detection Under Interference: Challenges and Solutions

Recent research has highlighted the detection of human respiration rate using commodity WiFi devices. Nevertheless, these devices encounter challenges in accurately discerning human respiration amidst the prevailing human motion interference encountered in daily life. To tackle this predicament, this paper introduces a passive sensing and communication system designed specifically for respiration detection in the presence of robust human motion interference. Operating within the 60.48GHz band, the proposed system aims to detect human respiration even when confronted with substantial human motion interference within close proximity. Subsequently, a neural network is trained using the collected data by us to enable human respiration detection. The experimental results demonstrate a consistently high accuracy rate over 90\% of the human respiration detection under interference, given an adequate sensing duration. Finally, an empirical model is derived analytically to achieve the respiratory rate counting in 10 seconds.