The emergence of fine-grained numerical formats like NVFP4 presents new opportunities for efficient Large Language Model (LLM) inference.However, it is difficult to adapt existing Post-Training Quantization (PTQ) strategies to these formats: rotation-based methods compromise fine-grained block isolation; smoothing techniques struggle with significant 4-bit quantization errors; and mixed-precision approaches often conflict with hardware constraints on unified-precision computation.To address these challenges, we propose ARC-Quant, a framework that boosts NVFP4 performance via Augmented Residual Channels.Distinct from methods that compromise block isolation or hardware uniformity, ARCQuant maintains a strictly unified NVFP4 format by augmenting the activation matrix with quantized residual channels.This design integrates the error compensation process directly into the matrix reduction dimension, enabling the use of standard, highly optimized GEMM kernels with minimal overhead.Theoretical analysis confirms that the worst-case error bound of our dual-stage NVFP4 quantization is comparable to that of standard 8-bit formats such as MXFP8.Extensive experiments on LLaMA and Qwen models demonstrate that ARCQuant achieves state-of-the-art accuracy, comparable to full-precision baselines in perplexity and downstream tasks.Furthermore, deployment on RTX 5090 and RTX PRO 6000 GPUs confirms practical benefits, achieving up to 3× speedup over FP16.Our code is available at https://github.com/ actypedef/ARCQuant.

One of the main roadblocks to Atomic Force Microscope (AFM) based nanomanipulation is lack of real time visual feedback. Although the model based visual feedback can partly solve this problem, its unguaranteed reliability due to the inaccurate models in nano-environment still limits the efficiency of AFM based nanomanipulation. This paper introduce a Real-time Fault Detection and Correction (RFDC) method to improve the reliability of the visual feedback. By utilizing Kalman filter and local scan technologies, the RFDC method not only can real-time detect the fault display caused by the modeling error, but also can on-line correct it without interrupting manipulation. In this way, the visual feedback keeps consistent with the true environment changes during manipulation, which makes several operations being finished without a image scanning in between. The theoretical study and the implementation of the RFDC method are elaborated in this paper. Experiments of manipulating nano-particles have been carried out to demonstrate the effectiveness and efficiency of the proposed method.

Atomic Force Microscope (AFM) has been used as a manipulation tool for a decade. The problem of lacking real time visual feedback still limits its efficiency and hinders its wide application. Although the model based visual feedback can partly solve this problem, due to the complexity of nano environment, it is difficult to use a model to accurately describe the object’s behavior. The modeling error will give the operator a false feedback and lead to a failed manipulation. In this paper, a strategy for visual feedback error on-line detection and correction is proposed to solve this problem. As the real time force information is a key factor for this strategy, an adaptable end effector is employed to accurately measure the interaction force between the probe and the nano-objects, and the system error is also compensated to improve the accuracy of interaction force measurement. Based on the true real time force information, an extended Kalman filter is developed to online detect whether there is a false feedback. Once a false feedback is detected, an optimal searching pattern is generated to get the real manipulation result in a short time. With the assistance of this strategy, the false visual feedback can be realtime detected and corrected without interrupting manipulation. Complex manipulation task can be finished without being interrupted by a new image scan. Experiments of manipulating nano-particles are performed to verify the effectiveness of this strategy, which demonstrated the improved efficiency of the AFM based nano-assembly system.

This paper describes the development of a human airbag system which is designed to reduce the impact force from falls. A micro inertial measurement unit (muIMU), based on MEMS accelerometers and gyro sensors is developed as the motion sensing part of the system. A recognition algorithm is used for real-time fall determination. With the algorithm, a microcontroller integrated with the muIMU can discriminate falling-down motion from normal human motions and trigger an airbag system when a fall occurs. Our airbag system is designed to have fast response with moderate input pressure, i.e., the experimental response time is less than 0.3 second under 0.4 MPa. In addition, we present our progress on using support vector machine (SVM) training together with the muIMU to better distinguish falling and normal motions. Experimental results show that selected eigenvector sets generated from 200 experimental data sets can be accurately separated into falling and other motions