69238 research outputs found
Sort by
Decomposable Image Diffusion for Accurate Semantic Translation
No full textRecent advances in text-to-image diffusion models have achieved remarkable success in generating
high-quality and realistic images from given prompt. However, these methods face
challenges in handling complex prompts and ensuring faithful alignment between the generated
image and the textual described in the prompts. In this paper, we propose a two-stage
image-text realignment method, called Decomposable Image Diffusion, to improve the alignment
between prompt and image in text-to-image generation diffusion models. Decomposable
Diffusion contains two stages. Stage 1: semantic decomposition and compositional image
generation, integrates linguistic structures into the diffusion guidance process by decomposing
complex textual inputs into multiple semantic concepts, which are then used to guide the image
generation process. Stage 2: Semantic realignment aims to realign the image with the prompt.
We propose global reward and local reward to evaluate the consistency between the generated
image and the prompt from both global and local perspectives. Then, we compute the global
realignment loss and local realignment loss, and fine-tune the pretrained diffusion model from
the first stage to achieve the realignment between the image and the prompt. Experimental
results from MS-COCO, ABC-6k, and CC-500 benchmarks show that the proposed Decomposable
Diffusion significantly outperforms other baseline realignment techniques in terms of
visual quality and semantic similarity with the input prompt
Scaling Craft
No full textThis thesis is motivated by a simple observation. We, human beings, are self-aware creatures with distinctive personalities and varying opinions, yet the buildings we occupy are often subject to standardisation and mass production. This growing mundanity, repetitiveness, and overall deterioration of the built environment has weakened the relationships between individuals and the spaces they inhabit.
The matter of well-crafted architecture is now in question as this growing trend of technology and automation has reduced authenticity and character in making processes. In order to gain a better understanding of the position of craft in the modern day, this project asks: what does craft mean in today’s context, and how can it negotiate between the convergence of hand, machine, and digital fabrication in the field of architecture?
Scaling Craft does not seek a formal conclusion, nor does it result in a conventional architectural position. Instead, it employs methodologies and tools in both practical and theoretical exploration, experimental in nature, in its critique of craft. The project is unique in its construction and approach as it aspires to combine contrasting interpretations of the term in an attempt to understand what crafting truly means in modern architectural practice, bringing attributes of skill, material sensitivity, innovation and engagement into discussion.
In an iterative journey of making and testing across various scales, this thesis proposes a new way of thinking about architecture and what it means to craft
An Investigation into Vaping: Modelling the Effect of E-Cigarette Induced Lung Inflammation on Organ-Level Function
No full textE-cigarette (EC) use has risen rapidly worldwide, particularly among young adults.
Aotearoa New Zealand has been especially affected, ranking among the highest global
vaping populations in 2023. Although often marketed as a safer alternative to cigarettes, EC use has been strongly linked to pulmonary inflammation, which in turn drives oxidative stress and cytotoxicity, contributing to asthma-like symptoms, small airway fibrosis, and acute injuries such as E-cigarette or Vaping Use-Associated Lung Injury (EVALI). The impact of ECs is challenging to understand experimentally, given their enormous variability in flavour, device type, and heating temperature. Computational modelling offers a useful alternative, enabling the understanding of the individual contributors to integrated inflammatory processes. To date, no comprehensive computational framework has been introduced for EC exposure.
This thesis presents the first multi-scale computational framework linking EC exposure to inflammation, tissue remodelling, and whole-lung mechanics. This was accomplished by expanding and integrating two models: an agent-based model of inflammatory airway remodelling and a simulation of lung ventilation.
Simulations across increasing EC exposure revealed that chronic EC use promotes
airway wall thickening, elevates stiffness, and narrows the small airways. It was found
that increased EC exposure correlated with increased deposition of airway smooth muscle and fibrosis alongside impaired epithelial recovery. These structural changes reduced dynamic compliance, increased airway resistance, and increased the work of breathing (WOB). The model identified potential inflammatory thresholds, beyond which airway stiffening accelerated, supporting the hypothesis that increased EC exposure drives progressive remodelling.
These findings suggest that EC use may trigger a distinct trajectory of small air-way disease, increasing ventilation heterogeneity and the risk of long-term respiratory
decline. While further empirical validation is needed, the model identifies qualitative
trends that can guide future research. This framework offers a novel platform for predictive respiratory toxicology, adaptable to other inhaled exposures and suitable for investigating chronic airway disease progression
Investigating the mechanism of action of thioesterbased polymyxin B variants in different membranes using molecular dynamics simulations
No full textAntimicrobial resistance poses a serious threat to global public health, and
polymyxins remain one of the last-choice treatments for drug-resistant Gram-negative
bacterial infections. The emergence of multidrug-resistant bacteria and the spread of
resistance genes such as mcr-1 have further exacerbated this challenge. However, the dosedependent
nephrotoxicity of polymyxins limits their clinical application, creating an urgent
need for low-toxicity and effective polymyxin variants. This study employed all-atom
molecular dynamics simulations to investigate the interactions between Polymyxin B3 and
three thioester-containing polymyxin variants with polymyxin-sensitive and resistant inner
and outer membranes of Eschericia coli. Specifically, we aimed to explore: (1) the
mechanism of action of polymyxins; (2) the pEtN-mediated resistance mechanism; and (3)
the differences in membrane interactions between thioester-containing polymyxin variants
and Polymyxin B3. The results support that polymyxins act through the self-promoted uptake
mechanism, progressing through three key stages: attraction, binding, and insertion. The
positively charged L-2,4-diaminobutyric acid residues form strong electrostatic interactions
with negatively charged phosphate groups on the membrane surface, followed by Mg²⁺
displacement, weakening of the hydrogen bond network, and ultimately, insertion of the
hydrophobic tail into the membrane core, leading to membrane destabilization. pEtN
modification reduces the net negative charge of lipid A, weakens the electrostatic attraction
between polymyxins and the membrane, enhances lipid A crosslinking, increases membrane
rigidity, and inhibits polymyxin binding and insertion. The thioester-containing polymyxin
variants exhibited distinct membrane interaction properties. The CAD variant demonstrated
the strongest binding affinity and membrane interactions, while the DMBA and BAS variants,
due to their shorter hydrophobic tails, had reduced binding affinity and insertion capability.
Overall, this study provides detailed insights into the mechanism of action and resistance of
polymyxins, offering theoretical guidance for the design of novel polymyxin variants with
improved efficacy and reduced toxicity, and contributing to antimicrobial resistance research
A non-native pathogen meets a native host: Austropuccinia psidii infection reduces photosynthesis and alters non-structural carbohydrates in seedlings of Metrosideros excelsa
No full textAbstract
Key message
Austropuccinia psidii infection and increase in diseased leaf area resulted in a reduction of photosynthesis, an upregulation of stomatal conductance, and an increase in leaf starch and sucrose content.
Abstract
Austropuccinia psidii is a biotrophic rust pathogen that causes myrtle rust, affecting over 480 species in the Myrtaceae family. The development of chlorotic and necrotic leaf areas following A. psidii infection has been shown to affect leaf gas exchange. In this study, we quantified photosynthesis, stomatal conductance, and non-structural carbohydrates in seedlings of a long-lived tree, Metrosideros excelsa (pōhutukawa), following A. psidii infection in a glasshouse experiment (infected and control seedlings) conducted over 20 weeks. The diseased leaf area rose from 8% in week 2 to 95% in week 20 after A. psidii inoculation. The photosynthetic rate declined by over 90% within 6 weeks after inoculation and was associated with biochemical damage in CO2 fixation. Stomatal conductance decreased over the first 4 weeks after inoculation and then increased. An increase in lesions and necrotic cells may inhibit stomatal regulation. Starch content was threefold higher in infected than control leaves 20 weeks after inoculation. Increased starch accumulation in the infected leaf area could be due to reduced export of newly fixed carbon from the infected leaves. Meanwhile, glucose + fructose content was 31% lower in infected leaves at the experiment’s end, likely because of leaf necrosis. If the pathogen-induced damage and loss of leaves, reduction in photosynthesis and changes in non-structural carbohydrates shown in this study also occur in wild M. excelsa seedlings and reduces their biomass, this may in turn reduce their competitive ability in the primary successions that they currently often dominate.
</jats:sec
Urban Safety Via Digitally Augmented Relationality
No full textThis article addresses urban safety issues, focusing on their relationship with the ongoing effective structural transformation of inborn instrumentalities of progressively vulnerable communities in the Global South. It responds to the United Nations' call for urgent action to reverse the negative trends towards the goals set with the 2030 Agenda for Sustainable Development by redefining urban safety as a crucial factor for the affirmation of sociospatial justice, resilience, and wellbeing in alignment with the objectives of the Global Digital Compact framework. We redress the conventional crime prevention approach by formulating a new framework for defining a differential right to digitally augmented urban safety that leverages locally embedded sociocultural practices and relational networks. We conjuncturally articulate this right for underserved communities within informal settlements in Bandung, Java, Republic of Indonesia, concentrating on the Indigenous gotong-royong cultural practice—a system of commoning combining differential forms of collaboration, relational empathy and imaginative re-enchantment. We develop and operationalise this right, to produce a pilot methodology for creating a mixed reality instrument that converts inadequate securitisation methods of surveillance into effective safety practices of relatedness. By combining digital twinning and gaming technologies, this instrument enhances community-centred care and accumulation of social, cultural, and infrastructural capital. We discuss the development of this instrument, focusing on the conception and implementation phases that involved multilevel, multistakeholder engagement, emphasising its supplementary transformative aspects. Our contribution to the urban safety discourse further articulates the Right to the Safety by introducing a novel approach that upholds the core good city principles—participation, relationality, and cosmopolitics—for regenerative and differential urban development in the increasingly mediatised society
Utilising the Nature Futures Framework to identify positive pathways for marine invasive seaweeds
No full textThe development of new tools to assist in stemming global biodiversity loss and environmental degradation has been one of the focus areas of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES). A tool, known as the Nature Futures Framework (NFF), aims to develop a desirable future for nature and people through the values ‘Nature for Nature’, ‘Nature for Society’, ‘Nature as Culture/One with Nature’, or a combination of these different perspectives.
In this study the NFF scenario-based approach was used to examine the interactions of the invasive marine macroalgae (Undaria pinnatifida, Asparagopsis taxiformis, Caulerpa brachypus and Caulerpa parvifolia) and their impacts on the NFF values. While invasive species can cause irreversible changes to biodiversity, communities, and people, they may also be able to be used or their damage to native species and ecosystems minimised.
The results of this study indicate that beneficial, harmful, and neutral interactions for all four macroalgal species were able to be identified with nature, society and culture, whilst commercial opportunities from in situ harvesting were identified for U. pinnatifida, and A. taxiformis. A higher number of positive interactions were highlighted for U. pinnatifida, compared to A. taxiformis, C. brachypus and C. parvifolia, which is likely a result of the widespread distribution of U. pinnatifida along New Zealand’s coastline, compared to A. taxiformis, C. brachypus and C. parvifolia that have only recently been detected.
The NFF values are not mutually exclusive, with connectivity, surveillance, and commercial value all having a positive influence. The findings also suggest that stakeholder engagement is an important element for the future, and confirmed the NFF is a useful tool to evaluate the potential pathways for management of invasive macroalgae species in New Zealand’s waters, and a tool to enable stakeholder views to be captured and different perspectives on opportunities to be explored
A Survey of Knowledge Editing: Technologies and Trends
No full textLarge language models (LLMs), storing enormous knowledge, have become one of the most prominent fields. However, deficiencies in training data, model learning errors, and the world change can lead to inaccurate or outdated knowledge within these models, resulting in hallucinations that significantly impair their capabilities or cause errors. Traditional solutions like retraining and fine-tuning are often prohibitively time-consuming, resource-intensive, and yield suboptimal results. Knowledge editing, characterized by its efficiency, accuracy, and resource-light nature, has emerged and rapidly developed as a promising alternative in recent years. While most existing surveys focus on specific techniques, they often fail to comprehensively provide an overview of the field's evolution. This paper pioneers a novel perspective by tracing the development of knowledge editing through the lens of editing complexity. We systematically categorize editing methods into four distinct classes based on their capabilities: whether they are qualified for batch and sequential editing. Progressing from foundational to advanced concepts, we also introduce the key metrics for evaluating knowledge editing techniques. This structured approach offers readers a clear grasp of the entire field, illuminates current research trends through the shift in focus over time, and highlights promising future directions
Potential Effects of an Emerging Disease on Tree Communities Around a Susceptible Foundation Tree Species
No full textABSTRACTDisease caused by plant pathogens can shape plant community structure and composition, especially if the host species plays important ecological roles. The New Zealand kauri, Agathis australis, is a foundation species that modifies its soil environment and, by doing so, it acts as an ecological filter selecting for a distinctive suite of plant species that co‐occur preferentially with it. Kauri is threatened by kauri dieback disease caused by Phytophthora agathidicida. Despite kauri's ecological importance in structuring plant communities, little is known about the consequences of kauri dieback on the plant species ecologically associated with kauri—kauri associates. This study examined (1) the environmental drivers of diseased kauri occurrences, (2) the correlations between the occurrences of kauri associates and environmental factors and (3) which kauri associates frequently co‐occurred with diseased kauri as a result of shared environmental responses. A joint species distribution model with nine environmental factors was used to analyse the occurrences of diseased kauri and 16 kauri‐associated plant species located within a 10‐m radius of 761 kauri trees in lowland kauri‐podocarp‐broadleaved forests of Waitākere Ranges, Auckland, New Zealand. Our results showed that diseased kauri occurred more frequently in sites at lower elevation, with shallower organic soil layers and a history of logging or timber processing disturbance. Additionally, successional strategies influenced the relationship between kauri associate occurrences and environmental factors, with kauri associates characteristic of early succession frequently co‐occurring with diseased kauri due to environmental factors. This suggests that kauri dieback could either cause the retrogression of succession in kauri forests or that kauri in younger stands were more likely to be diseased. Both of these possibilities indicate that kauri dieback would likely change the successional trajectory of kauri forests. Future studies should integrate longitudinal demographic data with disease progression to investigate the mechanistic effects of disease on kauri community dynamics.</jats:p
Resisting the new conservatism: Social work and the social justice imperative
No full textEditorial for 37(3). The continuing downgrading of social work’s social justice agenda alongside increasing professionalisation (narrowly defined) has led to debates about the future of social work in its current form