Related papers: Recent Results from the EVN MkIV Data Processor at…
The Arcminute Microkelvin Imager (AMI) telescopes located at the Mullard Radio Astronomy Observatory near Cambridge have been significantly enhanced by the implementation of a new digital correlator with 1.2 MHz spectral resolution. This…
Recent vision-language (VL) studies have shown remarkable progress by learning generic representations from massive image-text pairs with transformer models and then fine-tuning on downstream VL tasks. While existing research has been…
Computing-in-memory (CIM) is renowned in deep learning due to its high energy efficiency resulting from highly parallel computing with minimal data movement. However, current SRAM-based CIM designs suffer from long latency for loading…
Jacobian-vector products (JVPs) form the backbone of many recent developments in Deep Networks (DNs), with applications including faster constrained optimization, regularization with generalization guarantees, and adversarial example…
We present Qwen-Image-VAE-2.0, a suite of high-compression Variational Autoencoders (VAEs) that achieve significant advances in both reconstruction fidelity and diffusability. To address the reconstruction bottlenecks of high compression,…
As modern astronomy confronts unprecedented data volumes, automated pipelines and machine-learning techniques have become essential for processing and analysis. As these workflows grow more complex, astronomers also require input and…
Smart Internet of Vehicles (IoV) as a promising application in Internet of Things (IoT) emerges with the development of the fifth generation mobile communication (5G). Nevertheless, the heterogeneous requirements of sufficient battery…
We report on a search for radio transients at 340 MHz with the Jansky Very Large Array (VLA) Low band Ionospheric and Transient Experiment (VLITE). Between 2015 July 29 and 2015 September 27, operating in commensal mode, VLITE imaged…
We present the current status of two new fully automated reduction and analysis pipelines, built for the Euler telescope and the CORALIE spectrograph. Both pipelines have been designed and built independently at the Universidad de Chile and…
Multimodal vision language models (VLMs) have made significant progress with the support of continuously increasing model sizes and data volumes. Running VLMs on edge devices has become a challenge for their widespread application. There…
The Advanced Laser Interferometer Gravitational-wave Observatory (LIGO) detectors have completed their initial upgrade phase and will enter the first observing run in late 2015, with detector sensitivity expected to improve in future runs.…
We report on our experiences using the Metsahovi Radio Observatory's (MRO) VLBI Standard Interface (VSI, Whitney 2002) recorder in a number of astronomical applications. The PC-EVN device is a direct memory access (DMA) interface which…
Efficient image tokenization with high compression ratios remains a critical challenge for training generative models. We present SoftVQ-VAE, a continuous image tokenizer that leverages soft categorical posteriors to aggregate multiple…
Obtaining high-resolution images at centimeter-or-longer wavelengths is vital for understanding the physics of jets. We reconstructed images from the M87 22 GHz data observed with the East Asian VLBI Network (EAVN) by using the regularized…
Vision-Language Models (VLMs) face severe memory and latency bottlenecks due to high-resolution visual tokens. While current token reduction methods theoretically save FLOPs, post-hoc pruning introduces structural overhead, failing to yield…
Parallel accelerators, such as GPUs, are key enablers for large-scale Machine Learning (ML) applications. However, ML model developers often lack detailed knowledge of the underlying system architectures, while system programmers usually do…
SS 433 is a microquasar whose relativistic jets precess every ~162 days, providing a laboratory for jet-interstellar medium interactions. We present a comprehensive analysis of 16 years of Fermi Large Area Telescope data (August…
LLMs have low GPU efficiency and high latency due to autoregressive decoding. Speculative decoding (SD) mitigates this using a small draft model to speculatively generate multiple tokens, which are then verified in parallel by a target…
Key-Value (KV) cache has become a de facto component of modern Large Vision-Language Models (LVLMs) for inference. While it enhances decoding efficiency in Large Language Models (LLMs), its direct adoption in LVLMs introduces substantial…
Large vision-language models (LVLMs) have shown premise in a broad range of vision-language tasks with their strong reasoning and generalization capabilities. However, they require considerable computational resources for training and…