arXiv subjects

David Peng

Publications and source records attributed to David Peng.

4 recordsLinked to original sources

arXiv2026

Projected sensitivity of the ANUBIS detector to heavy neutral leptons

Long-Lived Particles (LLPs) are a common feature in various extensions to the Standard Model (SM) that seek to address known limitations. The ANUBIS detector has been proposed to extend the sensitivity of the ATLAS experiment at the LHC to LLPs by instrumenting the ceiling of the ATLAS detector cavern. This article presents the projected sensitivity of ANUBIS to Heavy Neutral Leptons (HNLs). For a minimal Majorana HNL model that only couples to a single flavour of lepton ($e$ or $\mu$) ANUBIS reaches a maximum sensitivity of $|V_{1e}|^2=1.8\times10^{-8}$ and $|V_{1\mu}|^2=1.9\times10^{-8}$ for a HNL mass of $m_{N_1}=6.4$ GeV and 6.3 GeV respectively. This provides complementary coverage to other proposed LLP experiments in the HNL parameter-space, with potential for significant improvement during ANUBIS data-taking through advances in analysis strategies. The results are obtained with SET-ANUBIS, a flexible framework to evaluate the sensitivity of ANUBIS to a variety of LLP models.

hep-ex↗

arXiv2022

FOLIO: Natural Language Reasoning with First-Order Logic

Large language models (LLMs) have achieved remarkable performance on a variety of natural language understanding tasks. However, existing benchmarks are inadequate in measuring the complex logical reasoning capabilities of a model. We present FOLIO, a human-annotated, logically complex and diverse dataset for reasoning in natural language (NL), equipped with first-order logic (FOL) annotations. FOLIO consists of 1,430 examples (unique conclusions), each paired with one of 487 sets of premises used to deductively reason for the validity of each conclusion. The logical correctness of the premises and conclusions is ensured by their FOL annotations, which are automatically verified by an FOL inference engine. In addition to the main NL reasoning task, NL-FOL pairs in FOLIO constitute a new NL-FOL translation dataset. Our experiments on FOLIO systematically evaluate the FOL reasoning ability of supervised fine-tuning on medium-sized language models. For both NL reasoning and NL-FOL translation, we benchmark multiple state-of-the-art language models. Our results show that a subset of FOLIO presents a challenge for one of the most capable {Large Language Model (LLM)} publicly available, GPT-4.

cs.CL↗

arXiv2021

NetDAM: Network Direct Attached Memory with Programmable In-Memory Computing ISA

Data-intensive applications like distributed AI-training may require multi-terabytes memory capacity with multi-terabits bandwidth. We directly attach the memory to the ethernet controller with some programable logic to design an efficient hardware "template" for Memory pooling and in-memory / in-network computing. We built an FPGA prototype of the NetDAM, andwe demonstrate MPI-Allreduce communication case, the NetDAM can be used as a software and hardware friendly programmable architeture with high performance alternative for RDMA.

cs.DC↗

arXiv2021

Exploring the Sensory Spaces of English Perceptual Verbs in Natural Language Data

In this study, we explore how language captures the meaning of words, in particular meaning related to sensory experiences learned from statistical distributions across texts. We focus on the most frequent perception verbs of English analyzed from an and Agentive vs. Experiential distinction across the five basic sensory modalities: Visual (to look vs. to see), Auditory (to listen vs. to hear), Tactile (to touch vs. to feel), Olfactory (to smell), and Gustatory (to taste). In this study we report on a data-driven approach based on distributional-semantic word embeddings and clustering models to identify and uncover the descriptor sensory spaces of the perception verbs. In the analysis, we identified differences and similarities of the generated descriptors based on qualitative and quantitative differences of the perceptual experience they denote. For instance, our results show that while the perceptual spaces of the experiential verbs like to see, to hear show a more detached, logical way of knowing and learning, their agentive counterparts (to look, listen) provide a more intentional as well as more intimate and intuitive way of discovering and interacting with the world around us. We believe that such an approach has a high potential to expand our understanding and the applicability of such sensory spaces to different fields of social and cultural analysis. Research on the semantic organization of sensory spaces for various applications might benefit from an the Agentive/Experiential account to address the complexity of multiple senses wired with each other in still unexplored ways.

cs.CL↗