Fully-sensorized smart-eyewear platform for on-device Machine Learning

This paper presents ARGO, a smart eyewear platform designed to bridge ergonomic comfort, high computational throughput, and energy efficiency. Unlike cloud-dependent solutions, ARGO leverages the STM32N6 microcontroller and its integrated Neural Processing Unit (NPU) to enable on-device machine learning, minimizing latency and preserving user privacy through local data processing. The primary contribution lies in the holistic co-design of hardware, firmware, and artificial intelligence, centered on the deployment of an optimized YOLOv11 model for real-time urban obstacle recognition. To ensure compatibility with the target NPU, we introduce Head-wise Parallel Attention (HPA), an architectural refinement that enables efficient accelerator execution while preserving the original computational logic. The model is trained on the Walking On The Road (WOTR) dataset, and the final deployed configuration achieves an mAP50-95 of 24 under strict memory constraints, with a memory footprint of only 2.483 MB. The platform integrates a multimodal sensor suite, RGB cameras, Time-of-Flight sensors, microphones, and ambient sensors, and delivers 10 FPS at a continuous autonomy of ~113 minutes on a 200 mAh battery. These results demonstrate the feasibility of a high-performance, privacy-preserving, and socially acceptable assistive device, and highlight how competitive edge AI solutions increasingly demand tightly integrated, multidisciplinary co-design approaches.

When Good Enough Is Optimal: Multiplication-Only Matrix Inversion Approximation for Quantized Gated DeltaNet

Matrix inversion in chunk-wise parallel linear attention is a major bottleneck for long-context modeling, particularly on NPUs, where forward-substitution-based methods exhibit limited parallelism and poor hardware utilization. We propose a fast, Matrix Multiplication (MatMul)-based algorithm tailored for strictly lower-triangular matrices arising in chunk-wise linear attention. Motivated by the rapid growth of Neumann-series terms and the diagonal concentration of the inverse matrix, we employ a truncated Neumann expansion with structural masking and parallel residual correction to eliminate sequential dependencies. We further extend our method to low-bits INT by mitigating the dynamic range expansion arising from repeated matrix power operations, and adapt the approximation order and residual step to the chunk size to minimize computational cost while preserving the model's accuracy. Experiments on Qwen3.5-family models demonstrate up to 5$\times$ kernel-level speedup and a 20% reduction in decode-layer overhead, while preserving accuracy under both floating-point and low-precision inference. Our method offers an efficient and hardware-friendly solution for scalable linear attention.

Fast and Stable Triangular Inversion for Delta-Rule Linear Transformers

Linear attention has emerged as a cornerstone for efficient long-context architectures, as evidenced by its integration into state-of-the-art open-source models including Qwen3.5/3.6, Kimi Linear, and RWKV-7. Models that incorporate linear attention layers with the so-called Delta-Rule involve the inversion of triangular matrices as a core sub-routine. This operation often forms a performance bottleneck, and, due to its high-sensitivity to numerical errors, it can significantly deteriorate end-to-end model accuracy if it is not carefully implemented. This work provides a systematic analysis of both direct and iterative triangular inversion algorithms, targeting methods that are rich in matrix products, and, therefore, have the potential to efficiently utilize modern hardware. To that end, our analysis covers a broad spectrum of mathematical and practical aspects, with a heavy focus on numerical stability, computational complexity, and, ultimately, hardware efficiency and practical considerations. We provide a rigorous experimental evaluation to verify these properties in practical scenarios, and in low-precision floating-point representations, highlighting the strengths and limitations of each method. Performance benchmarks on NPUs reveal up to $4.3\times$ speed-up against the state-of-the-art implementations of SGLang for triangular matrix inversion, leading to significant performance improvements on the entire layer level, while maintaining full end-to-end model accuracy.