arXiv subjects

Zhouteng Ye

Publications and source records attributed to Zhouteng Ye.

4 recordsLinked to original sources

arXiv2026

SkillSpec: Intent-Masked Specification Reasoning for Agent Skill Correctness

Autonomous agent systems increasingly depend on reusable skill abstractions for consolidating experiential knowledge and domain expertise. These artifacts typically bundle free-form instructions with heterogeneous resources. However, ensuring their correctness remains challenging. Their failure modes transcend conventional code defects to subtle semantic inconsistencies such as intent conflicts, which manifest as silent failures masked by the underlying model. Moreover, skill correctness must be grounded in intended task boundaries and generalizability. We propose SkillSpec, a Hoare-style framework that formulates skill correctness as a specification reasoning problem. It transforms a heterogeneous skill repository into a unified graph representation that aligns descriptions, instructions and code artifacts. For each node, SkillSpec derives an ExpectSpec from the surrounding declared intent, and infers FactSpecs from encoded behavior under partially disclosed intent. An intent mask regulates access to holistic, lineage, neighborhood, and local views to balance the bias introduced by excessive context against unsupported inference caused by insufficient context. SkillSpec jointly reasons over these views to flag candidate defects, and automatically validates them in an isolated sandbox. On 515 real-world skills from SkillsBench and widely downloaded repositories, SkillSpec identified 763 manually confirmed defects across 239 skills, achieving 61.2% precision. The node-level analysis across multiple model families shows that specification reasoning is consistently reliable for code nodes, whereas plain-text nodes remain a major bottleneck. Most defects arise at the boundaries between declared intent and implementation, demonstrating that explicit specifications provide a practical foundation for skill quality assurance in real-world agent ecosystems.

cs.SE↗

arXiv2025

LaPON: A Lagrange's-mean-value-theorem-inspired operator network for solving PDEs and its application on NSE

Accelerating the solution of nonlinear partial differential equations (PDEs) while maintaining accuracy at coarse spatiotemporal resolution remains a key challenge in scientific computing. Physics-informed machine learning (ML) methods such as Physics-Informed Neural Networks (PINNs) introduce prior knowledge through loss functions to ensure physical consistency, but their "soft constraints" are usually not strictly satisfied. Here, we propose LaPON, an operator network inspired by the Lagrange's mean value theorem, which embeds prior knowledge directly into the neural network architecture instead of the loss function, making the neural network naturally satisfy the given constraints. This is a hybrid framework that combines neural operators with traditional numerical methods, where neural operators are used to compensate for the effect of discretization errors on the analytical scale in under-resolution simulations. As evaluated on turbulence problem modeled by the Navier-Stokes equations (NSE), the multiple time step extrapolation accuracy and stability of LaPON exceed the direct numerical simulation baseline at 8x coarser grids and 8x larger time steps, while achieving a vorticity correlation of more than 0.98 with the ground truth. It is worth noting that the model can be well generalized to unseen flow states, such as turbulence with different forcing, without retraining. In addition, with the same training data, LaPON's comprehensive metrics on the out-of-distribution test set are at least approximately twice as good as two popular ML baseline methods. By combining numerical computing with machine learning, LaPON provides a scalable and reliable solution for high-fidelity fluid dynamics simulation, showing the potential for wide application in fields such as weather forecasting and engineering design.

physics.comp-ph↗

arXiv2021

a Decision-Tree based Moment-of-Fluid (DTMOF) Method in 3D rectangular hexahedrons

The moment-of-fluid (MOF) method is an extension of the volume-of-fluid method with piecewise linear interface construction (VOF-PLIC). By minimizing the least square error of the centroid of the cutting polyhedron, the MOF method reconstructs the linear interface without using any neighboring information. Traditional MOF involves iteration while finding the optimized linear reconstruction. Here, we propose an alternative approach based on a machine learning algorithm: Decision Tree algorithm. A training data set is generated from a list of random cuts of a unit cube by plane. The Decision Tree algorithm extracts the input-output relationship from the training data, so that the resulting function determines the normal vector of the reconstruction plane directly, without any iteration. The present method is tested on a range of popular interface advection test problems. Numerical results show that our approach is much faster than the iteration-based MOF method while provides compatible accuracy with the conventional MOF method.

math.NA↗

arXiv2020

Some improvements on Moment-of-Fluid method in 3D rectangular hexahedrons

The moment-of-fluid method (MOF) is an extension of the volume-of-fluid method with piecewise linear interface construction (VOF-PLIC). In MOF reconstruction, the optimized normal vector is determined from the reference centroid and the volume fraction by iteration. The state-of-art work by \citet{milcent_moment--fluid_2020} proposed an analytic gradient of the objective function, which greatly reduces the computational cost. In this study, we further accelerate the MOF reconstruction algorithm by using Gauss-Newton iteration instead of Broyden-Fletcher-Goldfarb-Shanno (BFGS) iteration. We also propose an improved initial guess for MOF reconstruction, which improves the efficiency and the robustness of the MOF reconstruction algorithm. Our implementation of the code and test cases are available on our Github repository.

physics.comp-ph↗