Search

Recent clinical xenotransplantation and human decedent studies demonstrate that clinical hyperacute rejection of genetically engineered porcine organs can be reliably avoided but that antibody mediated rejection continues to limit graft survival. We previously identified porcine glycans and proteins which are immunogenic after cardiac xenotransplantation in nonhuman primates, but the clinical immune response to antigens present in glycan depleted triple knockout (TKO) donor pigs is poorly understood. In this study we use fluorescence barcoded HEK cells and HEK cell lines expressing porcine glycans (Gal and SDa) or proteins (CD9, CD46, CD59, PROCR and ANXA2) to screen antibody reactivity in human serum from 160 swine veterinarians, a serum source with potential occupational immune challenge from porcine tissues and pathogens. High levels of anti-Gal IgM were present in all samples and lower levels of anti-SDa IgM were present in 41% of samples. IgM binding to porcine proteins, primarily CD9 and CD46, previously identified as immunogenic in pig to non-human primate cardiac xenograft recipients, was detected in 28 of the 160 swine veterinarian samples. These results suggest that barcoded HEK cell lines expressing porcine protein antigens can be useful for screening human patient serum. A comprehensive analysis of sera from clinical xenotransplant recipients to define a panel of commonly immunogenic porcine antigens will likely be necessary to establish an array of porcine non-Gal antigens for effective monitoring of patient immune responses and allow earlier therapies to reverse antibody mediated rejection., Comment: 13 pages, 2 Tables, 3 Figures

Recent research has shown that large language models (LLMs) can achieve remarkable translation performance through supervised fine-tuning (SFT) using only a small amount of parallel data. However, SFT simply instructs the model to imitate the reference translations at the token level, making it vulnerable to the noise present in the references. Hence, the assistance from SFT often reaches a plateau once the LLMs have achieved a certain level of translation capability, and further increasing the size of parallel data does not provide additional benefits. To overcome this plateau associated with imitation-based SFT, we propose a preference-based approach built upon the Plackett-Luce model. The objective is to steer LLMs towards a more nuanced understanding of translation preferences from a holistic view, while also being more resilient in the absence of gold translations. We further build a dataset named MAPLE to verify the effectiveness of our approach, which includes multiple translations of varying quality for each source sentence. Extensive experiments demonstrate the superiority of our approach in "breaking the plateau" across diverse LLMs and test settings. Our in-depth analysis underscores the pivotal role of diverse translations and accurate preference scores in the success of our approach., Comment: Accepted to NAACL 2024 (long, main)

Macrophages in atherosclerotic lesions exhibit a spectrum of behaviours or phenotypes. The phenotypic distribution of monocyte-derived macrophages (MDMs), its correlation with MDM lipid content, and relation to blood lipoprotein densities are not well understood. Of particular interest is the balance between low density lipoproteins (LDL) and high density lipoproteins (HDL), which carry bad and good cholesterol respectively. To address these issues, we have developed a mathematical model for early atherosclerosis in which the MDM population is structured by phenotype and lipid content. The model admits a simpler, closed subsystem whose analysis shows how lesion composition becomes more pathological as the blood density of LDL increases relative to the HDL capacity. We use asymptotic analysis to derive a power-law relationship between MDM phenotype and lipid content at steady-state. This relationship enables us to understand why, for example, lipid-laden MDMs have a more inflammatory phenotype than lipid-poor MDMs when blood LDL lipid density greatly exceeds HDL capacity. We show further that the MDM phenotype distribution always attains a local maximum, while the lipid content distribution may be unimodal, adopt a quasi-uniform profile or decrease monotonically. Pathological lesions exhibit a local maximum in both the phenotype and lipid content MDM distributions, with the maximum at an inflammatory phenotype and near the lipid content capacity respectively. These results illustrate how macrophage heterogeneity arises in early atherosclerosis and provide a framework for future model validation through comparison with single-cell RNA sequencing data.

The Directed Acyclic Transformer is a fast non-autoregressive (NAR) model that performs well in Neural Machine Translation. Two issues prevent its application to general Natural Language Generation (NLG) tasks: frequent Out-Of-Vocabulary (OOV) errors and the inability to faithfully generate entity names. We introduce Control-DAG, a constrained decoding algorithm for our Directed Acyclic T5 (DA-T5) model which offers lexical, vocabulary and length control. We show that Control-DAG significantly enhances DA-T5 on the Schema Guided Dialogue and the DART datasets, establishing strong NAR results for Task-Oriented Dialogue and Data-to-Text NLG., Comment: 11 pages. NAACL 2024

Sidewalk robots are increasingly common across the globe. Yet, their operation on public paths poses challenges for people with mobility disabilities (PwMD) who face barriers to accessibility, such as insufficient curb cuts. We interviewed 15 PwMD to understand how they perceive sidewalk robots. Findings indicated that PwMD feel they have to compete for space on the sidewalk when robots are introduced. We next interviewed eight robotics practitioners to learn about their attitudes towards accessibility. Practitioners described how issues often stem from robotic companies addressing accessibility only after problems arise. Both interview groups underscored the importance of integrating accessibility from the outset. Building on this finding, we held four co-design workshops with PwMD and practitioners in pairs. These convenings brought to bear accessibility needs around robots operating in public spaces and in the public interest. Our study aims to set the stage for a more inclusive future around public service robots.

Heavy metal subdwarfs are a class of hot subdwarfs with very high abundances of heavy elements, typically around 10 000 times solar. They include stars which are strongly enhanced in either lead or zirconium, as well as other elements. Vertical stratification of the enhanced elements, where the element is concentrated in a thin layer of the atmosphere, has been proposed as a mechanism to explain the apparent high abundances. This paper explores the effects of the vertical stratification of lead on theoretical spectra of hot subdwarfs. The concentration of lead in different regions of the model atmosphere is found to affect individual lines in a broadly wavelength-dependent manner, with the potential for lines to display modified profiles depending on the location of lead enhancement in the atmosphere. This wavelength dependence highlights the importance of observations in both the optical and the UV for determining whether stratification is present in real stars., Comment: 11 pages, 11 figures, accepted for publication in MNRAS

Of the hundreds of $z\gtrsim6$ quasars discovered to date, only one is known to be gravitationally lensed, despite the high lensing optical depth expected at $z\gtrsim6$. High-redshift quasars are typically identified in large-scale surveys by applying strict photometric selection criteria, in particular by imposing non-detections in bands blueward of the Lyman-$\alpha$ line. Such procedures by design prohibit the discovery of lensed quasars, as the lensing foreground galaxy would contaminate the photometry of the quasar. We present a novel quasar selection methodology, applying contrastive learning (an unsupervised machine learning technique) to Dark Energy Survey imaging data. We describe the use of this technique to train a neural network which isolates an 'island' of 11 sources, of which 7 are known $z\sim6$ quasars. Of the remaining four, three are newly discovered quasars (J0109-5424, $z=6.07$; J0122-4609, $z=5.99$; J0603-3923, $z=5.94$), as confirmed by follow-up Gemini-South/GMOS and archival NTT/EFOSC2 spectroscopy, implying a 91 per cent efficiency for our novel selection method; the final object on the island is a brown dwarf. In one case (J0109-5424), emission below the Lyman limit unambiguously indicates the presence of a foreground source, though high-resolution optical/near-infrared imaging is still needed to confirm the quasar's lensed (multiply-imaged) nature. Detection in the g band has led this quasar to escape selection by traditional colour cuts. Our findings demonstrate that machine learning techniques can thus play a key role in unveiling populations of quasars missed by traditional methods., Comment: 11 pages, 5 figures, 4 tables. Accepted 26 March 2024 for publication in MNRAS

We use both chirped-pulse Fourier transform and frequency modulated absorption spectroscopy to study the rotational spectrum of 2-methoxyethanol in several frequency regions ranging from 8.7-500 GHz. The resulting rotational parameters permitted a search for this molecule in Atacama Large Millimeter/submillimeter Array (ALMA) observations toward the massive protocluster NGC 6334I as well as source B of the low-mass protostellar system IRAS 16293-2422. 25 rotational transitions are observed in the ALMA Band 4 data toward NGC 6334I, resulting in the first interstellar detection of 2-methoxyethanol. A column density of $1.3_{-0.9}^{+1.4} \times 10^{17}$ cm$^{-2}$ is derived at an excitation temperature of $143_{-39}^{+31}$ K. However, molecular signal is not observed in the Band 7 data toward IRAS 16293-2422B and an upper limit column density of $2.5 \times 10^{15}$ cm$^{-2}$ is determined. Various possible formation pathways--including radical recombination and insertion reactions--are discussed. We also investigate physical differences between the two interstellar sources that could result in the observed abundance variations., Comment: Accepted for publication in The Astrophysical Journal Letters

Two approaches have emerged to input images into large language models (LLMs). The first is to caption images into natural language. The second is to map image feature embeddings into the domain of the LLM and pass the mapped embeddings directly to the LLM. The majority of recent few-shot multimodal work reports performance using architectures that employ variations of one of these two approaches. But they overlook an important comparison between them. We design a controlled and focused experiment to compare these two approaches to few-shot visual question answering (VQA) with LLMs. Our findings indicate that for Flan-T5 XL, a 3B parameter LLM, connecting visual embeddings directly to the LLM embedding space does not guarantee improved performance over using image captions. In the zero-shot regime, we find using textual image captions is better. In the few-shot regimes, how the in-context examples are selected determines which is better.

While Reinforcement Learning from Human Feedback (RLHF) effectively aligns pretrained Large Language and Vision-Language Models (LLMs, and VLMs) with human preferences, its computational cost and complexity hamper its wider adoption. To alleviate some of the computational burden of fine-tuning, parameter efficient methods, like LoRA were introduced. In this work, we empirically evaluate the setup of Parameter Efficient Reinforcement Learning from Human Feedback (PE-RLHF) that leverages LoRA fine-tuning for Reward Modeling, and Reinforcement Learning. We benchmark the PE-RLHF setup on six diverse datasets spanning summarization, harmless/helpful response generation, UI automation, and visual question answering in terms of effectiveness of the trained models, and the training resources required. Our findings show, for the first time, that PE-RLHF achieves comparable performance to RLHF, while significantly reducing training time (up to 90% faster for reward models, and 30% faster for RL), and memory footprint (up to 50% reduction for reward models, and 27% for RL). We provide comprehensive ablations across LoRA ranks, and model sizes for both reward modeling and reinforcement learning. By mitigating the computational burden associated with RLHF, we push for a broader adoption of PE-RLHF as an alignment technique for LLMs and VLMs.

A major focus of the planetary science and astrobiology community is the understanding of planetary habitability, including the myriad factors that control the evolution and sustainability of temperate surface environments such as that of Earth. The few substantial terrestrial planetary atmospheres within the Solar System serve as a critical resource in studying these habitability factors, from which models can be constructed for application to extrasolar planets. The recent Astronomy and Astrophysics and Planetary Science and Astrobiology Decadal Surveys both emphasise the need for an improved understanding of planetary habitability as an essential goal within the context of astrobiology. The divergence in climate evolution of Venus and Earth provides a major, accessible basis for understanding how the habitability of large rocky worlds evolves with time and what conditions limit the boundaries of habitability. Here, we argue that Venus can be considered an "anchor point" for understanding planetary habitability within the context of terrestrial planet evolution. We discuss the major factors that have influenced the respective evolutionary pathways of Venus and Earth, how these factors might be weighted in their overall influence, and the measurements that will shed further light on their impacts of these worlds' histories. We further discuss the importance of Venus with respect to both of the recent decadal surveys, and how these community consensus reports can help shape the exploration of Venus in the coming decades., Comment: 23 pages, 3 figures, 1 table, accepted for publication in Nature Astronomy

In this study, we explore the effectiveness of persuasive messages endorsing the adoption of a privacy protection technology (IoT Inspector) tailored to individuals' regulatory focus (promotion or prevention). We explore if and how regulatory fit (i.e., tuning the goal-pursuit mechanism to individuals' internal regulatory focus) can increase persuasion and adoption. We conducted a between-subject experiment (N = 236) presenting participants with the IoT Inspector in gain ("Privacy Enhancing Technology" -- PET) or loss ("Privacy Preserving Technology" -- PPT) framing. Results show that the effect of regulatory fit on adoption is mediated by trust and privacy calculus processes: prevention-focused users who read the PPT message trust the tool more. Furthermore, privacy calculus favors using the tool when promotion-focused individuals read the PET message. We discuss the contribution of understanding the cognitive mechanisms behind regulatory fit in privacy decision-making to support privacy protection.

This paper provides an introduction to quantum machine learning, exploring the potential benefits of using quantum computing principles and algorithms that may improve upon classical machine learning approaches. Quantum computing utilizes particles governed by quantum mechanics for computational purposes, leveraging properties like superposition and entanglement for information representation and manipulation. Quantum machine learning applies these principles to enhance classical machine learning models, potentially reducing network size and training time on quantum hardware. The paper covers basic quantum mechanics principles, including superposition, phase space, and entanglement, and introduces the concept of quantum gates that exploit these properties. It also reviews classical deep learning concepts, such as artificial neural networks, gradient descent, and backpropagation, before delving into trainable quantum circuits as neural networks. An example problem demonstrates the potential advantages of quantum neural networks, and the appendices provide detailed derivations. The paper aims to help researchers new to quantum mechanics and machine learning develop their expertise more efficiently., Comment: Published as a DTIC report under the title "Quantum Computing for Machine Learning - An Introduction". Distribution Statement A: Approved for Public Release. Distribution is Unlimited

Large Multimodal Models (LMMs) excel in natural language and visual understanding but are challenged by exacting tasks such as Knowledge-based Visual Question Answering (KB-VQA) which involve the retrieval of relevant information from document collections to use in shaping answers to questions. We present an extensive training and evaluation framework, M2KR, for KB-VQA. M2KR contains a collection of vision and language tasks which we have incorporated into a single suite of benchmark tasks for training and evaluating general-purpose multi-modal retrievers. We use M2KR to develop PreFLMR, a pre-trained version of the recently developed Fine-grained Late-interaction Multi-modal Retriever (FLMR) approach to KB-VQA, and we report new state-of-the-art results across a range of tasks. We also present investigations into the scaling behaviors of PreFLMR intended to be useful in future developments in general-purpose multi-modal retrievers., Comment: ACL 2024; Project page: https://preflmr.github.io/

Leveraging planning during learning and decision-making is central to the long-term development of intelligent agents. Recent works have successfully combined tree-based search methods and self-play learning mechanisms to this end. However, these methods typically face scaling challenges due to the sequential nature of their search. While practical engineering solutions can partly overcome this, they often result in a negative impact on performance. In this paper, we introduce SPO: Sequential Monte Carlo Policy Optimisation, a model-based reinforcement learning algorithm grounded within the Expectation Maximisation (EM) framework. We show that SPO provides robust policy improvement and efficient scaling properties. The sample-based search makes it directly applicable to both discrete and continuous action spaces without modifications. We demonstrate statistically significant improvements in performance relative to model-free and model-based baselines across both continuous and discrete environments. Furthermore, the parallel nature of SPO's search enables effective utilisation of hardware accelerators, yielding favourable scaling laws., Comment: Accepted to NeurIPS 2024. 34 pages, 3 main figures

We investigate the connection between the most luminous active galactic nuclei (AGN), galaxy pairs, and post-mergers in the IllustrisTNG simulation. We select galaxy pairs and post-mergers with a mass ratio between 1:10 $< \mu <$ 1:1 and a redshift between $010^{44}$ erg/s occur in $\sim $26\% of paired galaxies with a companion within 20 kpc, compared with $\sim $14\% in controls (a fractional excess of $\sim$2). The enhancement of AGN in galaxy pairs is luminosity dependent, with the highest excess (a factor of $\sim6\pm2$ at the closest separations) for AGN with $L_{\mathrm{bol}}>10^{45}$ erg/s. Additionally, pairs exhibit a modest yet statistically significant excess of luminous AGN up to $\sim$150 kpc in separation. For pairs which merge between $010^{45}$ erg/s, $\sim$55\% are interacting, compared with a 30\% interaction fraction in AGN with $10^{44}

Differential equation models are crucial to scientific processes. The values of model parameters are important for analyzing the behaviour of solutions. A parameter is called globally identifiable if its value can be uniquely determined from the input and output functions. To determine if a parameter estimation problem is well-posed for a given model, one must check if the model parameters are globally identifiable. This problem has been intensively studied for ordinary differential equation models, with theory and several efficient algorithms and software packages developed. A comprehensive theory of algebraic identifiability for PDEs has hitherto not been developed due to the complexity of initial and boundary conditions. Here, we provide theory and algorithms, based on differential algebra, for testing identifiability of polynomial PDE models. We showcase this approach on PDE models arising in the sciences.