arXiv subjects

Kevin Xu

Publications and source records attributed to Kevin Xu.

At least 19 recordsLinked to original sources

arXiv2026

Industry Classification of GitHub Repositories Using the North American Industry Classification System (NAICS)

GitHub hosts hundreds of millions of public repositories, but the platform exposes no native mapping from repositories to standardized industry sectors. This gap limits empirical work on the geography of innovation, the industrial composition of open-source production, and the diffusion of new technologies across economic sectors. We present NAICS-GH, a publicly released corpus of 6,588 GitHub repositories drawn from source pools covering the United States, the European Union, and Australia, each labeled with a 2-digit sector from the North American Industry Classification System (NAICS 2022). Labels are produced by a retrieve-and-verify pipeline that combines BAAI/bge-large-en embeddings, FAISS retrieval, and GPT-4.1 rubric scoring. The pipeline narrows about 1.37 million source repositories to 31,178 candidate repository-sector pairs and retains 6,588 high-confidence labels with score at least 8. Re-running the retrieval pipeline end to end reproduces the candidate set to within 0.03 percent. On a 2,421-repository human-validated random sample, the released labels attain 96.98 percent precision, with Wilson 95 percent confidence interval [96.23, 97.59]. We benchmark six pretrained encoders on the released corpus; RoBERTa-large reaches 86.45 percent F1 and 86.35 percent accuracy on a held-out 20 percent test set. The dataset, Croissant metadata, pipeline code, prompts, and fine-tuned checkpoint are released under CC-BY-4.0 and MIT licenses.

cs.SE↗

arXiv2026

Fermi gas of polar molecules in the Pauli-blocked regime

Quantum gases of polar molecules have recently emerged as a powerful platform for exploring exotic many-body dynamics and correlated quantum behavior. To achieve the full potential of this platform, the production of deeply degenerate quantum gases of molecules in arbitrary confinement geometries is necessary. Here, we successfully evaporate fermionic KRb molecules in both 3D and quasi-2D geometries to well below their Fermi temperatures utilizing dipolar collisions. As we evaporate deeper into degeneracy in both geometries, we enter the Pauli-blocked regime with polar molecules, which we independently confirm for the first time by measuring the Pauli suppression of elastic collisions. Moreover, the Pauli suppression of collisions contributes to the limitation of our final molecular temperature to about 25% of the Fermi temperature in both geometries, particularly limiting quasi-2D evaporation where the Pauli blockade drastically reduces an otherwise large elastic to inelastic scattering ratio. This work demonstrates the production of degenerate Fermi gases of polar molecules both in a 3D harmonic trap and in mono- and bi-layer 2D configurations. Further, our work explores the fundamental limits on evaporation of molecular Fermi gases set by the Pauli-exclusion principle, which could be overcome in the future by introducing distinguishable scattering partners.

cond-mat.quant-gas↗

arXiv2026

Agentic Delegation and the Language Frontier of Software Developers: A Model and Evidence from Claude Code on GitHub

We develop and test a model of agentic delegation in software production. Developers face language-specific entry thresholds; conversational AI mainly augments work in languages they already know, while agentic AI adds delegated execution under developer specification and verification. The model predicts an activation band of unfamiliar languages that become feasible only with an agent, expanding the observed language-production frontier of the developer. We test this prediction in a monthly GitHub panel of 5,346 developers, dating adoption by first Claude Code co-authorship and constructing commit-level language outcomes from 57 million changed files. Doubly robust staggered-adoption event studies with not-yet-treated comparisons show sharp expansion at adoption: active languages rise by 2.5 relative to a 0.9 baseline, newly used languages by 1.2, entropy by 0.38, and cumulative breadth continues to grow afterward. The pattern survives removing the treatment-defining language, excluding all Claude-coauthored commits, conditioning on activity, and screening users of competing agents. Consistent with the model, first uses of unfamiliar languages concentrate among narrow pre-adoption specialists at each activity level. Because adoption is voluntary and may coincide with project shocks, the estimates are event-time associations rather than definitive causal effects.

econ.GN↗

arXiv2026

Uber's Failover Architecture: Reconciling Reliability and Efficiency in Hyperscale Microservice Infrastructure

Operating a global, real-time platform at Uber's scale requires infrastructure that is both resilient and cost-efficient. Historically, reliability was ensured through a costly 2x capacity model--each service provisioned to handle global traffic independently across two regions--leaving half the fleet idle. We present Uber's Failover Architecture (UFA), which replaces the uniform 2x model with a differentiated architecture aligned to business criticality. Critical services retain failover guarantees, while non-critical services opportunistically use failover buffer capacity reserved for critical services during steady state. During rare "full-peak" failovers, non-critical services are selectively preempted and rapidly restored, with differentiated Service-Level Agreements (SLAs) using on-demand capacity. Automated safeguards, including dependency analysis and regression gates, ensure critical services continue to function even while non-critical services are unavailable. The quantitative impact is significant: UFA reduces steady-state provisioning from 2x to 1.3x, raising utilization from ~20% to ~30% while sustaining 99.97% availability. To date, UFA has hardened over 4,000 unsafe dependencies, eliminated over one million CPU cores from a baseline of about four million cores.

cs.DC↗

arXiv2025

A Formal Comparison Between Chain of Thought and Latent Thought

Chain of thought (CoT) elicits reasoning in large language models by explicitly generating intermediate tokens. In contrast, latent thought reasoning operates directly in the continuous latent space, enabling computation beyond discrete linguistic representations. While both approaches exploit iterative computation, their comparative capabilities remain underexplored. In this work, we present a formal analysis showing that latent thought admits more efficient parallel computation than inherently sequential CoT. In contrast, CoT enables approximate counting and sampling through stochastic decoding. These separations suggest the tasks for which depth-driven recursion is more suitable, thereby offering practical guidance for choosing between reasoning paradigms.

cs.AI↗

arXiv2025

Neural Cellular Automata for ARC-AGI

Cellular automata and their differentiable counterparts, Neural Cellular Automata (NCA), are highly expressive and capable of surprisingly complex behaviors. This paper explores how NCAs perform when applied to tasks requiring precise transformations and few-shot generalization, using the Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) as a domain that challenges their capabilities in ways not previously explored. Specifically, this paper uses gradient-based training to learn iterative update rules that transform input grids into their outputs from the training examples and apply them to the test inputs. Results suggest that gradient-trained NCA models are a promising and efficient approach to a range of abstract grid-based tasks from ARC. Along with discussing the impacts of various design modifications and training constraints, this work examines the behavior and properties of NCAs applied to ARC to give insights for broader applications of self-organizing systems.

cs.NE↗

arXiv2025

To CoT or To Loop? A Formal Comparison Between Chain-of-Thought and Looped Transformers

Chain-of-Thought (CoT) and Looped Transformers have been shown to empirically improve performance on reasoning tasks and to theoretically enhance expressivity by recursively increasing the number of computational steps. However, their comparative capabilities are still not well understood. In this paper, we provide a formal analysis of their respective strengths and limitations. We show that Looped Transformers can efficiently simulate parallel computations for deterministic tasks, which we formalize as evaluation over directed acyclic graphs. In contrast, CoT with stochastic decoding excels at approximate inference for compositional structures, namely self-reducible problems. These separations suggest the tasks for which depth-driven recursion is more suitable, thereby offering practical cues for choosing between reasoning paradigms.

cs.LG↗

arXiv2025

Measuring AI agent autonomy: Towards a scalable approach with code inspection

AI agents are AI systems that can achieve complex goals autonomously. Assessing the level of agent autonomy is crucial for understanding both their potential benefits and risks. Current assessments of autonomy often focus on specific risks and rely on run-time evaluations -- observations of agent actions during operation. We introduce a code-based assessment of autonomy that eliminates the need to run an AI agent to perform specific tasks, thereby reducing the costs and risks associated with run-time evaluations. Using this code-based framework, the orchestration code used to run an AI agent can be scored according to a taxonomy that assesses attributes of autonomy: impact and oversight. We demonstrate this approach with the AutoGen framework and select applications.

cs.AI↗

arXiv2024

Revisiting Absence withSymptoms that T Show up Decades Later to Recover Empty Categories

This paper explores null elements in English, Chinese, and Korean Penn treebanks. Null elements contain important syntactic and semantic information, yet they have typically been treated as entities to be removed during language processing tasks, particularly in constituency parsing. Thus, we work towards the removal and, in particular, the restoration of null elements in parse trees. We focus on expanding a rule-based approach utilizing linguistic context information to Chinese, as rule based approaches have historically only been applied to English. We also worked to conduct neural experiments with a language agnostic sequence-to-sequence model to recover null elements for English (PTB), Chinese (CTB) and Korean (KTB). To the best of the authors' knowledge, null elements in three different languages have been explored and compared for the first time. In expanding a rule based approach to Chinese, we achieved an overall F1 score of 80.00, which is comparable to past results in the CTB. In our neural experiments we achieved F1 scores up to 90.94, 85.38 and 88.79 for English, Chinese, and Korean respectively with functional labels.

cs.CL↗

arXiv2024

Measuring Software Innovation with Open Source Software Development Data

Existing innovation metrics inadequately capture software innovation, creating blind spots for researchers and policymakers seeking to understand and foster technological innovation in an increasingly software-defined economy. This paper introduces a novel measure of software innovation based on open source software (OSS) development activity on GitHub. We examine the dependency growth and release complexity among 350,000 unique releases from 33,000 unique packages across the JavaScript, Python, and Ruby ecosystems over two years post-release. We find that the semantic versioning types of OSS releases exhibit ecosystem-specific and maturity-dependent patterns in predicting one-year dependency growth, with minor releases showing relatively consistent adoption across contexts while major and patch releases vary significantly by ecosystem and package size. In addition, while semantic versioning correlates with the technical complexity of the change-set, complexity itself shows minimal correlation with downstream adoption, suggesting that versioning signals rather than technical change drive dependency growth. Overall, while semantic versioning release information can be used as a unit of innovation in OSS development complementary to common sources for innovation metrics (e.g. scientific publications, patents, and standards), this measure should be weighted by ecosystem culture, package maturity, and release type to accurately capture innovation dynamics. We conclude with a discussion of the theoretical and practical implications of this novel measure of software innovation as well as future research directions.

cs.SE↗

arXiv2024

Improving Performance of Commercially Available AI Products in a Multi-Agent Configuration

In recent years, with the rapid advancement of large language models (LLMs), multi-agent systems have become increasingly more capable of practical application. At the same time, the software development industry has had a number of new AI-powered tools developed that improve the software development lifecycle (SDLC). Academically, much attention has been paid to the role of multi-agent systems to the SDLC. And, while single-agent systems have frequently been examined in real-world applications, we have seen comparatively few real-world examples of publicly available commercial tools working together in a multi-agent system with measurable improvements. In this experiment we test context sharing between Crowdbotics PRD AI, a tool for generating software requirements using AI, and GitHub Copilot, an AI pair-programming tool. By sharing business requirements from PRD AI, we improve the code suggestion capabilities of GitHub Copilot by 13.8% and developer task success rate by 24.5% -- demonstrating a real-world example of commercially-available AI systems working together with improved outcomes.

cs.SE↗

arXiv2024

On Expressive Power of Looped Transformers: Theoretical Analysis and Enhancement via Timestep Encoding

Looped Transformers provide advantages in parameter efficiency, computational capabilities, and generalization for reasoning tasks. However, their expressive power regarding function approximation remains underexplored. In this paper, we establish the approximation rate of Looped Transformers by defining the modulus of continuity for sequence-to-sequence functions. This reveals a limitation specific to the looped architecture. That is, the analysis prompts the incorporation of scaling parameters for each loop, conditioned on timestep encoding. Experiments validate the theoretical results, showing that increasing the number of loops enhances performance, with further gains achieved through the timestep encoding.

cs.LG↗

arXiv2024

Tur[k]ingBench: A Challenge Benchmark for Web Agents

Can advanced multi-modal models effectively tackle complex web-based tasks? Such tasks are often found on crowdsourcing platforms, where crowdworkers engage in challenging micro-tasks within web-based environments. Building on this idea, we present TurkingBench, a benchmark consisting of tasks presented as web pages with textual instructions and multi-modal contexts. Unlike previous approaches that rely on artificially synthesized web pages, our benchmark uses natural HTML pages originally designed for crowdsourcing workers to perform various annotation tasks. Each task's HTML instructions are instantiated with different values derived from crowdsourcing tasks, creating diverse instances. This benchmark includes 32.2K instances spread across 158 tasks. To support the evaluation of TurkingBench, we have developed a framework that links chatbot responses to actions on web pages (e.g., modifying a text box, selecting a radio button). We assess the performance of cutting-edge private and open-source models, including language-only and vision-language models (such as GPT4 and InternVL), on this benchmark. Our results show that while these models outperform random chance, there is still significant room for improvement. We hope that this benchmark will drive progress in the evaluation and development of web-based agents.

cs.AI↗

arXiv2023

Open Data on GitHub: Unlocking the Potential of AI

GitHub is the world's largest platform for collaborative software development, with over 100 million users. GitHub is also used extensively for open data collaboration, hosting more than 800 million open data files, totaling 142 terabytes of data. This study highlights the potential of open data on GitHub and demonstrates how it can accelerate AI research. We analyze the existing landscape of open data on GitHub and the patterns of how users share datasets. Our findings show that GitHub is one of the largest hosts of open data in the world and has experienced an accelerated growth of open data assets over the past four years. By examining the open data landscape on GitHub, we aim to empower users and organizations to leverage existing open datasets and improve their discoverability -- ultimately contributing to the ongoing AI revolution to help address complex societal issues. We release the three datasets that we have collected to support this analysis as open datasets at https://github.com/github/open-data-on-github.

cs.LG↗

arXiv2023

THC: Accelerating Distributed Deep Learning Using Tensor Homomorphic Compression

Deep neural networks (DNNs) are the de facto standard for essential use cases, such as image classification, computer vision, and natural language processing. As DNNs and datasets get larger, they require distributed training on increasingly larger clusters. A main bottleneck is the resulting communication overhead where workers exchange model updates (i.e., gradients) on a per-round basis. To address this bottleneck and accelerate training, a widely-deployed approach is compression. However, previous deployments often apply bi-directional compression schemes by simply using a uni-directional gradient compression scheme in each direction. This results in significant computational overheads at the parameter server and increased compression error, leading to longer training and lower accuracy. We introduce Tensor Homomorphic Compression (THC), a novel bi-directional compression framework that enables the direct aggregation of compressed values and thus eliminating the aforementioned computational overheads. Moreover, THC is compatible with in-network aggregation (INA), which allows for further acceleration. Our evaluation shows that training representative vision and language models with THC reaches target accuracy by 1.40x to 1.47x faster using INA and 1.28x to 1.33x faster using a software PS compared with state-of-the-art systems.

cs.LG↗

arXiv2023

How Good Can 2D Excitonic Solar Cells Be?

Excitonic semiconductors have been a subject of research for photovoltaic applications for many decades. Among them, the organic polymers and small molecules based solar cells have now exceeded 19% power conversion efficiency (PCE). While organic photovoltaics (OPVs) are approaching maturity, the advent of strongly excitonic inorganic semiconductors such as two-dimensional transition metal dichalcogenides (TMDCs) has renewed interest in excitonic solar cells due to their high-optical constants, stable inorganic structure and sub-nm film thicknesses. While several reports have been published on TMDC based PVs, achieving power conversion efficiencies higher than 6% under one-sun AM1.5G illumination has remained challenging. Here, we perform a full optical and electronic analysis of design, structure and performance of monolayer TMDC based, single-junction excitonic PVs. Our computational model with optimized properties predicts a PCE of 9.22% in a superlattice device structure. Our analysis suggests that, while the PCE for 2D excitonic solar cells may be limited to < 10%, a specific power > 100 W g-1 may be achieved with our proposed designs, making them attractive in aerospace, distributed remote sensing, and wearable electronics.

physics.app-ph↗

arXiv2022

Predicting housing prices and analyzing real estate market in the Chicago suburbs using Machine Learning

The pricing of housing properties is determined by a variety of factors. However, post-pandemic markets have experienced volatility in the Chicago suburb area, which have affected house prices greatly. In this study, analysis was done on the Naperville/Bolingbrook real estate market to predict property prices based on these housing attributes through machine learning models, and to evaluate the effectiveness of such models in a volatile market space. Gathering data from Redfin, a real estate website, sales data from 2018 up until the summer season of 2022 were collected for research. By analyzing these sales in this range of time, we can also look at the state of the housing market and identify trends in price. For modeling the data, the models used were linear regression, support vector regression, decision tree regression, random forest regression, and XGBoost regression. To analyze results, comparison was made on the MAE, RMSE, and R-squared values for each model. It was found that the XGBoost model performs the best in predicting house prices despite the additional volatility sponsored by post-pandemic conditions. After modeling, Shapley Values (SHAP) were used to evaluate the weights of the variables in constructing models.

cs.LG↗

arXiv2022

Non-Abelian topological defects and strain mapping in 2D moir\'e materials

We present a general method to analyze the topological nature of the domain boundary connectivity that appeared in relaxed moir\'e superlattice patterns at the interface of 2-dimensional (2D) van der Waals (vdW) materials. At large enough moir\'e lengths, all moir\'e systems relax into commensurated 2D domains separated by networks of dislocation lines. The nodes of the 2D dislocation line network can be considered as vortex-like topological defects. We find that a simple analogy to common topological systems with an $S^1$ order parameter, such as a superconductor or planar ferromagnet, cannot correctly capture the topological nature of these defects. For example, in twisted bilayer graphene, the order parameter space for the relaxed moir\'e system is homotopy equivalent to a punctured torus. Here, the nodes of the 2D dislocation network can be characterized as elements of the fundamental group of the punctured torus, the free group on two generators, endowing these network nodes with non-Abelian properties. Extending this analysis to consider moir\'e patterns generated from any relative strain, we find that antivortices occur in the presence of anisotropic heterostrain, such as shear or anisotropic expansion, while arrays of vortices appear under twist or isotropic expansion between vdW materials. Experimentally, utilizing the dark field imaging capability of transmission electron microscopy (TEM), we demonstrate the existence of vortex and antivortex pair formation in a moir\'e system, caused by competition between different types of heterostrains in the vdW interfaces. We also present a methodology for mapping the underlying heterostrain of a moir\'e structure from experimental TEM data, which provides a quantitative relation between the various components of heterostrain and vortex-antivortex density in moir\'e systems.

cond-mat.mes-hall↗