83308 research outputs found
Sort by
Mosaic: composite projection pruning for resource-efficient LLMs
No full textExtensive compute and memory requirements limit the deployment of large language models (LLMs) on any hardware. Compression methods, such as pruning, can reduce model size, which in turn reduces resource requirements. State-of-the-art pruning is based on coarse-grained methods. They are time-consuming and inherently remove critical model parameters, adversely impacting the quality of the pruned model. This paper introduces projection pruning, a novel fine-grained method for pruning LLMs. In addition, LLM projection pruning is enhanced by a new approach we refer to as composite projection pruning — the synergistic combination of unstructured pruning that retains accuracy and structured pruning that reduces model size. We develop Mosaic, a novel system to create and deploy pruned LLMs using composite projection pruning. Mosaic is evaluated using a range of performance and quality metrics on multiple hardware platforms, LLMs, and datasets. Mosaic is 7.19 faster in producing models than existing approaches. Mosaic models achieve up to 84.2% lower perplexity and 31.4% higher accuracy than models obtained from coarse-grained pruning. Up to 67% faster inference and 68% lower GPU memory use is noted for Mosaic models
Carbon and nitrogen cycling in response to global environmental change during the Carnian Pluvial Episode (late Triassic)
No full textThe Carnian Pluvial Episode (CPE) (∼234 Ma) was marked by recurrent climate shifts, changes in the hydrological cycle, in particular precipitation pattern, which in turn affected biogeochemical cycles and nutrient supply to shelf and epicontinental seas. Among biogeochemical cycles, the nitrogen cycle represents a fundamental process governing the transformation and distribution of nitrogenous species, including dinitrogen (N₂), ammonium (NH₄+), nitrite (NO₂−), and nitrate (NO₃−). However, due to the lack of detailed studies, the response of the nitrogen cycle to the environmental changes during the CPE remains poorly understood. This study presents organic carbon (δ13Corg) and nitrogen isotopes (δ15N) along with bulk elemental geochemical data from the Carnian Quemo Co section in the Qiangtang Basin, China. The δ13Corg chemostratigraphy indicates that the sediment succession investigated spans the onset of the CPE, which is marked by successive short-term negative carbon isotope excursions (NCIEs). Lithological observations, coupled with elevated Si/Al and Al/K ratios suggest an increased input of silt and chemically altered clay minerals, in which may related to increased hydrological cycle during the CPE. A positive δ15N excursion during the NCIE interval suggests increased nutrient fluxes to marine environments, possibly driven by increased terrestrial organic matter input resulting from enhanced erosion and weathering, stimulating nitrogen uptake by marine phytoplankton and partial nitrate assimilation
Visualising historical university records to explore the colonial connections of the students and alumni of the University of St Andrews, 1700-1897
No full textThis paper uses the historical student records of the University of St Andrews to demonstrate how an exploration of the student community and the alumni network can be a valuable contribution to our understanding of the ways UK universities have been implicated in British colonial and imperial activities. It also contributes to the investigation of student origins, international student mobility, and alumni career patterns. We used digital technology to identify over 900 individuals who either came from the colonies, or later lived or worked there, and who studied at (or were accredited by) the University of St Andrews during the period 1700 - 1897. We developed a web-based, custom-built, and interactive data visualization to enable us – and other scholars – to gain new perspectives and a deeper understanding of the chronological, geographical, and career trends within the records. The visualisation allowed us to explore these trends in finer detail than with traditional tabular statistics, and it was crucial in directing our attention to particular groups within the data (e.g., people with medical careers in British India or women students in southern Africa). We close with reflections on the challenges involved in using information visualization on incomplete and inconsistent historical records
Unpacking willpower in unassisted smoking cessation:a qualitative analysis reveals a systematic profile of situational and cognitive strategies
No full textBackgroundOver half of those who quit smoking do so without formal assistance, yet the psychological processes supporting unassisted cessation remain little understood. Success is often attributed to willpower, an umbrella term that lacks explanatory precision and obscures the underlying tractable processes. Drawing on the Process Model of Self-Regulation and the Behavior Change Technique (BCT) Taxonomy, this study aimed to identify the concrete strategies that enable individuals to quit smoking unassisted, thereby clarifying what willpower might look like in practice.Materials and methodsThirty-two participants who had successfully quit smoking without formal support participated in semi-structured interviews. Inductive content analysis identified key challenges, while deductive coding mapped strategies addressing these challenges to the Process Model of Self-Regulation and the BCT Taxonomy.ResultsParticipants’ accounts reflected a diverse range of strategies, averaging seven distinct BCTs, spanning the Situation Selection and Modification, Attention Redeployment, and Cognitive Change stages from the Process Model. Common BCTs included avoiding environmental triggers, substituting smoking with alternative behaviors, and seeking social support. In contrast, Response Modulation (e.g. ‘just say no’) accounted for only 1% of the data.ConclusionUnassisted quitters drew from a sophisticated repertoire of strategies that are actionable, teachable, and embedded within the individual’s physical and social environment. The qualitative methodology used in this study offers an understanding of the lived experiences of self-quitters, potentially informing public health interventions that integrate individual and system-level approaches to behavior change that extend beyond brute-force willpower
Demagogy as a revolutionary phenomenon:the model of the popular tribune in French revolutionary discourse
No full textArtwork protection against unauthorized neural style transfer and aesthetic color distance metric
No full textNeural style transfer (NST) generates new images by combining the style of one image with the content of another. However, unauthorized NST can exploit artwork, raising concerns about artists’ rights and motivating the development of proactive protection methods. We propose Locally Adaptive Adversarial Color Attack (LAACA), enabling artists to conveniently protect their work from unauthorized NST by pre-processing the artwork image before public release, providing content-independent protection regardless of which content image it may later be combined with. LAACA introduces adaptive perturbations that significantly degrade NST quality while maintaining the visual integrity of the original image. We also develope LAACAv2, which resists the current SOTA adversarial perturbation removal method — SDEdit-based adversarial purification. Additionally, we introduce the Aesthetic Color Distance Metric (ACDM) to better evaluate color-sensitive tasks like NST. Extensive experiments across various NST techniques demonstrate our methods outperform baselines in structural similarity, color preservation, and perceptual quality. User studies with both general users and art experts confirm the practical applicability of our approach, addressing the social trust crisis in the art community while advancing adversarial machine learning at the intersection of art, technology, and intellectual property rights.</p
Kant and the Supposed Right of Necessity
No full textA study of Kant's rejection of the 'right' of necessity to kill innocent people in emergency situations (as proposed by the natural law tradition) and his curious argument that the murderer, though guilty, should yet go unpunished
A multinuclear approach to characterising disordered (Al,Sc)-MIL-53 using solid-state NMR spectroscopy
No full textA series of new mixed-metal (Al,Sc)-MIL-53 MOFs has been synthesised solvothermally using DMF. The materials have been characterised using a combination of solid-state NMR spectroscopy, X-ray diffraction, electron microscopy, and energy dispersive X-ray spectroscopy. The breathing behaviour was followed using 13C MAS NMR spectroscopy and showed that the mixed-metal MOFs exhibited different pore forms to the two end members. The cation distribution in the framework was investigated using 17O NMR spectroscopy. The low natural abundance of 17O (0.038 %) requires isotopic enrichment for spectra to be acquired on a reasonable timescale, but the high cost of enriched reagents necessitates a cost-effective and atom-efficient approach. Materials were enriched using post-synthetic steaming and the cation distribution in the enriched frameworks was shown to be close to random. Evidence for framework degradation and the formation of impurities (an alumina-based material and the small-pore MOF Sc2(BDC)3) upon any heating (i.e., during calcination, dehydration or enrichment) was provided from 27Al and 45Sc NMR spectroscopy, leading to the enriched mixed-metal frameworks having very similar compositions (with 70–77 % Al present). Further work would be required to overcome some of the synthetic challenges encountered, but the possibility of tailoring the breathing response of a mixed-metal MOF in the presence of a guest molecule or with changes in external conditions offers interesting possibilities for the future
Access Repository Dashboard
Do you manage University of St. Andrews - Pure? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!