Searcharxiv⌕ Search

arXiv subjects

Jian Cao

Publications and source records attributed to Jian Cao.

At least 19 recordsLinked to original sources

OverlapAdvisor: Modeling and Adapting Computation-Communication Overlap for Distributed MoE Serving

Mixture-of-Experts (MoE) inference with expert parallelism (EP) relies on All-to-All communication, making computation-communication overlap strategies critical for achieving high performance. However, no strategy is universally optimal. Across 103 production configurations, varying input lengths, and four hardware platforms, we observe that the fastest strategy shifts based on model shape, EP degree, input length, and hardware. Since no strategy dominates more than 80.6% of configurations, configuration-aware selection is necessary. In this work, we propose OverlapAdvisor, which models overlap stages using two granularities to capture computation-communication dependencies. OverlapAdvisor builds a unified performance model, accounting for deviations from peak compute, memory, and communication performance across workloads and hardware. This model enables launch-time, configuration-aware strategy selection within two serving engines. We conduct experiments on four MoE models (30B-235B) against three baselines, utilizing SGLang and vLLM. At the kernel level, OverlapAdvisor achieves 1.16x average and 1.59x maximum speedups, with a prediction MAPE of 3.3%-15.9%. In SGLang and vLLM, it improves end-to-end throughput up to 31.7%. Its granularity-guided optimizations further achieve median speedups of 1.10x for DeepEP-v1 and 1.22x for FlashMoE.

cs.DC↗

LogLLM: Log-based Anomaly Detection Using Large Language Models

Software systems often record important runtime information in logs to help with troubleshooting. Log-based anomaly detection has become a key research area that aims to identify system issues through log data, ultimately enhancing the reliability of software systems. Existing methods often fall short in capturing the semantic information, typically expressed in natural language, or the sequential dependencies inherent in log sequences. In this paper, we propose LogLLM, a framework that enables collaboration between heterogeneous LLMs for log-based anomaly detection. LogLLM exploits the complementary capabilities of different LLM architectures: a Transformer encoder-based LLM is employed to extract fine-grained semantic vectors from individual log messages, while a Transformer decoder-based LLM is utilized to model sequential dependencies and generate anomaly detection decisions. To enable effective collaboration between these heterogeneous LLMs, we introduce a learnable projector to align their vector representation spaces. Furthermore, we design a progressive three-stage training strategy to optimize the collaboration between heterogeneous LLMs by gradually aligning their representations and adapting them to log anomaly detection. Unlike conventional methods that require log parsers to extract templates, LogLLM preprocesses log messages with regular expressions, streamlining the entire process. Experimental results on four public real-world datasets demonstrate that LogLLM outperforms state-of-the-art methods, achieving an average F$_1$-score improvement of 6.6% over the strongest existing approach. Further analyses provide insights into the effectiveness of the key components of the model architecture and progressive three-stage training strategy.

cs.SE↗

Scalable generative modeling of non-Gaussian spatio-temporal fields via autoregressive Gaussian processes

Generative modeling of spatio-temporal fields is crucial for a variety of applications, including stochastic weather generators and climate-model surrogates. However, many such fields exhibit complex dependence structures that vary across space and time and are nonlinear, resulting in nonstationary and non-Gaussian joint distributions. Our approach represents the joint density of a spatio-temporal field as a product of univariate conditional distributions and models these conditionals using Gaussian processes within an autoregressive transport-map construction. This prior distribution provides regularization, making our method suitable for a small number of training samples. Data-dependent sparsity in the conditioning sets ensures scalability to high-dimensional distributions. We also propose a variant of the method designed to sample or predict forward in time from a given incomplete space-time trajectory. We demonstrate the accuracy and scalability of our approach on non-Gaussian climate-model output with tens of millions of data points.

stat.ME↗

Tracing the Evolution of Oracle Bone Characters Across Three Millennia

Of the approximately 4,500 Oracle Bone Inscription (OBI) characters discovered from the Shang dynasty, only about 1,600 have been deciphered. Many computational approaches compare OBI with glyphs from one historical period at a time. However, during the evolution of Chinese characters, significant structural or semantic changes often occur in uncertain dynasties. A single-period reference may be insufficient when relevant forms change substantially between observed eras. Therefore, we propose the \textbf{Manifold-based Script Evolution Framework (MSEF)}, a framework that models the evolution series (OBI, Bronze, Seal, Clerical, Regular) of Chinese characters as the continual evolution of a manifold space. MSEF represents each character as an era-specific manifold point and learns continuous inter-era transition rules via Neural Ordinary Differential Equations. Both manifold space and transition dynamics can be trained end-to-end through character evolution pairs across any two eras.

cs.CL↗

EDGC: Entropy-driven Dynamic Gradient Compression for Efficient LLM Training

Training large language models (LLMs) at scale incurs substantial communication overhead, while static gradient compression cannot adapt to gradient evolution and may degrade model quality. We propose EDGC, an entropy-driven dynamic gradient compression framework that adapts compression ranks to gradient entropy during training. EDGC combines efficient entropy estimation through gradient sampling, a theoretical model relating entropy to compression rank under a bounded-error constraint, and window-based rank adjustment across pipeline stages. Experiments on 32-V100 and 64-H100 GPU clusters training GPT2 models with 2.5B and 12.1B parameters show that EDGC reduces communication latency by up to 46.45% and end-to-end training time by 16.13%, while maintaining model quality.

cs.LG↗

STEP-LLM: Generating CAD STEP Models from Natural Language with Large Language Models

Computer-aided design (CAD) is vital to modern manufacturing, yet model creation remains labor-intensive and expertise-heavy. To enable non-experts to translate intuitive design intent into manufacturable artifacts, recent large language models-based text-to-CAD efforts focus on command sequences or script-based formats like CadQuery. However, these formats are kernel-dependent and lack universality for manufacturing. In contrast, the Standard for the Exchange of Product Data (STEP, ISO 10303) file is a widely adopted, neutral boundary representation (B-rep) format directly compatible with manufacturing, but its graph-structured, cross-referenced nature poses unique challenges for auto-regressive LLMs. To address this, we curate a dataset of ~40K STEP-caption pairs and introduce novel preprocessing tailored for the graph-structured format of STEP, including a depth-first search-based reserialization that linearizes cross-references while preserving locality and chain-of-thought(CoT)-style structural annotations that guide global coherence. We integrate retrieval-augmented generation to ground predictions in relevant examples for supervised fine-tuning, and refine generation quality through reinforcement learning with a specific Chamfer Distance-based geometric reward. Experiments demonstrate consistent gains of our STEP-LLM in geometric fidelity over the Text2CAD baseline, with improvements arising from multiple stages of our framework: the RAG module substantially enhances completeness and renderability, the DFS-based reserialization strengthens overall accuracy, and the RL further reduces geometric discrepancy. Both metrics and visual comparisons confirm that STEP-LLM generates shapes with higher fidelity than Text2CAD. These results show the feasibility of LLM-driven STEP model generation from natural language, showing its potential to democratize CAD design for manufacturing.

cs.AI↗

ALP photo-production in the parton and hadron cascade model PACIAE

In this work, the parton and hadron cascade model PACIAE is employed to simulate the events of Au+Au collisions at $\sqrt{s_{NN}}=$200 GeV within the 60-80 \% centrality interval. Based on the abundant photons in final hadronic state, the statistical-mechanics-based coalescence model DCPC is applied to phenomenologically recombine post-freeze-out photon pairs into axion-like particles (ALPs) for the first time. The coalescence probability is constrained by the branching-ratio $B_r(a \rightarrow γγ)=0.57$ under the threshold of twice of $K$ meson mass. The ALP observables including yields, transverse-momentum distributions and rapidity spectra are calculated for several benchmark ALP-mass bins. These theoretical predictions are worthy to be investigated by the experiments.

hep-ph↗

GLOW: Graph-Language Co-Encoding for Agentic Workflow Performance Prediction

Agentic Workflows (AWs) have emerged as a promising paradigm for solving complex tasks. However, automatically generating high-quality AWs remains expensive because AW optimization requires evaluating a large number of candidate AWs via execution, resulting in high computational cost and latency. Recently, AW performance prediction has become a hot research topic to avoid costly execution-based evaluation, but existing methods primarily use Graph Neural Networks (GNNs) to model workflow structures and insufficiently capture the semantic relationships among agents. To address this limitation, we propose GLOW, a unified framework for AW performance prediction that combines the graph-structure modeling ability of GNNs with the topology-aware semantic encoding capability of LLMs. Specifically, a graph-oriented LLM is first built through instruction-tuning on graph understanding tasks to extract topology-aware semantic representations from descriptive text of AWs. Meanwhile, a GNN explicitly models the structural information of AWs and produces corresponding structural representations. The semantic and structural representations are then fused in a shared latent space using a Transformer-based fusion module. A contrastive learning strategy is further introduced to learn more discriminative representations for AWs. Experiments on the FLORA-Bench benchmark demonstrate that GLOW consistently outperforms state-of-the-art baselines in both prediction accuracy and ranking utility. Moreover, when integrated into the AFLOW, an automatic AW generation framework, GLOW reduces optimization time by 98.7% with only a 0.031 average score decrease across three datasets, showing its effectiveness as an efficient surrogate evaluator for AW optimization.

cs.LG↗

Fast 360-Degree 3D Metrology for Directed Energy Deposition

Directed Energy Deposition (DED) is a metal additive manufacturing process capable of building and repairing large, complex metal parts from a wide range of alloys. Its flexibility makes it attractive for multiple industrial applications, e.g., in aerospace, automotive and biomedical fields. However, errors or defects introduced at any stage of the printing process can, if undetected, significantly impact the final result, rendering the printed part unusable. Potential in-situ correction methods of printing defects require fast and high-resolution on-the-fly 3D inspection inside the machine, but existing 3D monitoring methods often lack full 360° 3D coverage, require bulky setups, or are too slow for real-time layer-wise feedback. In this paper, we present a single-shot, multi-view polarized fringe projection profilometry (FPP) system designed for real-time in-situ 3D inspection during DED printing. Multiple camera-projector pairs are arranged around the deposition surface to measure depth from different viewpoints in single-shot, while cross-polarized image filtering suppresses specular reflections caused by varying surface reflectance across different alloys. The final 360° reconstruction is obtained via joint registration of the captured multi-view measurements. Our prototype has been deployed in a DED system and our first experiments demonstrate a depth precision better than $δz < 60\,μ\mathrm{m}$ on partially reflective and "shiny" metal surfaces, enabling accurate, layer-wise monitoring for closed-loop DED control.

physics.optics↗

Self-Correcting Long-Horizon Search Agents via Tree-Structured Memory

Large language model (LLM)-based search agents answer questions through multi-step interactions with external environments. However, providing complete execution trajectories to the LLM causes unbounded context growth and introduces noise. Existing compression methods reduce context at the cost of important details and often replace erroneous facts without repairing downstream reasoning derived from them. To address this problem, we propose ReTree, a self-correcting tree-structured memory mechanism for search agents. ReTree constructs a bounded per-step reasoning context while preserving source-linked evidence. It models search as an evidence tree whose nodes store bounded summaries, evidence, and revision histories. When newly retrieved evidence contradicts an earlier claim, ReTree traces back to the node where the claim was introduced, replaces outdated evidence, regenerates summaries, prunes affected branches, and resumes search. Source-grounded evidence provenance supports reliable conflict localization and keeps final claims traceable to retrieved passages. Experiments on four public question-answering and search benchmarks show that ReTree consistently outperforms Full-Trajectory ReAct, improving answer accuracy by up to 25.6 percentage points (pp); the average maximum per-step reasoning context of Full-Trajectory ReAct is $1.27$--$1.51\times$ that of ReTree. These results establish ReTree as an effective self-correcting memory abstraction for long-horizon search.

cs.AI↗

Continue or Replan? Bernoulli-Continuation Policy Learning for Adaptive Horizon Execution

Existing chunk-based Vision-Language-Action (VLA) models execute a fixed number of actions (i.e., execution horizon) before replanning, turning replanning into a task-agnostic periodic schedule that is independent of task progress. As a result, when no replanning boundary falls before a critical manipulation stage, it is executed from a stale chunk rather than a freshly replanned one. To address this limitation, we propose Bernoulli-Continuation Policy (BCP), a lightweight, plug-and-play framework for adaptive horizon execution that keeps the base VLA frozen. Given a fixed-length action chunk, its continuation head decomposes execution-horizon selection into a sequence of continue-or-replan decisions, which imposes an ordinal, prefix-sharing inductive bias over candidate horizons rather than treating them as independent classes. Since the optimal horizon for each chunk is not observable, we train this head with reinforcement learning from trajectory-level outcomes and introduce a Replanning-Efficiency Reward that jointly rewards task success and efficient VLA usage, discouraging the policy from collapsing to unnecessarily short horizons. On RoboTwin 2.0 with LingBot-VLA as the base policy, BCP improves the average success rate by +11.08% on 13 low-success tasks and from 89.88% to 93.94% (+4.06%) across all 50 tasks. Although trained only under the Clean setting, BCP generalizes to the Randomized setting, raising the average success rate by +4.06%. It also transfers to a different base policy $π_{0.5}$, achieving a better result on LIBERO (+1.7%) and, notably, on the harder LIBERO-PRO (+6.8%). On a real robot, BCP lifts success from 74% to 92% and from 44% to 84% on two manipulation tasks. Meanwhile, its negligible overhead, combined with higher success, makes BCP's overall runtime even lower than the fixed-horizon baselines.

cs.RO↗

Thermo-Structural Topology Optimization Considering Nonlinear Creep

Creep is a primary life-limiting mechanism for metallic components operating at high temperature, producing permanent deformation under sustained loads even when stresses remain below yield. The design of structures to minimize this deformation is critical to extending the service life of components. Incorporating creep into topology optimization (TO) remains open because the response is nonlinear, history-dependent, and thermomechanically coupled, and prior work often relies on linear viscoelastic models, which do not capture the behavior of metals at high temperatures. To bridge this gap, we introduce a differentiable thermo-structural TO framework. The approach considers creep deformation using the Norton model and leverages JAX's automatic differentiation to perform adjoint sensitivity analysis, enabling efficient gradient-based optimization. The transient material response is solved via a backward Euler scheme over a prescribed service life. Our objective is to minimize creep deformation subject to a volume constraint. We first demonstrate the framework on canonical two-dimensional benchmarks, showing that the proposed formulation significantly reduces permanent deformation compared to designs optimized solely for elastic stiffness. We then pose, as a challenge problem, the compositional design of a three-dimensional graded material turbine blade in which the local mixture of two candidate alloys is optimized. This challenge problem exercises the full capability of the framework, including transient nonlinear creep, coupled thermal loading, three-dimensional geometry, and gradient-based multi-material design, highlighting the need for creep-aware design in high-temperature applications.

cs.CE↗

JAX-FEM-ANISO: Differentiable GPU-Accelerated Finite Element Framework for Inverse Identification of Finite-Strain Anisotropic Plasticity

We present a fully differentiable, GPU-accelerated finite element framework JAX-FEM-ANISO, for forward simulation and inverse parameter identification of finite-strain anisotropic plasticity. Built on JAX-FEM, the framework exploits modern accelerator architectures by parallelizing the three major computational bottlenecks in nonlinear FEM: elemental weak-form and tangent-stiffness evaluation, global sparse matrix assembly, and sparse linear solution. For a large-scale forward problem with 3 million degrees of freedom, JAX-FEM-ANISO on a single NVIDIA H100 GPU achieves up to 9.4$\times$ speed-up over a 24-core CPU Abaqus baseline. Automatic differentiation is applied through the constitutive update and solver workflow, providing consistent Jacobians for complex constitutive models without manual derivation and accurate gradients for PDE-constrained inverse analysis. Compared with finite differences, the JAX-AD gradients avoid step-size sensitivity and provide the required sensitivities at substantially lower computational cost. For inverse characterization, we combine information-rich, topology-optimized heterogeneous specimens with full-field displacement data to identify advanced constitutive model parameters from a single test, replacing what would otherwise require many conventional experiments. We demonstrate accurate recovery of anisotropic yield and hardening parameters in progressively challenging settings, including uniform and spatially varying material properties. The resulting AD-based formulation enables efficient optimization in high-dimensional parameter spaces where finite-difference approaches are computationally infeasible. These results establish differentiable, GPU-accelerated FEM as a practical high-throughput engine for simulation, characterization, and optimization workflows in advanced manufacturing.

cs.CE↗

Shedding light on the nature of $ϕ(2170)$ with the parton and hadron cascade model PACIAE

The nature of $ϕ(2170)$ remains open. We simulate its production in $e^+e^-$ collisions at $\sqrt{s}=4.95$ GeV using PACIAE 4.0, which sequentially generates the final partonic state (FPS) and the final hadronic state (FHS). While previous studies have interpreted $ϕ(2170)$ as an $ss\bar{s}\bar{s}$ or a $u\bar{u}s\bar{s}$ state, the $U(1)$ anomaly coupling allows non-strange quarks to couple to a vector $s\bar{s}$ component via soft-gluon interactions. This motivates us to also explore the $d\bar{d}s\bar{s}$ tetraquark configuration. In addition, we consider $ϕ(2170)$ as an excited strangeonium state, an $s\bar{s}g$ hybrid state, a $\barΛΛ$ bound state, and a $ϕK^+K^-$ resonance state. The strangeonium, hybrid, and tetraquark candidates are formed by coalescing their constituent partons in the FPS using the dynamically constrained phase-space coalescence model. The $\barΛΛ$ and $ϕK^+K^-$ states are produced via recombination of their constituent hadrons in the FHS. We calculate the orbital angular momentum quantum number of each candidate in its rest frame and perform spectral classification. Given $J^{PC}=1^{--}$, $ϕ(2170)$ can be interpreted as a $D$-wave $s\bar{s}$, a $P$-wave $s\bar{s}g$, a $P$-wave $u\bar{u}s\bar{s}/d\bar{d}s\bar{s}/ss\bar{s}\bar{s}$, an $S$-wave $\barΛΛ$, or an $S$-wave $ϕK^+K^-$ state. The yields of the $D$-wave $s\bar{s}$, $P$-wave $s\bar{s}g$, $u\bar{u}s\bar{s}$ and $d\bar{d}s\bar{s}$ states are of order $10^{-4}$; those for the $S$-wave $\barΛΛ$ and $ϕK^+K^-$ states are of order $10^{-5}$; while the $P$-wave $ss\bar{s}\bar{s}$ yield is of order $10^{-6}$. Moreover, significant discrepancies are observed in the rapidity distributions and the $p_T$ spectra among the various candidates. These discrepancies could serve as valuable criteria for unraveling the nature of $ϕ(2170)$.

hep-ph↗

MOSAIC: Mobile Object Segmentation under Adverse Imaging Conditions for Rapid L-PBF Keyhole Behavior Characterization

In laser powder bed fusion (L-PBF) processes, the rapid evolution of gas and fluid interactions complicates our ability to properly monitor or control the process, with unstable keyholes leading to porosity and spatter formation. High-speed operando x-ray imaging of the keyhole has been used to better understand the impact of these interactions on the monitoring and control of the L-PBF process. MOSAIC, a Mobile Object Segmentation algorithm for experiments under Adverse Imaging Conditions, is designed to perform rapid analysis of keyhole dynamics during active beamline experimentation without needing time consuming manual labeling or model training. Validation studies performed on 12 unique samples proved the robustness of MOSAIC with an average F1 score of 0.894 and a precision of 0.953 when compared to manually segmented images, performing equally or better than the SAM and YOLO machine learning methods tested. MOSAIC is efficient, processing frames cropped to a moving window approximately 150x250 pixels at 19.9 milliseconds per image on CPU, compared to 54 and 5284 milliseconds per image for inference on CPU for YOLO and SAM models.

eess.IV↗

Permutation-preserving Functions and Neural Vecchia Covariance Kernels

We introduce a novel framework for constructing scalable and flexible covariance kernels for Gaussian processes (GPs) by directly learning the covariance structure under a regression-type parameterization induced by Vecchia approximations, using deep neural architectures. Specifically, we model kriging coefficients and conditional standard deviations, deterministic quantities that uniquely characterize the covariance, providing stable and informative learning targets. Exploiting the permutation-equivariant structure of conditioning sets in the Vecchia factorization, we derive a universal representation for permutation-preserving functions and design neural architectures that respect this symmetry, leading to improved training stability and data efficiency. The proposed approach enables expressive, non-stationary kernel learning while maintaining computational scalability, thereby bridging classical GP methodology with modern deep learning.

stat.ML↗

PipeWeave: Synergizing Analytical and Learning Models for Unified GPU Performance Prediction

The rapid expansion of Transformer-based large language models has dramatically increased the need for high-performance GPUs. As a result, there is growing demand for fast, accurate, and widely generalizable GPU performance models to support next-generation hardware selection and system-level exploration. However, current data-driven methods are limited, exhibiting poor generalization across hardware and inadequate modeling of complex production-level kernels common in modern inference stacks. To address these issues, we present PipeWeave, a unified GPU modeling framework. This approach first employs an analytical model to quantify a given kernel's demands on the GPU's heterogeneous instruction pipelines. These analytical features are then fed into a machine learning (ML) model to capture complex cross-pipeline interactions and resource dependencies, enabling high-fidelity performance prediction. Our evaluation across 11 GPU types from four generations of major architectures on two widely-used serving systems demonstrates that PipeWeave delivers high fidelity and strong generalizability. It achieves accurate predictions, with only 6.1% average error at the kernel level and 8.5% for end-to-end inference -- reducing the error of state-of-the-art methods by 6.7x and 4.4x, respectively. We also demonstrate PipeWeave's value "beyond simulation" by utilizing its performance ceiling to diagnose implementation shortcomings and guide the optimization of a production fused MoE Triton kernel, achieving up to 1.7x speedup. Code is available https://github.com/zksainx/pipeweave.

cs.PF↗

A Neural-Mean Vecchia Gaussian Process for Unified Argo Modeling

Argo is an international program that collects temperature and salinity observations in the upper two kilometers of the global ocean. Most existing approaches for modeling Argo temperature rely on localized modeling within moving windows, first estimating a prescribed mean structure and then fitting Gaussian processes (GPs) to the mean-subtracted anomalies. Such strategies introduce challenges in designing suitable mean structures and defining local moving windows, often resulting in case-specific modeling choices. In this work, we propose a one-stop Gaussian process regression framework with a flexible mean structure and a generic spatio-temporal covariance function to jointly model Argo temperature data across broad spatial domains. Our fully data-driven approach achieves predictive performance that compares favorably with the established benchmarks that require moving-window regression and separate parametric mean estimation. To ensure scalability over large spatial regions, we employ the Vecchia approximation, which reduces the computational complexity from cubic to quasi-linear in the number of observations while preserving predictive accuracy. Using Argo data from January to March over the years 2007-2016, the same dataset used in prior benchmark studies, we demonstrate that our approach provides a unified and data-driven alternative for large-scale oceanographic analysis.

stat.CO↗