Search

We present the results of the Distant Giants Survey, a three-year radial velocity (RV) campaign to search for wide-separation giant planets orbiting Sun-like stars known to host an inner transiting planet. We defined a distant giant to have $a$ = 1--10 AU and $M_{p} \sin i = 70-4000$ \mearth~ = 0.2-12.5 \mj, and required transiting planets to have $a<1$ AU and $R_{p} = 1-4$ \rearth. We assembled our sample of 47 stars using a single selection function, and observed each star at monthly intervals to obtain $\approx$30 RV observations per target. The final catalog includes a total of twelve distant companions: four giant planets detected during our survey, two previously known giant planets, and six objects of uncertain disposition identified through RV/astrometric accelerations. Statistically, half of the uncertain objects are planets and the remainder are stars/brown dwarfs. We calculated target-by-target completeness maps to account for missed planets. We found evidence for a moderate enhancement of distant giants (DG) in the presence of close-in small planets (CS), P(DG|CS) = $30^{+14}_{-12}\%$, over the field rate of P(DG) = $16^{+2}_{-2}\%$. No enhancement is disfavored ($p \sim$ 8%). In contrast to a previous study, we found no evidence that stellar metallicity enhances P(DG|CS). We found evidence that distant giant companions are preferentially found in systems with multiple transiting planets and have lower eccentricities than randomly selected giant planets. This points toward dynamically cool formation pathways for the giants that do not disturb the inner systems., Comment: 32 pages, 20 figures, 4 tables. Comments welcome

The Ni-Au-Ga interdiffusion mechanisms taking place during rapid thermal annealing (RTA) under oxygen atmosphere of a Ni-Au/p-GaN contact are investigated by high-resolution transmission electron microscopy (HR-TEM) coupled to energy dispersive X-ray spectroscopy (EDX). It is shown that oxygen-assisted, Ni diffusion to the top surface of the metallic contact through the formation of a nickel oxide (NiOx) is accompanied by Au diffusion down to the GaN surface, and by Ga out-diffusion through the GaN/metal interface. Electrical characterizations of the contact by Transmission Line Method (TLM) show that an ohmic contact is obtained as soon as a thin, Au-Ga interfacial layer is formed, even after complete diffusion of Ni or NiOx to the top surface of the contact. Our results clarify that the presence of Ni or NiOx at the interface is not the main origin of the ohmic-like behavior in such contacts. Auto-cleaning of the interface during the interdiffusion process may play a role, but TEM-EDX analysis evidences that the creation of Ga vacancies associated to the formation of a Ga-Au interfacial layer is crucial for reducing the Schottky barrier height, and maximizing the amount of current flowing through the contact.

Foresight 2 (FS2) is a large language model fine-tuned on hospital data for modelling patient timelines (GitHub 'removed for anon'). It can understand patients' clinical notes and predict SNOMED codes for a wide range of biomedical use cases, including diagnosis suggestions, risk forecasting, and procedure and medication recommendations. FS2 is trained on the free text portion of the MIMIC-III dataset, firstly through extracting biomedical concepts and then creating contextualised patient timelines, upon which the model is then fine-tuned. The results show significant improvement over the previous state-of-the-art for the next new biomedical concept prediction (P/R - 0.73/0.66 vs 0.52/0.32) and a similar improvement specifically for the next new disorder prediction (P/R - 0.69/0.62 vs 0.46/0.25). Finally, on the task of risk forecast, we compare our model to GPT-4-turbo (and a range of open-source biomedical LLMs) and show that FS2 performs significantly better on such tasks (P@5 - 0.90 vs 0.65). This highlights the need to incorporate hospital data into LLMs and shows that small models outperform much larger ones when fine-tuned on high-quality, specialised data.

Low-mass X-ray binaries with a neutron star as the primary object show a complex array of phenomenology during outbursts. The observed variability in X-ray emission primarily arises from changes in the innermost regions of the accretion disk, neutron star surface, and corona. In this work, we present the results of a comprehensive X-ray spectral and timing analysis of the neutron star transient MAXI J1807+132 during its 2023 outburst using data from the NICER observatory. The outburst is marked by a very rapid rise in the count rate by about a factor of 20 in a day. The source undergoes full state transitions and displays hysteresis effect in the hardness and rms intensity diagrams. Spectral analysis with a three-component model is consistent with disk truncation during the hard states and reaching the last stable orbit during the intermediate and soft states. We discuss the different values of the last stable radius in the context of possible distance of the source and magnetic field strength. The characteristic frequencies throughout the hard and intermediate states are found to be strongly correlated with the inner radius of the disk. Together with the spectral and fast variability properties, we attempt to trace the evolution of the size of the corona along the outburst. Following the main outburst, the source undergoes a high amplitude reflare wherein it shows a complex behavior with relatively high variability (10 %), but low hardness., Comment: ApJ (in press)

In 1975, Stein made a wide generalisation of the Ryser-Brualdi-Stein conjecture on transversals in Latin squares, conjecturing that every equi-$n$-square (an $n\times n$ array filled with $n$ symbols where each symbol appears exactly $n$ times) has a transversal of size $n-1$. That is, it should have a collection of $n-1$ entries that share no row, column, or symbol. In 2017, Aharoni, Berger, Kotlar, and Ziv showed that equi-$n$-squares always have a transversal with size at least $2n/3$. In 2019, Pokrovskiy and Sudakov disproved Stein's conjecture by constructing equi-$n$-squares without a transversal of size $n-\frac{\log n}{42}$, but asked whether Stein's conjecture is approximately true. I.e., does an equi-$n$-square always have a transversal with size $(1-o(1))n$? We answer this question in the positive. More specifically, we improve both known bounds, showing that there exist equi-$n$-squares with no transversal of size $n-\Omega(\sqrt{n})$ and that every equi-$n$-square contains $n-n^{1-\Omega(1)}$ disjoint transversals of size $n-n^{1-\Omega(1)}$., Comment: 16 pages, 2 figures

We report on the fault-tolerant operation of logical qubits on a neutral atom quantum computer, with logical performance surpassing physical performance for multiple circuits including Bell states (12x error reduction), random circuits (15x), and a prototype Anderson Impurity Model ground state solver for materials science applications (up to 6x, non-fault-tolerantly). The logical qubits are implemented via the [[4, 2, 2]] code (C4). Our work constitutes the first complete realization of the benchmarking protocol proposed by Gottesman 2016 [1] demonstrating results consistent with fault-tolerance. In light of recent advances on applying concatenated C4/C6 detection codes to achieve error correction with high code rates and thresholds, our work can be regarded as a building block towards a practical scheme for fault tolerant quantum computation. Our demonstration of a materials science application with logical qubits particularly demonstrates the immediate value of these techniques on current experiments.

A recent report on the detection of very-high-energy gamma rays from V4641 Sagittarii (V4641 Sgr) up to ~0.8 peta-electronvolt has made it the second confirmed "PeVatron" microquasar. Here we report on the observation of V4641 Sgr with X-Ray Imaging and Spectroscopy Mission (XRISM) in September 2024. Thanks to the large field of view and low background, the CCD imager Xtend successfully detected for the first time X-ray extended emission around V4641 Sgr with a significance of > 4.5 sigma and > 10 sigma based on our imaging and spectral analysis, respectively. The spatial extent is estimated to have a radius of $7 \pm 3$ arcmin ($13 \pm 5$ pc at a distance of 6.2 kpc) assuming a Gaussian-like radial distribution, which suggests that the particle acceleration site is within ~10 pc of the microquasar. If the X-ray morphology traces the diffusion of accelerated electrons, this spatial extent can be explained by either an enhanced magnetic field (~80 uG) or a suppressed diffusion coefficient (~$10^{27}$ cm$^2$ s$^{-1}$ at 100 TeV). The integrated X-ray flux, (4-6)$\times 10^{-12}$ erg s$^{-1}$ cm$^{-2}$ (2-10 keV), would require a magnetic field strength higher than the galactic mean (> 8 uG) if the diffuse X-ray emission originates from synchrotron radiation and the gamma-ray emission is predominantly hadronic. If the X-rays are of thermal origin, the measured extension, temperature, and plasma density can be explained by a jet with a luminosity of ~$2\times 10^{39}$ erg s$^{-1}$, which is comparable to the Eddington luminosity of this system., Comment: 9 pages, 5 figures, accepted for publication in ApJL

We demonstrate that including the second viscosity of an electron gas in the hydrodynamic model allows for highly accurate modeling of the optical response of heavily doped semiconductors. In our setup, which improves resonance visibility compared to previous approaches, plasmon resonances become more distinct, allowing for detailed analysis of the underlying physics. With advanced fitting techniques based on a physics-informed cost function and a tailored optimization algorithm, we obtain close agreement between simulations and experimental data across different sample thicknesses. This enhanced resonance visibility, combined with our integrated approach, shows that key parameters such as doping level and effective electron mass can be retrieved from a single optical measurement. The spatial dispersion taken into account in the hydrodynamic framework is essential for accurately describing the optical response of plasmonic materials in this frequency range and is likely to become a standard modeling approach.

In this work, we show that for all $t\geq e$, \[|\zeta(1+it)|\leq 0.6443 \log t. \] The equality is achieved when $t=17.7477$. We also use the Riemann-Siegel formula and numerical computations to show that \[|\zeta(1+it)|\leq\frac{1}{2}\log t+0.6633\hspace{1cm}\text{when}\;t\geq e.\]When $t\geq 100$, the bound $\frac{1}{2}\log t+0.6633$ is better than the bound $0.6443\log t$., Comment: 31 pages, 6 figures

Quantum-information processing and computation with bosonic qubits are corruptible by noise channels. Using interferometers and photon-subtraction gadgets (PSGs) accompanied by linear amplification and attenuation, we establish linear-optical methods to mitigate and suppress bosonic noise channels. We first show that by employing amplifying and attenuating PSGs respectively at the input and output of either a thermal or random-displacement channel, probabilistic error cancellation (PEC) can be carried out to mitigate errors in expectation-value estimation. We also derive optimal physical estimators that are properly constrained to improve the sampling accuracy of PEC. Next, we prove that a purely-dephasing channel is coherently suppressible using a multimode Mach--Zehnder interferometer and conditional vacuum measurements (VMZ). In the limit of infinitely-many ancillas, with nonvanishing success rates, VMZ using either Hadamard or two-design interferometers turns any dephasing channel into a phase-space-rotated linear-attenuation channel that can subsequently be inverted with (rotated) linear amplification without Kerr nonlinearity. Moreover, for weak central-Gaussian dephasing, the suppression fidelity increases monotonically with the number of ancillas and most optimally with Hadamard interferometers. We demonstrate the performance of these linear-optical mitigation and suppression schemes on common noise channels (and their compositions) and popular bosonic codes. While the theoretical formalism pertains to idling noise channels, we also provide numerical evidence supporting mitigation and suppression capabilities with respect to noise from universal gate operations., Comment: 25 pages, 13 figures

Stacking, a potent ensemble learning method, leverages a meta-model to harness the strengths of multiple base models, thereby enhancing prediction accuracy. Traditional stacking techniques typically utilize established learning models, such as logistic regression, as the meta-model. This paper introduces a novel approach that integrates computational geometry techniques, specifically solving the maximum weighted rectangle problem, to develop a new meta-model for binary classification. Our method is evaluated on multiple open datasets, with statistical analysis showing its stability and demonstrating improvements in accuracy compared to current state-of-the-art stacking methods with out-of-fold predictions. This new stacking method also boasts two significant advantages: enhanced interpretability and the elimination of hyperparameter tuning for the meta-model, thus increasing its practicality. These merits make our method highly applicable not only in stacking ensemble learning but also in various real-world applications, such as hospital health evaluation scoring and bank credit scoring systems, offering a fresh evaluation perspective., Comment: 11 pages

Treatment selection in breast cancer is guided by molecular subtypes and clinical characteristics. Recurrence risk assessment plays a crucial role in personalizing treatment. Current methods, including genomic assays, have limited accuracy and clinical utility, leading to suboptimal decisions for many patients. We developed a test for breast cancer patient stratification based on digital pathology and clinical characteristics using novel AI methods. Specifically, we utilized a vision transformer-based pan-cancer foundation model trained with self-supervised learning to extract features from digitized H&E-stained slides. These features were integrated with clinical data to form a multi-modal AI test predicting cancer recurrence and death. The test was developed and evaluated using data from a total of 8,161 breast cancer patients across 15 cohorts originating from seven countries. Of these, 3,502 patients from five cohorts were used exclusively for evaluation, while the remaining patients were used for training. Our test accurately predicted our primary endpoint, disease-free interval, in the five external cohorts (C-index: 0.71 [0.68-0.75], HR: 3.63 [3.02-4.37, p<0.01]). In a direct comparison (N=858), the AI test was more accurate than Oncotype DX, the standard-of-care 21-gene assay, with a C-index of 0.67 [0.61-0.74] versus 0.61 [0.49-0.73], respectively. Additionally, the AI test added independent information to Oncotype DX in a multivariate analysis (HR: 3.11 [1.91-5.09, p<0.01)]). The test demonstrated robust accuracy across all major breast cancer subtypes, including TNBC (C-index: 0.71 [0.62-0.81], HR: 3.81 [2.35-6.17, p=0.02]), where no diagnostic tools are currently recommended by clinical guidelines. These results suggest that our AI test can improve accuracy, extend applicability to a wider range of patients, and enhance access to treatment selection tools.

Recently, prompt learning has emerged as the state-of-the-art (SOTA) for fair text-to-image (T2I) generation. Specifically, this approach leverages readily available reference images to learn inclusive prompts for each target Sensitive Attribute (tSA), allowing for fair image generation. In this work, we first reveal that this prompt learning-based approach results in degraded sample quality. Our analysis shows that the approach's training objective -- which aims to align the embedding differences of learned prompts and reference images -- could be sub-optimal, resulting in distortion of the learned prompts and degraded generated images. To further substantiate this claim, as our major contribution, we deep dive into the denoising subnetwork of the T2I model to track down the effect of these learned prompts by analyzing the cross-attention maps. In our analysis, we propose a novel prompt switching analysis: I2H and H2I. Furthermore, we propose new quantitative characterization of cross-attention maps. Our analysis reveals abnormalities in the early denoising steps, perpetuating improper global structure that results in degradation in the generated samples. Building on insights from our analysis, we propose two ideas: (i) Prompt Queuing and (ii) Attention Amplification to address the quality issue. Extensive experimental results on a wide range of tSAs show that our proposed method outperforms SOTA approach's image generation quality, while achieving competitive fairness. More resources at FairQueue Project site: https://sutd-visual-computing-group.github.io/FairQueue, Comment: Accepted in NeurIPS24

Sparse Mixture of Experts (SMoE) has become the key to unlocking unparalleled scalability in deep learning. SMoE has the potential to exponentially increase parameter count while maintaining the efficiency of the model by only activating a small subset of these parameters for a given sample. However, it has been observed that SMoE suffers from unstable training and has difficulty adapting to new distributions, leading to the model's lack of robustness to data contamination. To overcome these limitations, we first establish a connection between the dynamics of the expert representations in SMoEs and gradient descent on a multi-objective optimization problem. Leveraging our framework, we then integrate momentum into SMoE and propose a new family of SMoEs named MomentumSMoE. We theoretically prove and numerically demonstrate that MomentumSMoE is more stable and robust than SMoE. In particular, we verify the advantages of MomentumSMoE over SMoE on a variety of practical tasks including ImageNet-1K object recognition and WikiText-103 language modeling. We demonstrate the applicability of MomentumSMoE to many types of SMoE models, including those in the Sparse MoE model for vision (V-MoE) and the Generalist Language Model (GLaM). We also show that other advanced momentum-based optimization methods, such as Adam, can be easily incorporated into the MomentumSMoE framework for designing new SMoE models with even better performance, almost negligible additional computation cost, and simple implementations., Comment: 10 pages in the main text. Published at NeurIPS 2024. The code is available at https://github.com/rachtsy/MomentumSMoE

This study investigates the application of Transfer Learning (TL) on Transformer architectures to enhance building energy consumption forecasting. Transformers are a relatively new deep learning architecture, which has served as the foundation for groundbreaking technologies such as ChatGPT. While TL has been studied in the past, prior studies considered either one data-centric TL strategy or used older deep learning models such as Recurrent Neural Networks or Convolutional Neural Networks. Here, we carry out an extensive empirical study on six different data-centric TL strategies and analyse their performance under varying feature spaces. In addition to the vanilla Transformer architecture, we also experiment with Informer and PatchTST, specifically designed for time series forecasting. We use 16 datasets from the Building Data Genome Project 2 to create building energy consumption forecasting models. Experimental results reveal that while TL is generally beneficial, especially when the target domain has no data, careful selection of the exact TL strategy should be made to gain the maximum benefit. This decision largely depends on the feature space properties such as the recorded weather features. We also note that PatchTST outperforms the other two Transformer variants (vanilla Transformer and Informer). Our findings advance the building energy consumption forecasting using advanced approaches like TL and Transformer architectures.

Algebraic Data Types (ADTs) are an increasingly common feature in modern programming languages. In many implementations, values of non-nullary, multi-case ADTs are allocated on the heap, which may reduce performance and increase memory usage. This work explores annotation-guided optimizations to ADT representation in Virgil, a systems-level programming language that compiles to x86, x86-64, Wasm and the Java Virtual Machine. We extend Virgil with annotations: #unboxed to eliminate the overhead of heap allocation via automatic compiler transformation to a scalar representation, and #packed, to enable programmer-expressed bit-layouts. These annotations allow programmers to both save memory and manipulate data in formats dictated by hardware. We dedicate this work as an homage and echo of work done in collaboration with Jens in the work entitled "A Declarative Approach to Generating Machine Code Tools", an unpublished manuscript from 2005. In fact, this work inherits some syntactic conventions from that prior work. The performance impact of these representation changes was evaluated on a variety of workloads in terms of execution time and memory usage, but we don't include it because Jens like semantics and type systems better!

This paper examines how the sequencing of images and text within multi-modal prompts influences the reasoning performance of large language models (LLMs). We performed empirical evaluations using three commercial LLMs. Our results demonstrate that the order in which modalities are presented can significantly affect performance, particularly in tasks of varying complexity. For simpler tasks involving a single image, modality sequencing had a clear impact on accuracy. However, in more complex tasks involving multiple images and intricate reasoning steps, the effect of sequencing diminished, likely due to the increased cognitive demands of the task. Our findings also highlight the importance of question/prompt structure. In nested and multi-step reasoning tasks, modality sequencing played a key role in shaping model performance. While LLMs excelled in the initial stages of reasoning, they struggled to re-incorporate earlier information, underscoring the challenges of multi-hop reasoning within transformer architectures. This suggests that aligning the sequence of modalities with the logical flow of reasoning steps is more critical than modality order alone. These insights offer valuable implications for improving multi-modal prompt design, with broader applications across fields such as education, medical imaging, and cross-modal learning.

We study the topological order that arises from chiral states with ${\rm SU}(N)$ or ${\rm SO}(N)$ edge-state symmetry. This extends our previous study of topological orders that descend from the bosonic $E_8$ quantum Hall state. We use exactly solvable models of coupled electron wires to construct states with ${\rm SU}(m)_n$, ${\rm SO}(m)_n$, or ${\rm Sp}(m)_n$ topological order for various levels $n$. We use our constructions to write down string operators for various non-Abelian anyons. We thereby provide a systematic, microscopic understanding of quantum Hall states, topological superconductors, and spin liquids with emergent non-Abelian quasiparticle excitations, including those of Ising, metaplectic, and Fibonacci type., Comment: 32 pages, 2 figures

In order to solve problems of practical importance, quantum computers will likely need to incorporate quantum error correction, where a logical qubit is redundantly encoded in many noisy physical qubits. The large physical-qubit overhead typically associated with error correction motivates the search for more hardware-efficient approaches. Here, using a microfabricated superconducting quantum circuit, we realize a logical qubit memory formed from the concatenation of encoded bosonic cat qubits with an outer repetition code of distance $d=5$. The bosonic cat qubits are passively protected against bit flips using a stabilizing circuit. Cat-qubit phase-flip errors are corrected by the repetition code which uses ancilla transmons for syndrome measurement. We realize a noise-biased CX gate which ensures bit-flip error suppression is maintained during error correction. We study the performance and scaling of the logical qubit memory, finding that the phase-flip correcting repetition code operates below threshold, with logical phase-flip error decreasing with code distance from $d=3$ to $d=5$. Concurrently, the logical bit-flip error is suppressed with increasing cat-qubit mean photon number. The minimum measured logical error per cycle is on average $1.75(2)\%$ for the distance-3 code sections, and $1.65(3)\%$ for the longer distance-5 code, demonstrating the effectiveness of bit-flip error suppression throughout the error correction cycle. These results, where the intrinsic error suppression of the bosonic encodings allows us to use a hardware-efficient outer error correcting code, indicate that concatenated bosonic codes are a compelling paradigm for reaching fault-tolerant quantum computation., Comment: Comments on the manuscript welcome!

Two-dimensional van der Waals semiconductors are promising for future nanoelectronics. However, integrating high-k gate dielectrics for device applications is challenging as the inert van der Waals material surfaces hinder uniform dielectric growth. Here, we report a liquid metal oxide-assisted approach to integrate ultrathin, high-k HfO2 dielectric on 2D semiconductors with atomically smooth interfaces. Using this approach, we fabricated 2D WS2 top-gated transistors with subthreshold swings down to 74.5 mV/dec, gate leakage current density below 10-6 A/cm2, and negligible hysteresis. We further demonstrate a one-step van der Waals integration of contacts and dielectrics on graphene. This can offer a scalable approach toward integrating entire prefabricated device stack arrays with 2D materials. Our work provides a scalable solution to address the crucial dielectric engineering challenge for 2D semiconductors, paving the way for high-performance 2D electronics.

Recent advances in large language models (LLMs) have gained interest in speech-text multimodal foundation models, achieving strong performance on instruction-based speech translation (ST). However, expanding language pairs from an existing instruction-tuned ST system is costly due to the necessity of re-training on a combination of new and previous datasets. We propose to expand new language pairs by merging the model trained on new language pairs and the existing model, using task arithmetic. We find that the direct application of task arithmetic for ST causes the merged model to fail to follow instructions; thus, generating translation in incorrect languages. To eliminate language confusion, we propose an augmented task arithmetic method that merges an additional language control model. It is trained to generate the correct target language token following the instructions. Our experiments demonstrate that our proposed language control model can achieve language expansion by eliminating language confusion. In our MuST-C and CoVoST-2 experiments, it shows up to 4.66 and 4.92 BLEU scores improvement, respectively. In addition, we demonstrate the use of our task arithmetic framework can expand to a language pair where neither paired ST training data nor a pre-trained ST model is available. We first synthesize the ST system from machine translation (MT) systems via task analogy, then merge the synthesized ST system to the existing ST model.

Third-party Software Development Kits (SDKs) are widely adopted in Android app development, to effortlessly accelerate development pipelines and enhance app functionality. However, this convenience raises substantial concerns about unauthorized access to users' privacy-sensitive information, which could be further abused for illegitimate purposes like user tracking or monetization. Our study offers a targeted analysis of user privacy protection among Android third-party SDKs, filling a critical gap in the Android software supply chain. It focuses on two aspects of their privacy practices, including data exfiltration and behavior-policy compliance (or privacy compliance), utilizing techniques of taint analysis and large language models. It covers 158 widely-used SDKs from two key SDK release platforms, the official one and a large alternative one. From them, we identified 338 instances of privacy data exfiltration. On the privacy compliance, our study reveals that more than 30% of the examined SDKs fail to provide a privacy policy to disclose their data handling practices. Among those that provide privacy policies, 37% of them over-collect user data, and 88% falsely claim access to sensitive data. We revisit the latest versions of the SDKs after 12 months. Our analysis demonstrates a persistent lack of improvement in these concerning trends. Based on our findings, we propose three actionable recommendations to mitigate the privacy leakage risks and enhance privacy protection for Android users. Our research not only serves as an urgent call for industry attention but also provides crucial insights for future regulatory interventions., Comment: 16 pages

Co-speech gestures are fundamental for communication. The advent of recent deep learning techniques has facilitated the creation of lifelike, synchronous co-speech gestures for Embodied Conversational Agents. "In-the-wild" datasets, aggregating video content from platforms like YouTube via human pose detection technologies, provide a feasible solution by offering 2D skeletal sequences aligned with speech. Concurrent developments in lifting models enable the conversion of these 2D sequences into 3D gesture databases. However, it is important to note that the 3D poses estimated from the 2D extracted poses are, in essence, approximations of the ground-truth, which remains in the 2D domain. This distinction raises questions about the impact of gesture representation dimensionality on the quality of generated motions - a topic that, to our knowledge, remains largely unexplored. Our study examines the effect of using either 2D or 3D joint coordinates as training data on the performance of speech-to-gesture deep generative models. We employ a lifting model for converting generated 2D pose sequences into 3D and assess how gestures created directly in 3D stack up against those initially generated in 2D and then converted to 3D. We perform an objective evaluation using widely used metrics in the gesture generation field as well as a user study to qualitatively evaluate the different approaches., Comment: arXiv admin note: substantial text overlap with arXiv:2406.15111

Electronic Health Records are large repositories of valuable clinical data, with a significant portion stored in unstructured text format. This textual data includes clinical events (e.g., disorders, symptoms, findings, medications and procedures) in context that if extracted accurately at scale can unlock valuable downstream applications such as disease prediction. Using an existing Named Entity Recognition and Linking methodology, MedCAT, these identified concepts need to be further classified (contextualised) for their relevance to the patient, and their temporal and negated status for example, to be useful downstream. This study performs a comparative analysis of various natural language models for medical text classification. Extensive experimentation reveals the effectiveness of transformer-based language models, particularly BERT. When combined with class imbalance mitigation techniques, BERT outperforms Bi-LSTM models by up to 28% and the baseline BERT model by up to 16% for recall of the minority classes. The method has been implemented as part of CogStack/MedCAT framework and made available to the community for further research.

This paper deals with a Tikhonov regularized second-order plus first-order primal-dual dynamical system with time scaling for separable convex optimization problems with linear equality constraints. This system consists of two second-order ordinary differential equations for the primal variables and one first-order ordinary differential equation for the dual variable.By utilizing the Lyapunov analysis approach, we obtain the convergence properties of the primal-dual gap, the objective function error, the feasibility measure and the gradient norm of the objective function along the trajectory. We also establish the strong convergence of the primal trajectory generated by the dynamical system towards the minimal norm solution of the separable convex optimization problem. Furthermore, we give numerical experiments to illustrate the theoretical results, showing that our dynamical system performs better than those in the literature in terms of convergence rates.

Modern artificial intelligence (AI) systems are powered by foundation models. This paper presents a new set of foundation models, called Llama 3. It is a herd of language models that natively support multilinguality, coding, reasoning, and tool usage. Our largest model is a dense Transformer with 405B parameters and a context window of up to 128K tokens. This paper presents an extensive empirical evaluation of Llama 3. We find that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks. We publicly release Llama 3, including pre-trained and post-trained versions of the 405B parameter language model and our Llama Guard 3 model for input and output safety. The paper also presents the results of experiments in which we integrate image, video, and speech capabilities into Llama 3 via a compositional approach. We observe this approach performs competitively with the state-of-the-art on image, video, and speech recognition tasks. The resulting models are not yet being broadly released as they are still under development.

Activities such as the movement of passengers and goods, the transfer of physical or digital assets, web navigation and even successive passes in football, result in timestamped paths through a physical or virtual network. The need to analyse such paths has produced a new modelling paradigm in the form of higher-order networks which are able to capture temporal and topological characteristics of sequential data. This has been complemented by sequence mining approaches, a key example being sequential motifs measuring the prevalence of recurrent subsequences. Previous work on higher-order networks has focused on how to identify the optimal order for a path dataset, where the order can be thought of as the number of steps of memory encoded in the model. In this paper, we build on these approaches to consider which orders are necessary to reproduce different path characteristics, from path lengths to counts of sequential motifs, viewing paths generated from different higher-order models as null models which capture features of the data up to a certain order, and randomised otherwise. Furthermore, we provide an important extension to motif counting, whereby cases with self-loops, starting nodes, and ending nodes of paths are taken into consideration. Conducting a thorough analysis using path lengths and sequential motifs on a diverse range of path datasets, we show that our approach can shed light on precisely where models of different order overperform or underperform, and what this may imply about the original path data., Comment: 23 pages, 8 figures, submitted to the Journal of Complex Networks

Research in 3D mapping is crucial for smart city applications, yet the cost of acquiring 3D data often hinders progress. Visual localization, particularly monocular camera position estimation, offers a solution by determining the camera's pose solely through visual cues. However, this task is challenging due to limited data from a single camera. To tackle these challenges, we organized the AISG-SLA Visual Localization Challenge (VLC) at IJCAI 2023 to explore how AI can accurately extract camera pose data from 2D images in 3D space. The challenge attracted over 300 participants worldwide, forming 50+ teams. Winning teams achieved high accuracy in pose estimation using images from a car-mounted camera with low frame rates. The VLC dataset is available for research purposes upon request via vlc-dataset@aisingapore.org.

In 2010, Bakos and collaborators discovered a Neptune-sized planet transiting the K-dwarf HAT-P-11 every five days. Later in 2018, Yee and collaborators reported an additional Jovian-mass companion on a nine year orbit based on a decade of Doppler monitoring. The eccentric outer giant HAT-P-11c may be responsible for the peculiar polar orbit of the inner planet HAT-P-11b. However, Basilicata et al. (2024) recently suggested that the HAT-P-11c Doppler signal could be caused by stellar activity. In this research note, we extend the Yee et al. (2018) Doppler time series by six years. The combined dataset spanning 17 years covers nearly two orbits of the outer planet. Importantly, we observe two periastron passages of planet c and do not observe a coherent activity signature. Together with the previously reported astrometric acceleration of HAT-P-11 from Hipparcos and Gaia, we believe there is strong evidence for HAT-P-11c as a bona fide planet., Comment: 4 pages, 1 figure

We prove that if an orientable 3-manifold $M$ admits a complete Riemannian metric whose scalar curvature is positive and has a subquadratic decay at infinity, then it decomposes as a (possibly infinite) connected sum of spherical manifolds and $\mathbb{S}^2 \times \mathbb{S}^1$ summands. This generalises a theorem of Gromov and Wang by using a different, more topological, approach. As a result, the manifold $M$ carries a complete Riemannian metric of uniformly positive scalar curvature, which partially answers a conjecture of Gromov. More generally, the topological decomposition holds without any scalar curvature assumption under a weaker condition on the filling discs of closed curves in the universal cover based on the notion of fill radius. Moreover, the decay rate of the scalar curvature is optimal in this decomposition theorem. Indeed, the manifold $\mathbb{R}^2 \times \mathbb{S}^1$ supports a complete metric of positive scalar curvature with exactly quadratic decay, but does not admit a decomposition as a connected sum., Comment: 20 pages, 7 figures

Large Language Models (LLMs) have seen widespread adoption due to their remarkable natural language capabilities. However, when deploying them in real-world settings, it is important to align LLMs to generate texts according to acceptable human standards. Methods such as Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO) have made significant progress in refining LLMs using human preference data. However, the privacy concerns inherent in utilizing such preference data have yet to be adequately studied. In this paper, we investigate the vulnerability of LLMs aligned using human preference datasets to membership inference attacks (MIAs), highlighting the shortcomings of previous MIA approaches with respect to preference data. Our study has two main contributions: first, we introduce a novel reference-based attack framework specifically for analyzing preference data called PREMIA (\uline{Pre}ference data \uline{MIA}); second, we provide empirical evidence that DPO models are more vulnerable to MIA compared to PPO models. Our findings highlight gaps in current privacy-preserving practices for LLM alignment.

In heterogeneous scenarios where the data distribution amongst the Federated Learning (FL) participants is Non-Independent and Identically distributed (Non-IID), FL suffers from the well known problem of data heterogeneity. This leads the performance of FL to be significantly degraded, as the global model tends to struggle to converge. To solve this problem, we propose Differentially Private Synthetic Data Aided Federated Learning Using Foundation Models (DPSDA-FL), a novel data augmentation strategy that aids in homogenizing the local data present on the clients' side. DPSDA-FL improves the training of the local models by leveraging differentially private synthetic data generated from foundation models. We demonstrate the effectiveness of our approach by evaluating it on the benchmark image dataset: CIFAR-10. Our experimental results have shown that DPSDA-FL can improve class recall and classification accuracy of the global model by up to 26% and 9%, respectively, in FL with Non-IID issues.

We present a novel, and effective, approach to the long-standing problem of mesh adaptivity in finite element methods (FEM). FE solvers are powerful tools for solving partial differential equations (PDEs), but their cost and accuracy are critically dependent on the choice of mesh points. To keep computational costs low, mesh relocation (r-adaptivity) seeks to optimise the position of a fixed number of mesh points to obtain the best FE solution accuracy. Classical approaches to this problem require the solution of a separate nonlinear "meshing" PDE to find the mesh point locations. This incurs significant cost at remeshing and relies on certain a-priori assumptions and guiding heuristics for optimal mesh point location. Recent machine learning approaches to r-adaptivity have mainly focused on the construction of fast surrogates for such classical methods. Our new approach combines a graph neural network (GNN) powered architecture, with training based on direct minimisation of the FE solution error with respect to the mesh point locations. The GNN employs graph neural diffusion (GRAND), closely aligning the mesh solution space to that of classical meshing methodologies, thus replacing heuristics with a learnable strategy, and providing a strong inductive bias. This allows for rapid and robust training and results in an extremely efficient and effective GNN approach to online r-adaptivity. This method outperforms classical and prior ML approaches to r-adaptive meshing on the test problems we consider, in particular achieving lower FE solution error, whilst retaining the significant speed-up over classical methods observed in prior ML work.

In this study, we explored the progression trajectories of artificial intelligence (AI) systems through the lens of complexity theory. We challenged the conventional linear and exponential projections of AI advancement toward Artificial General Intelligence (AGI) underpinned by transformer-based architectures, and posited the existence of critical points, akin to phase transitions in complex systems, where AI performance might plateau or regress into instability upon exceeding a critical complexity threshold. We employed agent-based modelling (ABM) to simulate hypothetical scenarios of AI systems' evolution under specific assumptions, using benchmark performance as a proxy for capability and complexity. Our simulations demonstrated how increasing the complexity of the AI system could exceed an upper criticality threshold, leading to unpredictable performance behaviours. Additionally, we developed a practical methodology for detecting these critical thresholds using simulation data and stochastic gradient descent to fine-tune detection thresholds. This research offers a novel perspective on AI advancement that has a particular relevance to Large Language Models (LLMs), emphasising the need for a tempered approach to extrapolating AI's growth potential and underscoring the importance of developing more robust and comprehensive AI performance benchmarks.

We consider the problem of one-dimensional function approximation using shallow neural networks (NN) with a rectified linear unit (ReLU) activation function and compare their training with traditional methods such as univariate Free Knot Splines (FKS). ReLU NNs and FKS span the same function space, and thus have the same theoretical expressivity. In the case of ReLU NNs, we show that their ill-conditioning degrades rapidly as the width of the network increases. This often leads to significantly poorer approximation in contrast to the FKS representation, which remains well-conditioned as the number of knots increases. We leverage the theory of optimal piecewise linear interpolants to improve the training procedure for a ReLU NN. Using the equidistribution principle, we propose a two-level procedure for training the FKS by first solving the nonlinear problem of finding the optimal knot locations of the interpolating FKS. Determining the optimal knots then acts as a good starting point for training the weights of the FKS. The training of the FKS gives insights into how we can train a ReLU NN effectively to give an equally accurate approximation. More precisely, we combine the training of the ReLU NN with an equidistribution based loss to find the breakpoints of the ReLU functions, combined with preconditioning the ReLU NN approximation (to take an FKS form) to find the scalings of the ReLU functions, leads to a well-conditioned and reliable method of finding an accurate ReLU NN approximation to a target function. We test this method on a series or regular, singular, and rapidly varying target functions and obtain good results realising the expressivity of the network in this case.

Co-speech gestures play a crucial role in the interactions between humans and embodied conversational agents (ECA). Recent deep learning methods enable the generation of realistic, natural co-speech gestures synchronized with speech, but such approaches require large amounts of training data. "In-the-wild" datasets, which compile videos from sources such as YouTube through human pose detection models, offer a solution by providing 2D skeleton sequences that are paired with speech. Concurrently, innovative lifting models have emerged, capable of transforming these 2D pose sequences into their 3D counterparts, leading to large and diverse datasets of 3D gestures. However, the derived 3D pose estimation is essentially a pseudo-ground truth, with the actual ground truth being the 2D motion data. This distinction raises questions about the impact of gesture representation dimensionality on the quality of generated motions, a topic that, to our knowledge, remains largely unexplored. In this work, we evaluate the impact of the dimensionality of the training data, 2D or 3D joint coordinates, on the performance of a multimodal speech-to-gesture deep generative model. We use a lifting model to convert 2D-generated sequences of body pose to 3D. Then, we compare the sequence of gestures generated directly in 3D to the gestures generated in 2D and lifted to 3D as post-processing., Comment: 8 pages. Paper accepted at WACAI 2024

Pairwise dot-product self-attention is key to the success of transformers that achieve state-of-the-art performance across a variety of applications in language and vision. This dot-product self-attention computes attention weights among the input tokens using Euclidean distance, which makes the model prone to representation collapse and vulnerable to contaminated samples. In this paper, we propose using a Mahalanobis distance metric for computing the attention weights to stretch the underlying feature space in directions of high contextual relevance. In particular, we define a hyper-ellipsoidal neighborhood around each query to increase the attention weights of the tokens lying in the contextually important directions. We term this novel class of attention Elliptical Attention. Our Elliptical Attention provides two benefits: 1) reducing representation collapse and 2) enhancing the model's robustness as Elliptical Attention pays more attention to contextually relevant information rather than focusing on some small subset of informative features. We empirically demonstrate the advantages of Elliptical Attention over the baseline dot-product attention and state-of-the-art attention methods on various practical tasks, including object classification, image segmentation, and language modeling across different data modalities., Comment: 10 pages in the main text. Published at NeurIPS 2024. The code is available at https://github.com/stefvk/Elliptical-Attention

The remarkable success of transformers in sequence modeling tasks, spanning various applications in natural language processing and computer vision, is attributed to the critical role of self-attention. Similar to the development of most deep learning models, the construction of these attention mechanisms relies on heuristics and experience. In our work, we derive self-attention from kernel principal component analysis (kernel PCA) and show that self-attention projects its query vectors onto the principal component axes of its key matrix in a feature space. We then formulate the exact formula for the value matrix in self-attention, theoretically and empirically demonstrating that this value matrix captures the eigenvectors of the Gram matrix of the key vectors in self-attention. Leveraging our kernel PCA framework, we propose Attention with Robust Principal Components (RPC-Attention), a novel class of robust attention that is resilient to data contamination. We empirically demonstrate the advantages of RPC-Attention over softmax attention on the ImageNet-1K object classification, WikiText-103 language modeling, and ADE20K image segmentation task., Comment: 10 pages in the main text. Published at NeurIPS 2024. The code is available at https://github.com/rachtsy/KPCA_code

Measuring the obliquities of stars hosting giant planets may shed light on the dynamical history of planetary systems. Significant efforts have been made to measure the obliquities of FGK stars with hot Jupiters, mainly based on observations of the Rossiter-McLaughlin effect. In contrast, M dwarfs with hot Jupiters have hardly been explored, because such systems are rare and often not favorable for such precise observations. Here, we report the first detection of the Rossiter-McLaughlin effect for an M dwarf with a hot Jupiter, TOI-4201, using the Gemini-North/MAROON-X spectrograph. We find TOI-4201 to be well-aligned with its giant planet, with a sky-projected obliquity of $\lambda=-3.0_{-3.2}^{+3.7}\ ^{\circ}$ and a true obliquity of $\psi=21.3_{-12.8}^{+12.5}\ ^{\circ}$ with an upper limit of $40^{\circ}$ at a 95% confidence level. The result agrees with dynamically quiet formation or tidal obliquity damping that realigned the system. As the first hot Jupiter around an M dwarf with its obliquity measured, TOI-4201b joins the group of aligned giant planets around cool stars ($T_{\rm eff}<6250\ K$), as well as the small but growing sample of planets with relatively high planet-to-star mass ratio ($M_p/M_\ast\gtrsim 3\times 10^{-3}$) that also appear to be mostly aligned., Comment: 12 pages, 5 figures, 3 tables, accepted to ApJL

We report the $p+p$ and $p+d$ differential cross sections measured in the SeaQuest experiment for $J/\psi$ and $\psi\left(2S\right)$ production at 120 GeV beam energy covering the forward $x$-Feynman ($x_F$) range of $0.5 < x_F <0.9$. The measured cross sections are in good agreement with theoretical calculations based on the nonrelativistic QCD (NRQCD) using the long-distance matrix elements deduced from a recent global analysis of proton- and pion-induced charmonium production data. The $\sigma_{\psi\left(2S\right)} / \sigma_{J/\psi}$ cross section ratios are found to increase as $x_F$ increases, indicating that the $q \bar{q}$ annihilation process has larger contributions in the $\psi\left(2S\right)$ production than the $J/\psi$ production. The $\sigma_{pd}/2\sigma_{pp}$ cross section ratios are observed to be significantly different for the Drell-Yan process and $J/\psi$ production, reflecting their different production mechanisms. We find that the $\sigma_{pd}/2\sigma_{pp}$ ratios for $J/\psi$ production at the forward $x_F$ region are sensitive to the $\bar{d}/ \bar{u}$ flavor asymmetry of the proton sea, analogous to the Drell-Yan process. The transverse momentum ($p_T$) distributions for $J/\psi$ and $\psi\left(2S\right)$ production are also presented and compared with data collected at higher center-of-mass energies., Comment: 8 pages, 7 figures, with 1 page of supplementary material which has 2 figures

In the event of natural or man-made disasters in an urban environment, such as fires, floods, and earthquakes, a swarm of unmanned aerial vehicles (UAVs) can rapidly sweep and provide coverage to monitor the area of interest and locate survivors. We propose a modular framework and patrol strategy that enables a swarm of UAVs to perform cooperative and periodic coverage in such scenarios. Our approach first discretizes the area of interest into viewpoints connected via closed paths. UAVs are assigned to teams via task allocation to cooperatively patrol these closed paths. We propose a minimal, scalable, and robust patrol strategy where UAVs within a team move in a random direction along their assigned closed path and "bounce" off each other when they meet. Our simulation results show that such a minimal strategy can exhibit an emergent behaviour that provides periodic and complete coverage in a 3D urban environment., Comment: Accepted for the 2024 International Conference on Unmanned Aircraft Systems (ICUAS 2024) in Chania, Greece

Distributed hybrid rendering (DHR) is a real-time rendering approach that incorporates cloud-based ray tracing with locally rasterized graphics for interactive thin client metaverse and game applications. With cloud assistance, DHR can generate high-fidelity ray-traced graphics contents remotely and deliver them to thin clients with low graphics capability, including standalone extended reality devices and mobile phones, while maintaining interactive frame rates for users under adverse network conditions. DHR can already achieve the effect of ray-traced hard shadows that form with the occlusion of direct illumination. We enhance the realism of these shadows by softening their edges with the direction of rays traced and approximating the occlusion of indirect illumination by reconstructing ray-traced ambient occlusion with a modified version of spatiotemporal variance-guided filtering. Our technique uses only 20-30% of the bandwidth of remote rendering and is also tolerant of delays of up to 200 ms with only slight distortion to the shadows along object edges.

We introduce the framework of Bayesian relative belief that directly evaluates whether or not the experimental data at hand supports a given hypothesis regarding a quantum system by directly comparing the prior and posterior probabilities for the hypothesis. In model-dimension certification tasks, we show that the relative belief procedure typically chooses Hilbert spaces that are never smaller in dimension than those selected from optimizing a broad class of information criteria, including Akaike's criterion. As a concrete and focused exposition of this powerful evidence-based technique, we apply the relative belief procedure to an important application: state reconstruction of imperfect quantum sources. In particular, just by comparing prior and posterior probabilities based on data, we demonstrate its capability of tracking multiphoton emissions using (realistically lossy) single-photon detectors in order to assess the actual quality of photon sources without making ad hoc assumptions, thereby reliably safeguarding source integrity for general quantum-information and communication tasks with Bayesian reasoning. Finally, we discuss how relative belief can be exploited to carry out parametric model certification and estimate the total dimension of the quantum state for the combined (measured) physical and interacting external systems described by the Tavis--Cummings model., Comment: 16 pages, 8 figures, companion letter in arXiv:2401.01562