Search

Functions are a fundamental object in mathematics, with countless applications to different fields, and are usually classified based on certain properties, given their domains and images. An important property of a real-valued function is its convexity, which plays a very crucial role in many areas, such as thermodynamics and geometry. Motivated by recent advances in quantum computation as well as the quest for quantum advantage, we give a quantum algorithm for testing convexity of polynomial functions, which appears frequently in multiple contexts, such as optimization, machine learning, physics, etc. We show that quantum computers can reveal the convexity property superpolynomially faster than classical computers with respect to number of variables. As a corollary, we provide a significant improvement and extension on quantum Newton's method constructed in earlier work of Rebentrost et al [New J. Phys. \textbf{21} 073023 (2019)]. We further discuss our algorithm in a broader context, such as potential application in the study of geometric structure of manifold, testing training landscape of variational quantum algorithm and also gradient descent/Newton's method for optimization.

Absorption and scattering properties of nanoparticles immersed in an absorbing medium are essential in understanding the overall properties of composites and in designing materials with expected functionalities. In this paper, we establish the model that links the absorption and scattering coefficients of individual particles with the reflectance and transmittance spectra of thin-film composite, supported by well-controlled experiments. Thin films consisting of TiO$_2$ nanoparticles embedded in PMMA are fabricated on glass substrates using spin-coating and then peeled off to form standalone samples for spectroscopy measurements. The absorption $K$ and scattering $S$ coefficients of multiple nanoparticles are calculated from the measured transmittance and reflectance using the Kubelka-Munk theory in combination with the Saunderson correction. From the theoretical side, the absorption $K$ and scattering $S$ coefficients are the accumulation of absorption and scattering of individual particles, which are derived from the Mie theory. The agreement between the $K$ and $S$ coefficients extracted from experimental data and from theoretical calculations gives deep insight into the significant attenuating effect of absorption and scattering on each particle due to the surrounding medium. Furthermore, the validated model of nanoparticles immersed in an absorbing medium can be used to obtain the preliminary results for materials designed toward radiative cooling., Comment: 27 pages, 13 figures

Quantum singular value transformation (QSVT) is a framework that has been shown to unify many primitives in quantum algorithms. In this work, we leverage the QSVT framework in two directions. We first show that the QSVT framework can accelerate one recently introduced quantum power method, which substantially improves its running time. Additionally, we incorporate several elementary numerical integration techniques, such as the rectangular method, Monte Carlo method, and quadrature method, into the QSVT framework, which results in polynomial speedup with respect to the size or the number of points of the grid. Our results thus provide further examples to demonstrate the potential of the QSVT and how it may enhance quantum algorithmic tasks.

Prompt learning methods are gaining increasing attention due to their ability to customize large vision-language models to new domains using pre-trained contextual knowledge and minimal training data. However, existing works typically rely on optimizing unified prompt inputs, often struggling with fine-grained classification tasks due to insufficient discriminative attributes. To tackle this, we consider a new framework based on a dual context of both domain-shared and class-specific contexts, where the latter is generated by Large Language Models (LLMs) such as GPTs. Such dual prompt methods enhance the model's feature representation by joining implicit and explicit factors encoded in LLM knowledge. Moreover, we formulate the Unbalanced Optimal Transport (UOT) theory to quantify the relationships between constructed prompts and visual tokens. Through partial matching, UOT can properly align discrete sets of visual tokens and prompt embeddings under different mass distributions, which is particularly valuable for handling irrelevant or noisy elements, ensuring that the preservation of mass does not restrict transport solutions. Furthermore, UOT's characteristics integrate seamlessly with image augmentation, expanding the training sample pool while maintaining a reasonable distance between perturbed images and prompt inputs. Extensive experiments across few-shot classification and adapter settings substantiate the superiority of our model over current state-of-the-art baselines., Comment: Version 1

Social science research has shown that candidates with names indicative of certain races or genders often face discrimination in employment practices. Similarly, Large Language Models (LLMs) have demonstrated racial and gender biases in various applications. In this study, we utilize GPT-3.5-Turbo and Llama 3-70B-Instruct to simulate hiring decisions and salary recommendations for candidates with 320 first names that strongly signal their race and gender, across over 750,000 prompts. Our empirical results indicate a preference among these models for hiring candidates with White female-sounding names over other demographic groups across 40 occupations. Additionally, even among candidates with identical qualifications, salary recommendations vary by as much as 5% between different subgroups. A comparison with real-world labor data reveals inconsistent alignment with U.S. labor market characteristics, underscoring the necessity of risk investigation of LLM-powered systems., Comment: preprint, 18 pages

After providing an explicit K-stability condition for a $\mathbb{Q}$-Gorenstein log spherical cone, we prove the existence and uniqueness of an equivariant K-stable degeneration of the cone, and deduce uniqueness of the asymptotic cone of a given complete $K$-invariant Calabi-Yau metric in the trivial class of an affine $G$-spherical manifold, $K$ being the maximal compact subgroup of $G$. Next, we prove that the valuation induced by $K$-invariant Calabi-Yau metrics on affine $G$-spherical manifolds is in fact $G$-invariant. As an application, we point out an affine smoothing of a Calabi-Yau cone that does not admit any $K$-invariant Calabi-Yau metrics asymptotic to the cone. Another corollary is that on $\mathbb{C}^3$, there are no other complete Calabi-Yau metrics with maximal volume growth and spherical symmetry other than the standard flat metric and the Li-Conlon-Rochon-Sz\'ekelyhidi metrics with horospherical asymptotic cone. This answers the question whether there is a nontrivial asymptotic cone with smooth cross section on $\mathbb{C}^{3}$ raised by Conlon-Rochon when the symmetry is spherical., Comment: 33 pages. Comments welcome !

Nonlinear equations are challenging to solve due to their inherently nonlinear nature. As analytical solutions typically do not exist, numerical methods have been developed to tackle their solutions. In this article, we give a quantum algorithm for solving a system of nonlinear algebraic equations, in which each equation is a multivariate polynomial of known coefficients. Building upon the classical Newton method and some recent works on quantum algorithm plus block encoding from the quantum singular value transformation, we show how to invert the Jacobian matrix to execute Newton's iterative method for solving nonlinear equations, where each contributing equation is a homogeneous polynomial of an even degree. A detailed analysis are then carried out to reveal that our method achieves polylogarithmic time in relative to the number of variables. Furthermore, the number of required qubits is logarithmic in the number of variables. In particular, we also show that our method can be modified with little effort to deal with polynomial of various types, thus implying the generality of our approach. Some examples coming from physics and algebraic geometry, such as Gross-Pitaevski equation, Lotka-Volterra equations, and intersection of algebraic varieties, involving nonlinear partial differential equations are provided to motivate the potential application, with a description on how to extend our algorithm with even less effort in such a scenario. Our work thus marks a further important step towards quantum advantage in nonlinear science, enabled by the framework of quantum singular value transformation.

This paper investigates the problem of informative path planning for a mobile robotic sensor network in spatially temporally distributed mapping. The robots are able to gather noisy measurements from an area of interest during their movements to build a Gaussian Process (GP) model of a spatio-temporal field. The model is then utilized to predict the spatio-temporal phenomenon at different points of interest. To spatially and temporally navigate the group of robots so that they can optimally acquire maximal information gains while their connectivity is preserved, we propose a novel multistep prediction informative path planning optimization strategy employing our newly defined local cost functions. By using the dual decomposition method, it is feasible and practical to effectively solve the optimization problem in a distributed manner. The proposed method was validated through synthetic experiments utilizing real-world data sets.

As a research-product hybrid group in AI for Software Engineering (AI4SE), we present four key takeaways from our experience developing in-IDE AI coding assistants. AI coding assistants should set clear expectations for usage, integrate with advanced IDE capabilities and existing extensions, use extendable backend designs, and collect app data responsibly for downstream analyses. We propose open questions and challenges that academia and industry should address to realize the vision of next-generation AI coding assistants.

The widespread use of social media necessitates reliable and efficient detection of offensive content to mitigate harmful effects. Although sophisticated models perform well on individual datasets, they often fail to generalize due to varying definitions and labeling of "offensive content." In this paper, we introduce HateCOT, an English dataset with over 52,000 samples from diverse sources, featuring explanations generated by GPT-3.5Turbo and curated by humans. We demonstrate that pretraining on HateCOT significantly enhances the performance of open-source Large Language Models on three benchmark datasets for offensive content detection in both zero-shot and few-shot settings, despite differences in domain and task. Additionally, HateCOT facilitates effective K-shot fine-tuning of LLMs with limited data and improves the quality of their explanations, as confirmed by our human evaluation., Comment: Preprint Version 3

Topological data analysis (TDA) is a fast-growing field that utilizes advanced tools from topology to analyze large-scale data. A central problem in topological data analysis is estimating the so-called Betti numbers of the underlying simplicial complex. While the difficulty of this problem has been established as NP-hard, previous works have showcased appealing quantum speedup. In this article, we provide an alternative method for estimating Betti numbers and normalized Betti numbers of given simplicial complex, based on some recent advances in quantum algorithm, specifically, quantum singular value transformation. Our method can be faster than the best-known classical method for finding Betti numbers, and interestingly, it can also find the Betti numbers of the complement graph to our original one. Comparing to the best known quantum algorithm, our method generally requires lower depth circuit, in trade-off for longer running time. Regarding normalized Betti numbers, our method could match the running time of best-known quantum method in the case of dense simplices.

In this manuscript, consideration is given to the existence of periodic traveling-wave solutions to the $abcd$-system. This system was derived by Bona, Saut, and Chen to describe small amplitude, long wavelength gravity waves on the surface of water. These exact solutions are formulated in terms of the Jacobi elliptic function cnoidal. The existence of explicit traveling-wave solutions is very useful in theoretical investigations such as stability of solutions, as well as other numerical analysis of the system., Comment: 19 pages, 6 figures, submitted to Water Waves

Question answering (QA) can only make progress if we know if an answer is correct, but for many of the most challenging and interesting QA examples, current efficient answer correctness (AC) metrics do not align with human judgments, particularly verbose, free-form answers from large language models (LLMs). There are two challenges: a lack of diverse evaluation data and that models are too big and non-transparent; LLM-based scorers correlate better with humans, but this expensive task has only been tested on limited QA datasets. We rectify these issues by providing guidelines and datasets for evaluating machine QA adopted from human QA community. We also propose an efficient, low-resource, and interpretable QA evaluation method more stable than an exact match and neural methods., Comment: Efficient PEDANTS Classifier for short-form QA in github: https://github.com/zli12321/qa_metrics. arXiv admin note: text overlap with arXiv:2401.13170

In this study, we employ a first-principles approach to conduct a comprehensive investigation of the properties of nine common native point defects in cubic boron nitride. This analysis combines standard semi-local and dielectric hybrid density-exchange-correlation functional calculations, encompassing vacancies, interstitials, antisites, and their complexes. Our findings elucidate the influence of these defects on the structural and electronic characteristics of cubic boron nitride, such as local structures, formation energy, magnetism, and the energies of defect states within the band gap. Notably, we accurately simulate the photoluminescent spectra of cubic boron nitride induced by these defects, demonstrating excellent agreement with experimental observations. This outcome indicates that the prominent peaks in the photoluminescent spectrum at 2.5 and 2.8 eV can be attributed to the nitrogen to boron antisite (N$_{\rm B}$) and boron interstitial (B$_{\rm i}$) defects, respectively. Additionally, we investigate the energetic stability of defects under various charge states, providing valuable references for benchmarking purposes., Comment: 11 pages, 5 figures, 2 tables

The propagation of offensive content through social media channels has garnered attention of the research community. Multiple works have proposed various semantically related yet subtle distinct categories of offensive speech. In this work, we explore meta-earning approaches to leverage the diversity of offensive speech corpora to enhance their reliable and efficient detection. We propose a joint embedding architecture that incorporates the input's label and definition for classification via Prototypical Network. Our model achieves at least 75% of the maximal F1-score while using less than 10% of the available training data across 4 datasets. Our experimental findings also provide a case study of training strategies valuable to combat resource scarcity., Comment: Accepted to Main Conference, EACL 2024

Question answering (QA) can only make progress if we know if an answer is correct, but for many of the most challenging and interesting QA examples, current evaluation metrics to determine answer equivalence (AE) often do not align with human judgments, particularly more verbose, free-form answers from large language models (LLM). There are two challenges: a lack of data and that models are too big: LLM-based scorers can correlate better with human judges, but this task has only been tested on limited QA datasets, and even when available, update of the model is limited because LLMs are large and often expensive. We rectify both of these issues by providing clear and consistent guidelines for evaluating AE in machine QA adopted from professional human QA contests. We also introduce a combination of standard evaluation and a more efficient, robust, and lightweight discriminate AE classifier-based matching method (CFMatch, smaller than 1 MB), trained and validated to more accurately evaluate answer correctness in accordance with adopted expert AE rules that are more aligned with human judgments., Comment: A duplicate and polished version is in arXiv:2402.11161

We show that on every non-$G_2$ complex symmetric space of rank two, there are complete Calabi-Yau metrics of Euclidean volume growth with prescribed horospherical singular tangent cone at infinity, providing the first examples of affine Calabi-Yau smoothings of singular and irregular tangent cone. As a corollary, we obtain infinitely many examples of Calabi-Yau manifolds degenerating to the tangent cone in a single step, supporting a recent conjecture by Sun-Zhang, which was only proved for asymptotically conical Calabi-Yau manifolds in the sense of Conlon-Hein., Comment: 45 pages

Block encoding is a key ingredient in the recently developed quantum signal processing that forms a unifying framework for quantum algorithms. Initially showcased for simplifying and optimizing resource utilization in several problems, such as searching, amplitude estimation, and Hamiltonian simulation, the capabilities of the quantum signal processing go beyond these and offer untapped potential for devising new quantum algorithms. In this article, we utilize block encoding to substantially enhance two previously proposed quantum algorithms: largest eigenvalue estimation and quantum gradient descent. Unlike previous works that involve sophisticated procedures, our findings, using the unitary block encoding, demonstrate that even with elementary operations, these new quantum algorithms can eliminate major scaling factors present in their original counterparts. This yields much more efficient quantum algorithms capable of tackling complex computational problems with remarkable efficiency. Furthermore, we show how to extend our proposed method to different contexts, including matrix inversion and multiple eigenvalues estimation.

Following the celebrated quantum algorithm for solving linear equations (so-called HHL algorithm), Childs, Kothari and Somma [SIAM Journal on Computing, {\bf 46}: 1920, (2017)] provided an approach to solve a linear system of equations with exponentially improved dependence on precision. In this note, we aim to complement such a result for applying a matrix to some quantum state, based upon their Chebyshev polynomial approach. A few examples that motivate this application are included and we further discuss an application of this improved matrix application algorithm explicitly with an efficient quantum procedure.

Constructing a robust model that can effectively generalize to test samples under distribution shifts remains a significant challenge in the field of medical imaging. The foundational models for vision and language, pre-trained on extensive sets of natural image and text data, have emerged as a promising approach. It showcases impressive learning abilities across different tasks with the need for only a limited amount of annotated samples. While numerous techniques have focused on developing better fine-tuning strategies to adapt these models for specific domains, we instead examine their robustness to domain shifts in the medical image segmentation task. To this end, we compare the generalization performance to unseen domains of various pre-trained models after being fine-tuned on the same in-distribution dataset and show that foundation-based models enjoy better robustness than other architectures. From here, we further developed a new Bayesian uncertainty estimation for frozen models and used them as an indicator to characterize the model's performance on out-of-distribution (OOD) data, proving particularly beneficial for real-world applications. Our experiments not only reveal the limitations of current indicators like accuracy on the line or agreement on the line commonly used in natural image applications but also emphasize the promise of the introduced Bayesian uncertainty. Specifically, lower uncertainty predictions usually tend to higher out-of-distribution (OOD) performance., Comment: Advances in Neural Information Processing Systems (NeurIPS) 2023, Workshop on robustness of zero/few-shot learning in foundation models

In the classic measurement error framework, covariates are contaminated by independent additive noise. This paper considers parameter estimation in such a linear errors-in-variables model where the unknown measurement error distribution is heteroscedastic across observations. We propose a new generalized method of moment (GMM) estimator that combines a moment correction approach and a phase function-based approach. The former requires distributions to have four finite moments, while the latter relies on covariates having asymmetric distributions. The new estimator is shown to be consistent and asymptotically normal under appropriate regularity conditions. The asymptotic covariance of the estimator is derived, and the estimated standard error is computed using a fast bootstrap procedure. The GMM estimator is demonstrated to have strong finite sample performance in numerical studies, especially when the measurement errors follow non-Gaussian distributions.

We consider the problem of surrogate sufficient dimension reduction, that is, estimating the central subspace of a regression model, when the covariates are contaminated by measurement error. When no measurement error is present, a likelihood-based dimension reduction method that relies on maximizing the likelihood of a Gaussian inverse regression model on the Grassmann manifold is well-known to have superior performance to traditional inverse moment methods. We propose two likelihood-based estimators for the central subspace in measurement error settings, which make different adjustments to the observed surrogates. Both estimators are computed based on maximizing objective functions on the Grassmann manifold and are shown to consistently recover the true central subspace. When the central subspace is assumed to depend on only a few covariates, we further propose to augment the likelihood function with a penalty term that induces sparsity on the Grassmann manifold to obtain sparse estimators. The resulting objective function has a closed-form Riemann gradient which facilitates efficient computation of the penalized estimator. We leverage the state-of-the-art trust region algorithm on the Grassmann manifold to compute the proposed estimators efficiently. Simulation studies and a data application demonstrate the proposed likelihood-based estimators perform better than inverse moment-based estimators in terms of both estimation and variable selection accuracy.

Some systems were recently put forth by Nguyen et. al. as models for studying the interaction of long and short waves in dispersive media. These systems were shown to possess synchronized Jacobi elliptic solutions as well as synchronized solitary wave solutions under certain constraints, i.e., vector solutions where the two components are proportional to one another. In this paper, the exact periodic traveling wave solutions to these systems in general are found to be given by Jacobi elliptic functions. Moreover, these cnoidal wave solutions are unique. Thus, the explicit synchronized solutions under some conditions obtained by Nguyen et. al. are also indeed unique., Comment: 21 pages, 4 figures, to appear in AIMS Mathematics. This is the final version, incorporating the referee comments

Topological data analysis has emerged as a powerful tool for analyzing large-scale data. High-dimensional data form an abstract simplicial complex, and by using tools from homology, topological features could be identified. Given a simplex, an important feature is so-called Betti numbers. Calculating Betti numbers classically is a daunting task due to the massive volume of data and its possible high-dimension. While most known quantum algorithms to estimate Betti numbers rely on homology, here we consider the `dual' approach, which is inspired by Hodge theory and de Rham cohomology, combined with recent advanced techniques in quantum algorithms. Our cohomology method offers a relatively simpler, yet more natural framework that requires exponentially less qubits, in comparison with the known homology-based quantum algorithms. Furthermore, our algorithm can calculate its $r$-th Betti number $\beta_r$ up to some multiplicative error $\delta$ with running time $\mathcal{O}\big( \log(c_r) c_r^2 / (c_r - \beta_r)^2 \delta^2 \big)$, where $c_r$ is the number of $r$-simplex. It thus works best when the $r$-th Betti number is considerably smaller than the number of the $r$-simplex in the given triangulated manifold.

Geometry and topology have generated impacts far beyond their pure mathematical primitive, providing a solid foundation for many applicable tools. Typically, real-world data are represented as vectors, forming a linear subspace for a given data collection. Computing distances between different subspaces is generally a computationally challenging problem with both theoretical and applicable consequences, as, for example, the results can be used to classify data from different categories. Fueled by the fast-growing development of quantum algorithms, we consider such problems in the quantum context and provide a quantum algorithm for estimating two kinds of distance: Grassmann distance and ellipsoid distance. Under appropriate assumptions and conditions, the speedup of our quantum algorithm is exponential with respect to both the dimension of the given data and the number of data points. Some extensions regarding estimating different kinds of distance are then discussed as a corollary of our main quantum algorithmic method.

Physics-informed machine learning (PIML) is a set of methods and tools that systematically integrate machine learning (ML) algorithms with physical constraints and abstract mathematical models developed in scientific and engineering domains. As opposed to purely data-driven methods, PIML models can be trained from additional information obtained by enforcing physical laws such as energy and mass conservation. More broadly, PIML models can include abstract properties and conditions such as stability, convexity, or invariance. The basic premise of PIML is that the integration of ML and physics can yield more effective, physically consistent, and data-efficient models. This paper aims to provide a tutorial-like overview of the recent advances in PIML for dynamical system modeling and control. Specifically, the paper covers an overview of the theory, fundamental concepts and methods, tools, and applications on topics of: 1) physics-informed learning for system identification; 2) physics-informed learning for control; 3) analysis and verification of PIML models; and 4) physics-informed digital twins. The paper is concluded with a perspective on open challenges and future research opportunities., Comment: 16 pages, 4 figures, to be published in 2023 American Control Conference (ACC)

Obtaining large pre-trained models that can be fine-tuned to new tasks with limited annotated samples has remained an open challenge for medical imaging data. While pre-trained deep networks on ImageNet and vision-language foundation models trained on web-scale data are prevailing approaches, their effectiveness on medical tasks is limited due to the significant domain shift between natural and medical images. To bridge this gap, we introduce LVM-Med, the first family of deep networks trained on large-scale medical datasets. We have collected approximately 1.3 million medical images from 55 publicly available datasets, covering a large number of organs and modalities such as CT, MRI, X-ray, and Ultrasound. We benchmark several state-of-the-art self-supervised algorithms on this dataset and propose a novel self-supervised contrastive learning algorithm using a graph-matching formulation. The proposed approach makes three contributions: (i) it integrates prior pair-wise image similarity metrics based on local and global information; (ii) it captures the structural constraints of feature embeddings through a loss function constructed via a combinatorial graph-matching objective; and (iii) it can be trained efficiently end-to-end using modern gradient-estimation techniques for black-box solvers. We thoroughly evaluate the proposed LVM-Med on 15 downstream medical tasks ranging from segmentation and classification to object detection, and both for the in and out-of-distribution settings. LVM-Med empirically outperforms a number of state-of-the-art supervised, self-supervised, and foundation models. For challenging tasks such as Brain Tumor Classification or Diabetic Retinopathy Grading, LVM-Med improves previous vision-language models trained on 1 billion masks by 6-7% while using only a ResNet-50., Comment: Accepted at NeurIPS 2023

A method for the optimisation and advanced studies of a laser-plasma electron injector is presented, based on a truncated ionisation injection scheme for high quality beam production. The SMILEI code is used with laser envelope approximation and a low number of particles per cell to reach computation time performances enabling the production of a large number of accelerator configurations. The developed and tested workflow is a possible approach for the production of large dataset for laser-plasma accelerator optimisation. A selection of functions of merit used to grade generated electron beams is discussed. Among the significant number of configurations, two specific working points are presented in details. All data generated are left open to the scientific community for further study and optimisation.

Modeling and simulations of pandemic dynamics play an essential role in understanding and addressing the spreading of highly infectious diseases such as COVID-19. In this work, we propose a novel deep learning architecture named Attention-based Multiresolution Graph Neural Networks (ATMGNN) that learns to combine the spatial graph information, i.e. geographical data, with the temporal information, i.e. timeseries data of number of COVID-19 cases, to predict the future dynamics of the pandemic. The key innovation is that our method can capture the multiscale structures of the spatial graph via a learning to cluster algorithm in a data-driven manner. This allows our architecture to learn to pick up either local or global signals of a pandemic, and model both the long-range spatial and temporal dependencies. Importantly, we collected and assembled a new dataset for New Zealand. We established a comprehensive benchmark of statistical methods, temporal architectures, graph neural networks along with our spatio-temporal model. We also incorporated socioeconomic cross-sectional data to further enhance our prediction. Our proposed model have shown highly robust predictions and outperformed all other baselines in various metrics for our new dataset of New Zealand along with existing datasets of England, France, Italy and Spain. For a future work, we plan to extend our work for real-time prediction and global scale. Our data and source code are publicly available at https://github.com/HySonLab/pandemic_tgnn

Poly(methyl methacrylate), PMMA, is a popular polymer for optical applications due to its superior transmission and reflection. However, its optical properties in the ultraviolet regime still need to be thoroughly studied. Using the reflection-transmission method, we determine its complex refractive index by numerically analyzing the measured data from thin films. The PMMA standalone film is fabricated by peeling off its substrate after spin-coating. Its transmittance and reflectance are then measured in a spectrophotometer using an integrated sphere. The complex refractive index $n+ik$ is extracted theoretically from the measured transmittance $T$ and reflectance $R$. The uncertainties of measured $n$ and $k$ are discussed in the two limits of strong absorption and weak absorption of the materials to illustrate the advantage and disadvantages of the approach., Comment: 10 pages, 4 figures

We present The Vault, a dataset of high-quality code-text pairs in multiple programming languages for training large language models to understand and generate code. We present methods for thoroughly extracting samples that use both rule-based and deep learning-based methods to ensure that they contain high-quality pairs of code and text, resulting in a dataset of 43 million high-quality code-text pairs. Our extensive evaluations on common coding tasks including code generation, code search and code summarization show that when fine-tuning Code Large Language Models on The Vault, such models outperform the same models trained on other datasets such as CodeSearchNet. We also provide detailed analyses of our datasets to assess the effects of various programming languages and docstrings on the performance of such models., Comment: Accepted at EMNLP 2023, Long Findings

The retrieval of 3D objects has gained significant importance in recent years due to its broad range of applications in computer vision, computer graphics, virtual reality, and augmented reality. However, the retrieval of 3D objects presents significant challenges due to the intricate nature of 3D models, which can vary in shape, size, and texture, and have numerous polygons and vertices. To this end, we introduce a novel SHREC challenge track that focuses on retrieving relevant 3D animal models from a dataset using sketch queries and expedites accessing 3D models through available sketches. Furthermore, a new dataset named ANIMAR was constructed in this study, comprising a collection of 711 unique 3D animal models and 140 corresponding sketch queries. Our contest requires participants to retrieve 3D models based on complex and detailed sketches. We receive satisfactory results from eight teams and 204 runs. Although further improvement is necessary, the proposed task has the potential to incentivize additional research in the domain of 3D object retrieval, potentially yielding benefits for a wide range of applications. We also provide insights into potential areas of future research, such as improving techniques for feature extraction and matching and creating more diverse datasets to evaluate retrieval performance. https://aichallenge.hcmus.edu.vn/sketchanimar, Comment: Accepted to Computers & Graphics (3DOR 2023, Journal track)

3D object retrieval is an important yet challenging task that has drawn more and more attention in recent years. While existing approaches have made strides in addressing this issue, they are often limited to restricted settings such as image and sketch queries, which are often unfriendly interactions for common users. In order to overcome these limitations, this paper presents a novel SHREC challenge track focusing on text-based fine-grained retrieval of 3D animal models. Unlike previous SHREC challenge tracks, the proposed task is considerably more challenging, requiring participants to develop innovative approaches to tackle the problem of text-based retrieval. Despite the increased difficulty, we believe this task can potentially drive useful applications in practice and facilitate more intuitive interactions with 3D objects. Five groups participated in our competition, submitting a total of 114 runs. While the results obtained in our competition are satisfactory, we note that the challenges presented by this task are far from fully solved. As such, we provide insights into potential areas for future research and improvements. We believe we can help push the boundaries of 3D object retrieval and facilitate more user-friendly interactions via vision-language technologies. https://aichallenge.hcmus.edu.vn/textanimar, Comment: Accepted to Computers and Graphics (3DOR, Journal Track)