arXiv subjects

Kaiyuan Wang

Publications and source records attributed to Kaiyuan Wang.

At least 19 recordsLinked to original sources

arXiv2026

Automated Modernization of Machine Learning Engineering Notebooks for Reproducibility

Interactive computational notebooks (e.g., Jupyter notebooks) are widely used in machine learning engineering (MLE) to program and share end-to-end pipelines, from data preparation to model training and evaluation. However, environmental erosion-the rapid evolution of hardware and software ecosystems for machine learning-has rendered many published MLE notebooks non-reproducible in contemporary environments, hindering code reuse and scientific progress. To quantify this gap, we study 12,106 notebooks selected from 75 popular Kaggle competitions: only 26% remain reproducible today. Crucially, we find that environment backporting, i.e., downgrading dependencies to match the submission time, does not improve reproducibility (decreased to 12%) but rather introduces additional failure modes. To address environmental erosion, we design and implement MLEModernizer, an LLM-driven agentic framework that treats the contemporary environment as a fixed constraint and modernizes notebook code to restore reproducibility. MLEModernizer iteratively executes notebooks, collects execution feedback, and applies three types of targeted fixes: error-repair, runtime-reduction, and score-calibration. Evaluated on 8,210 notebooks that are non-reproducible under the baseline environment, MLEModernizer makes 3,292 (40.1%, GPT-5.2) and 3,683 (44.9%, GPT-OSS-120b) notebooks reproducible. MLEModernizer presents a best-effort automated recovery and modernization technique that can improve reproducibility for a subset of notebooks. Practitioners can leverage MLEModernizer to validate, reuse, and maintain MLE artifacts as the hardware and software ecosystems continue to evolve.

cs.SE↗

arXiv2026

Programmable Hybrid Exceptional Points in Passive Scattering Networks

Exceptional points (EPs) have long promised enhanced sensing of physical signals, but have practically been limited by simultaneous enhancement of noise. Aligning an EP's non-analytic response with target perturbations while suppressing noise has, however, remained challenging. Here we show that fully passive, phase-tuned multi-port scattering networks enable scattering EPs with tailored anisotropic response to perturbations. We leverage projection-induced non-unitarity to realize effective non-Hermitian behavior when measuring only a subset of system ports. By formulating EP design in terms of the discriminant of the projected scattering sub-block and its directional derivatives, we give control-counting rules relating the number of programmable link phases to achievable Riemann surface topologies. We demonstrate our framework in a four-port photonic network by designing both an anisotropic EP and a Dirac-type EP with linear splitting along two parametric directions. We further suppress the global thermal drift response of a network-based sensor to a 3/2 power-law scaling while retaining square-root sensitivity to localized signals. Since the effective non-Hermiticity arises purely from port projection, our approach transfers to integrated photonic and microwave meshes, acoustic circuits, and projected metasurfaces, offering a phase-only route to reconfigurable non-Hermitian response and noise-robust EP sensing.

physics.optics↗

arXiv2026

TestEvo-Bench: An Executable and Live Benchmark for Test and Code Co-Evolution

Software tests and code evolve together: a code change should be followed by new or updated tests that record the new software behavior. Yet existing test generation and update benchmarks often isolate the test from the code change, and rely on static metadata that does not verify whether a test is executable or semantically tied to the code change. This makes it difficult to evaluate whether a test automation agent understands how a code change should propagate into the test suite. We introduce TestEvo-Bench, a benchmark of test and code co-evolution tasks mined from software repositories, with two tracks: in test generation, the agent shall write new tests to capture the new software behavior; in test update, the agent shall adapt failing existing tests to the changed software behavior. Each task is anchored to a real commit history and packaged with environment configuration to support execution-grounded metrics such as pass rate, coverage, and mutation score. TestEvo-Bench is also a live benchmark: each task records the timestamp of the test and code changes, and new tasks are periodically mined by our automated pipeline, so evaluation can be restricted to tasks postdating a model's training cutoff to reduce data leakage risk. The current snapshot contains 746 test generation and 509 test update tasks, curated from 59,950 candidate co-evolution records across 152 open-source Java projects. We experiment with four state-of-the-art agents that combine strong harnesses (Claude Code, Gemini CLI, and SWE-Agent) with strong foundation models (Claude Opus 4.7 and Gemini 3.1 Pro). Results show that they achieve up to 77.5% success rate on test generation and 74.6% on test update. However, success rate is materially lower on the most recent benchmark tasks and drops significantly under limited per-task cost.

cs.SE↗

arXiv2026

PhysInOne: Visual Physics Learning and Reasoning in One Suite

We present PhysInOne, a large-scale synthetic dataset addressing the critical scarcity of physically-grounded training data for AI systems. Unlike existing datasets limited to merely hundreds or thousands of examples, PhysInOne provides 2 million videos across 153,810 dynamic 3D scenes, covering 71 basic physical phenomena in mechanics, optics, fluid dynamics, and magnetism. Distinct from previous works, our scenes feature multiobject interactions against complex backgrounds, with comprehensive ground-truth annotations including 3D geometry, semantics, dynamic motion, physical properties, and text descriptions. We demonstrate PhysInOne's efficacy across four emerging applications: physics-aware video generation, long-/short-term future frame prediction, physical property estimation, and motion transfer. Experiments show that fine-tuning foundation models on PhysInOne significantly enhances physical plausibility, while also exposing critical gaps in modeling complex physical dynamics and estimating intrinsic properties. As the largest dataset of its kind, orders of magnitude beyond prior works, PhysInOne establishes a new benchmark for advancing physics-grounded world models in generation, simulation, and embodied AI.

cs.CV↗

arXiv2026

Latent Representations for Control Design with Provable Stability and Safety Guarantees

We initiate a formal study on the use of low-dimensional latent representations of dynamical systems for verifiable control synthesis. Our main goal is to enable the application of verification techniques -- such as Lyapunov or barrier functions -- that might otherwise be computationally prohibitive when applied directly to the full state representation. Towards this goal, we first provide dynamics-aware approximate conjugacy conditions which formalize the notion of reconstruction error necessary for systems analysis. We then utilize our conjugacy conditions to transfer the stability and invariance guarantees of a latent certificate function (e.g., a Lyapunov or barrier function) for a latent space controller back to the original system. Importantly, our analysis contains several important implications for learning latent spaces and dynamics, by highlighting the necessary geometric properties which need to be preserved by the latent space, in addition to providing concrete loss functions for dynamics reconstruction that are directly related to control design. We conclude by demonstrating the applicability of our theory to two case studies: (1) stabilization of a cartpole system, and (2) collision avoidance for a two vehicle system.

eess.SY↗

arXiv2025

Generalized Wigner-Smith theory for perturbations at exceptional and diabolic point degeneracies

Spectral degeneracies, including diabolic (DP) and exceptional (EP) points, exhibit unique sensitivity to external perturbations, enabling powerful control and engineering of wave phenomena. We present a residue-based perturbation theory that quantifies complex resonance splitting of DP and EP type spectral degeneracies using generalized Wigner-Smith operators. We validate our theory using both analytic Hamiltonian models and numerical electromagnetic simulations, demonstrating excellent agreement across a range of cases. Our approach accurately predicts degenerate resonance splitting using only scattering data, offering a powerful framework for precision tuning, inverse design, and practical exploitation of non-Hermitian phenomena.

physics.optics↗

arXiv2025

Fast-Converging and Asymptotic-Preserving DSMC

Improving the efficiency of the direct simulation Monte Carlo (DSMC) method has become increasingly urgent with the rapid development of space exploration. To address this issue, the direct intermittent general synthetic iteration (DIG) scheme has recently been proposed to enable DSMC's rapid and accurate convergence to steady-state solutions, even when the cell size is much larger than the mean free path in near-continuum flow regimes. The first part of the paper is devoted to the mathematical analysis of DIG's fast-converging and asymptotic-preserving properties. Because the Boltzmann equation is analytically intractable, the analysis is conducted using the linearized BGK model. It is found that, in the near continuum flow regime, the DIG method asymptotically recovers the Navier Stokes equations when the cell size is O(1), rather than being constrained by the mean free path. Moreover, after a single cycle of DIG evolution, the deviation from the final steady state solution is reduced by more than a factor of five. In the second part of the paper, the Poiseuille flow and hypersonic flow passing over cylinder are investigated using the DIG scheme, with different time step and cell sizes, thereby demonstrating its efficiency and accuracy in multiscale flow simulation. Specifically, when the Knudsen number is 0.01, the DIG method is found to be faster than the traditional DSMC method by two orders of magnitude. The performance gain becomes even greater at smaller Knudsen numbers. The proposed method holds great potential for engineering applications.

physics.comp-ph↗

arXiv2025

Topological Engineering of High-Order Exceptional Points through Transformation Optics

Exceptional points (EPs) in non-Hermitian photonic systems have attracted considerable research interest due to their singular eigenvalue topology and associated anomalous physical phenomena. These properties enable diverse applications ranging from enhanced quantum metrology to chiral light-matter interactions. Practical implementation of high order EPs in optical platforms however remains fundamentally challenging, requiring precise multi-parameter control that often exceeds conventional design capabilities. This work presents a novel framework for engineering high order EPs through transformation optics (TO) principles, establishing a direct correspondence between mathematical singularities and physically controllable parameters. Our TO-based paradigm addresses critical limitations in conventional Hamiltonian approaches, where abstract parameter spaces lack explicit connections to experimentally accessible degrees of freedom, while simultaneously providing full-field mode solutions. In contrast to prevailing parity-time-symmetric architectures, our methodology eliminates symmetry constraints in EP design, significantly expanding the possibilities in non-Hermitian photonic engineering. The proposed technique enables unprecedented control over EP formation and evolution in nanophotonic systems, offering new pathways for developing topological optical devices with enhanced functionality and robustness.

physics.optics↗

arXiv2025

C2PSA-Enhanced YOLOv11 Architecture: A Novel Approach for Small Target Detection in Cotton Disease Diagnosis

This study presents a deep learning-based optimization of YOLOv11 for cotton disease detection, developing an intelligent monitoring system. Three key challenges are addressed: (1) low precision in early spot detection (35% leakage rate for sub-5mm2 spots), (2) performance degradation in field conditions (25% accuracy drop), and (3) high error rates (34.7%) in multi-disease scenarios. The proposed solutions include: C2PSA module for enhanced small-target feature extraction; Dynamic category weighting to handle sample imbalance; Improved data augmentation via Mosaic-MixUp scaling. Experimental results on a 4,078-image dataset show: mAP50: 0.820 (+8.0% improvement); mAP50-95: 0.705 (+10.5% improvement); Inference speed: 158 FPS. The mobile-deployed system enables real-time disease monitoring and precision treatment in agricultural applications.

cs.CV↗

arXiv2025

SuPerPM: A Surgical Perception Framework Based on Deep Point Matching Learned from Physical Constrained Simulation Data

A major source of endoscopic tissue tracking errors during deformations stems from wrong data association between observed sensor measurements with previously tracked scene. To mitigate this issue, we present a surgical perception framework, SuPerPM, that leverages learning-based non-rigid point cloud matching for data association, thus accommodating larger deformations than previous approaches which relied on Iterative Closest Point (ICP) for point associations. The learning models typically require training data with ground truth point cloud correspondences, which is challenging or even impractical to collect in surgical environments. Thus, for tuning the learning model, we gather endoscopic data of soft tissue being manipulated by a surgical robot and then establish correspondences between point clouds at different time points to serve as ground truth. This was achieved by employing a position-based dynamics (PBD) simulation to ensure that the correspondences adhered to physical constraints. The proposed framework is demonstrated on several challenging surgical datasets that are characterized by large deformations, achieving superior performance over advanced surgical scene tracking algorithms.

cs.CV↗

arXiv2025

General synthetic iterative scheme for multiscale radiative transfer in the finite-volume framework

Achieving efficient and accurate simulation of the radiative transfer has long been a research challenge. Here we introduce the general synthetic iterative scheme as an easy-to-implement approach to address this issue. First, a macroscopic synthetic equation, which combines the asymptotic equation at the diffusion limit and the "high-order terms" extracted from the transport equation to account for transport effects, is introduced to accelerate the simulation of the radiative transfer equation. Second, the asymptotic preserving property is directly provided by the macroscopic process, eliminating the need for fine spatial discretization in optically thick media, as well as the need for consistency enforcement. Third, to address the issue of opacity discontinuity in the finite volume method, an adaptive least square method for gradient approximation is proposed. Numerical results on several canonical tests demonstrate that, in optically thick problems, our method achieves significant speed-up over the conventional iterative schemes. Finally, with our newly developed method, we reveal the importance of resolving the Knudsen layer in the initial stage of Tophat problem, while in steady-state the Knudsen layer can be under-resolved.

physics.comp-ph↗

arXiv2025

Fully Tunable On-Chip Meta-Generator for Multidimensional Poincaré Sphere mapping

The angular momentum of light can be elegantly mapped onto high-order Poincare spheres, providing a powerful framework for describing structured light beams. Such beams have shown extraordinary potential across diverse applications, including high-capacity optical communications, precision metrology, and quantum information processing. While various methods exist for generating structured light beams, the dynamic synthesis and flexible control of arbitrary vectorial states on diverse, multidimensional Poincare spheres still rely on bulky free-space optical components, posing significant challenges for scalability and integration. To date, a fully tunable solution implemented on a single photonic chip has yet to be realized. Here, we present the first fully tunable on-chip meta-generator capable of dynamically mapping arbitrary scalar, vectorial, and hybrid modes onto the full hierarchy of Poincare spheres, and even extending to a high-dimensional Poincare hypersphere within a four-dimensional Hilbert space. Our device is implemented on an eight-channel space-multiplexed multimode silicon photonic integrated circuit, where densely integrated mode multiplexers, amplitude-phase modulators, and an inverse-designed multimode meta-waveguide together enable compact, precise, and programmable control of structured light. The multimode meta-waveguide directly maps eight on-chip guided modes to orbital angular momentum (OAM), supporting broadband generation of high-purity OAM modes with diverse polarization states and topological charges. By simultaneously engineering amplitude, phase, polarization, and topological charge, we achieve full-field control over OAM mode bases, enabling fully tunable access to arbitrary scalar and vectorial states across more than eight distinct Poincare spheres.

physics.optics↗

arXiv2025

DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset

The creation of large, diverse, high-quality robot manipulation datasets is an important stepping stone on the path toward more capable and robust robotic manipulation policies. However, creating such datasets is challenging: collecting robot manipulation data in diverse environments poses logistical and safety challenges and requires substantial investments in hardware and human labour. As a result, even the most general robot manipulation policies today are mostly trained on data collected in a small number of environments with limited scene and task diversity. In this work, we introduce DROID (Distributed Robot Interaction Dataset), a diverse robot manipulation dataset with 76k demonstration trajectories or 350 hours of interaction data, collected across 564 scenes and 84 tasks by 50 data collectors in North America, Asia, and Europe over the course of 12 months. We demonstrate that training with DROID leads to policies with higher performance and improved generalization ability. We open source the full dataset, policy learning code, and a detailed guide for reproducing our robot hardware setup.

cs.RO↗

arXiv2025

Scalable Machine Learning Training Infrastructure for Online Ads Recommendation and Auction Scoring Modeling at Google

Large-scale Ads recommendation and auction scoring models at Google scale demand immense computational resources. While specialized hardware like TPUs have improved linear algebra computations, bottlenecks persist in large-scale systems. This paper proposes solutions for three critical challenges that must be addressed for efficient end-to-end execution in a widely used production infrastructure: (1) Input Generation and Ingestion Pipeline: Efficiently transforming raw features (e.g., "search query") into numerical inputs and streaming them to TPUs; (2) Large Embedding Tables: Optimizing conversion of sparse features into dense floating-point vectors for neural network consumption; (3) Interruptions and Error Handling: Minimizing resource wastage in large-scale shared datacenters. To tackle these challenges, we propose a shared input generation technique to reduce computational load of input generation by amortizing costs across many models. Furthermore, we propose partitioning, pipelining, and RPC (Remote Procedure Call) coalescing software techniques to optimize embedding operations. To maintain efficiency at scale, we describe novel preemption notice and training hold mechanisms that minimize resource wastage, and ensure prompt error resolution. These techniques have demonstrated significant improvement in Google production, achieving a 116% performance boost and an 18% reduction in training costs across representative models.

cs.DC↗

arXiv2024

Efficient Incremental Code Coverage Analysis for Regression Test Suites

Code coverage analysis has been widely adopted in the continuous integration of open-source and industry software repositories to monitor the adequacy of regression test suites. However, computing code coverage can be costly, introducing significant overhead during test execution. Plus, re-collecting code coverage for the entire test suite is usually unnecessary when only a part of the coverage data is affected by code changes. While regression test selection (RTS) techniques exist to select a subset of tests whose behaviors may be affected by code changes, they are not compatible with code coverage analysis techniques -- that is, simply executing RTS-selected tests leads to incorrect code coverage results. In this paper, we present the first incremental code coverage analysis technique, which speeds up code coverage analysis by executing a minimal subset of tests to update the coverage data affected by code changes. We implement our technique in a tool dubbed iJaCoCo, which builds on Ekstazi and JaCoCo -- the state-of-the-art RTS and code coverage analysis tools for Java. We evaluate iJaCoCo on 1,122 versions from 22 open-source repositories and show that iJaCoCo can speed up code coverage analysis time by an average of 1.86x and up to 8.20x compared to JaCoCo.

cs.SE↗

arXiv2024

Inverse-designed Photonic Computing Core for Parallel Matrix-vector Multiplication

On-chip optical neural networks (ONNs) have recently emerged as an attractive hardware accelerator for deep learning applications, characterized by high computing density, low latency, and compact size. As these networks rely heavily on massive matrix multiplication, photonic computing cores for matrix computation become crucial components for on-chip ONNs, which harness the degree of freedoms (DOFs) in photonics including space, wavelength and mode dimensions. However, previous photonic computing devices have not fully utilized the orthogonality and the conversion characteristic of the waveguide modes, which as we show here, allows for the simultaneous parallel computing of several independent matrix-vector multiplications within the same device. In this work, we propose an inverse-designed photonic computing core for parallel matrix-vector multiplication. The matrices are implemented through a mode conversion process, where the input fundamental modes are simultaneously converted into several orthogonal output modes. Specifically, we target the complex-valued conversion matrices between input and output modes and inversely design the dielectric distribution within the device to achieve parallel matrix-vector multiplication. As a demonstration, the proposed photonic computing core supports simultaneous parallel computing of two independent matrix-vector multiplications, with an ultra-compact footprint and high computing precision (relative error < 8%) at 1550 nm wavelength. The inverse-designed photonic computing devices hold great potential for high-performance on-chip ONNs with low energy consumption and high computing density.

physics.optics↗

arXiv2024

FinLLMs: A Framework for Financial Reasoning Dataset Generation with Large Language Models

Large Language models (LLMs) usually rely on extensive training datasets. In the financial domain, creating numerical reasoning datasets that include a mix of tables and long text often involves substantial manual annotation expenses. To address the limited data resources and reduce the annotation cost, we introduce FinLLMs, a method for generating financial question-answering data based on common financial formulas using Large Language Models. First, we compile a list of common financial formulas and construct a graph based on the variables these formulas employ. We then augment the formula set by combining those that share identical variables as new elements. Specifically, we explore formulas obtained by manual annotation and merge those formulas with shared variables by traversing the constructed graph. Finally, utilizing GPT-3.5, we generate financial question-answering data that encompasses both tabular information and long textual content, building on the collected formula set. Our experiments demonstrate that synthetic data generated by FinLLMs effectively enhances the performance of several large-scale numerical reasoning models in the financial domain, outperforming two established benchmark financial question-answering datasets.

cs.AI↗

arXiv2023

AnyOKP: One-Shot and Instance-Aware Object Keypoint Extraction with Pretrained ViT

Towards flexible object-centric visual perception, we propose a one-shot instance-aware object keypoint (OKP) extraction approach, AnyOKP, which leverages the powerful representation ability of pretrained vision transformer (ViT), and can obtain keypoints on multiple object instances of arbitrary category after learning from a support image. An off-the-shelf petrained ViT is directly deployed for generalizable and transferable feature extraction, which is followed by training-free feature enhancement. The best-prototype pairs (BPPs) are searched for in support and query images based on appearance similarity, to yield instance-unaware candidate keypoints.Then, the entire graph with all candidate keypoints as vertices are divided to sub-graphs according to the feature distributions on the graph edges. Finally, each sub-graph represents an object instance. AnyOKP is evaluated on real object images collected with the cameras of a robot arm, a mobile robot, and a surgical robot, which not only demonstrates the cross-category flexibility and instance awareness, but also show remarkable robustness to domain shift and viewpoint change.

cs.CV↗