arXiv subjects

Xinyu Shi

Publications and source records attributed to Xinyu Shi.

At least 19 recordsLinked to original sources

arXiv2026

Bridging Behavior and Implementation: Automated Java Glue Code Generation for Behavior-Driven Development

Behavior-Driven Development (BDD) helps technical and non-technical stakeholders share a common understanding of software requirements through natural-language scenarios. Glue code makes these scenarios executable by mapping each step to the corresponding project code. However, developing and maintaining glue code requires knowledge of both the intended behavior and the underlying codebase, making it a labor-intensive part of BDD as requirements evolve. Although large language models (LLMs) have shown strong code generation capabilities, their use for automated glue code generation remains unexplored. This task requires reasoning over underspecified behavior, related BDD artifacts, and large project codebases. We present AutoGlue, a hierarchical multi-agent framework for automated Java glue code generation. AutoGlue follows a behavior-first workflow that separates behavior interpretation, context retrieval, and code generation. A Behavior Interpreter derives the intent of a step from its scenario context, while a Developer agent retrieves relevant BDD artifacts and project code before generating the final glue code. We evaluate AutoGlue on 1,307 steps from eight open-source Java projects. Compared with few-shot prompting, AutoGlue improves API F1 by 58.7% and CodeBLEU by 43.7%. It produces directly usable glue code for 46.1% of the evaluated steps, while most partially correct outputs require only minor revisions, such as adding missing actions or refining parameters. Ablation results show that behavior interpretation and project-aware context retrieval both contribute substantially to generation quality. These findings demonstrate that LLMs can effectively connect natural-language behavior specifications with project code and support specification-driven software development.

cs.SE↗

arXiv2026

Hybrid functional calculation of electrical activity and complexing mechanism of Cu-related defects

Copper is a detrimental impurity in silicon with high diffusivity and a high tendency to precipitate. Interaction between Cu and other defects is essential for understanding the nature of Cu precipitation in silicon. Despite extensive experimental investigations of Cu-related defects in silicon, a comprehensive understanding remains elusive due to limitations of techniques in resolving defect configurations, as well as inconsistencies between theoretical and experimental results regarding transition levels. Moreover, the underlying formation mechanism of the well-known $\mathrm{Cu_{PL}}$ line is still unclear. In this work, configurations, formation energies, and transition levels of Cu-related defects in silicon are calculated using the HSE06 functional and finite-size correction. Defects involved in this study include $\mathrm{Cu_i}$, $\mathrm{Cu_{Si}}$, Cu-B, Cu-P, and Cu-H. A $\mathrm{Cu_{i4}V}$ model is proposed to explain the discrepancies between theory and experiment about $\mathrm{Cu_{PL}}$ defect. Our calculations may provide insight into the electrically active defects and the early states of Cu precipitation in silicon.

cond-mat.mtrl-sci↗

arXiv2026

FlexVector: A SpMM Vector Processor with Flexible VRF for GCNs on Varying-Sparsity Graphs

Graph Convolutional Networks (GCNs) are widely adopted for tasks involving relational or graph-structured data and can be formulated as two-stage sparse-dense matrix multiplication (SpMM) during inference. However, existing accelerators often struggle with the irregular workloads induced by power-law node degree distributions. In this work, we propose FlexVector, a vector-processor-based architecture that efficiently accelerates SpMM for GCN inference. To address irregular computation patterns, FlexVector adopts a row-wise, product-based dataflow that regularizes SpMM execution and exposes vector parallelism through full-row access to vector registers, eliminating the need for multi-banked register file designs. Building on this dataflow, it introduces software-managed, flexible vector register files (VRFs) that adapt to irregular data access patterns, without sacrificing memory access efficiency. To further exploit these architectural capabilities, we develop a graph-aware preprocessing and node partitioning strategy that restructures irregular graph workloads to better match the row-wise dataflow and VRF capacity. This hardware-software co-design reduces memory traffic, leading to significant performance and energy efficiency gains on real-world GCN workloads. Experimental results on five real-world GCN datasets show that the VRF-centric FlexVector achieves a 3.78x speedup and 40.5% lower energy at comparable area cost relative to a state-of-the-art cache-centric baseline with buffers of the same size.

cs.DC↗

arXiv2026

Cascade of Spin Liquids in a Bilayer Triangular-lattice Antiferromagnet Rb_2Co_2(SeO_3)_3

In frustrated Ising magnets, classical spin liquids (CSLs) with macroscopic ground-state degeneracy can survive against conventional magnetic order, as exemplified by systems on triangular, kagome and pyrochlore lattices at zero field. Here we report the discovery of a high-field route toward spin liquids in a bilayer triangular lattice antiferromagnet, Rb$_2$Co$_2$(SeO$_3$)$_3$. We demonstrate that a cascade of CSLs -- characterized by doubly degenerate one-up-one-down local spin configurations and a residual entropy of 1/2(1-M/M_s)Rln2 per mole -- emerges through field-controlled dilution of Ising dimers. Owing to the interplay of intra- and inter-layer interactions, these CSLs are further stabilized by lattice symmetry breaking at fractional magnetization plateaus. Such field-induced spin liquids can be understood as a consequence of generalized ice rules, analogous to those governing in pyrochlore antiferromagnets. In particular, the 5/6-plateau state is a candidate quantum spin liquid. Our results thereby establish a new pathway for exploring diverse spin liquid states across both classical and quantum regimes.

cond-mat.str-el↗

arXiv2026

Notational Animating: An Interactive Approach to Creating and Editing Animation Keyframes

We introduce the concept of notational animating, an interaction paradigm for animation authoring where users sketch high-level notations over static drawings to indicate intended motions, which are then interpreted by automatic methods (e.g., GenAI models) to generate animation keyframes. Sketched notations have long served as cognitive instruments for animators, capturing forces, poses, dynamics, paths, and other animation features. However, such notations are often context-dependent, non-categorical, ambiguous, and composable based on our analysis of real-world animator-produced sketches. To facilitate interpretation, we first formalize these notations into a structured animation representation (i.e., source, path, and target). We then built an animation authoring system that translates high-level notations into the formalized intended animation, provides dynamic UI widgets for fine-grained parameter control, and establishes a closed feedback loop to resolve ambiguity. Finally, through a preliminary study with animators, we assess the usability of notational animating, reflect its affordance, and identify its contexts of use.

cs.HC↗

arXiv2026

NMR evidence of spin supersolid and Pomeranchuk effect behaviors in the triangular-lattice antiferromagnet Rb$_2$Ni$_2$(SeO$_3$)$_3$

We performed $^{85}$Rb nuclear magnetic resonance (NMR) measurements on the $S$ = 1 bilayer triangular-lattice antiferromagnet Rb$_2$Ni$_2$(SeO$_3$)$_3$ in magnetic fields up to 26 T. In the field range from 3 T to 26 T, the NMR spectral lines split and their respective spectral weight ratios reveal the existence of the magnetic up-up-down (UUD) phase, although the 1/3-plateau phase is only reached at fields above 16 T. Two distinct gapless regimes are further identified: one at low fields and low temperatures, and the other at high fields and high temperatures, consistent with the spin supersolid Y and V phases. Notably, the UUD-V phase boundary exhibits a negative slope in $dT/dH$, where the supersolid phase is located at temperatures above the solid phase due to strong low-energy spin fluctuations.

cond-mat.str-el↗

arXiv2026

Architectural Classification of XR Workloads: Cross-Layer Archetypes and Implications

Edge and mobile platforms for augmented and virtual reality, collectively referred to as extended reality (XR) must deliver deterministic ultra-low-latency performance under stringent power and area constraints. However, the diversity of XR workloads is rapidly increasing, characterized by heterogeneous operator types and complex dataflow structures. This trend poses significant challenges to conventional accelerator architectures centered around convolutional neural networks (CNNs), resulting in diminishing returns for traditional compute-centric optimization strategies. Despite the importance of this problem, a systematic architectural understanding of the full XR pipeline remains lacking. In this paper, we present an architectural classification of XR workloads using a cross-layer methodology that integrates model-based high-level design space exploration (DSE) with empirical profiling on commercial GPU and CPU hardware. By analyzing a representative set of workloads spanning 12 distinct XR kernels, we distill their complex architectural characteristics into a small set of cross-layer workload archetypes (e.g., capacity-limited and overhead-sensitive). Building on these archetypes, we further extract key architectural insights and provide actionable design guidelines for next-generation XR SoCs. Our study highlights that XR architecture design must shift from generic resource scaling toward phase-aware scheduling and elastic resource allocation in order to achieve greater energy efficiency and high performance in future XR systems.

cs.AR↗

arXiv2025

Training Report of TeleChat3-MoE

TeleChat3-MoE is the latest series of TeleChat large language models, featuring a Mixture-of-Experts (MoE) architecture with parameter counts ranging from 105 billion to over one trillion,trained end-to-end on Ascend NPU cluster. This technical report mainly presents the underlying training infrastructure that enables reliable and efficient scaling to frontier model sizes. We detail systematic methodologies for operator-level and end-to-end numerical accuracy verification, ensuring consistency across hardware platforms and distributed parallelism strategies. Furthermore, we introduce a suite of performance optimizations, including interleaved pipeline scheduling, attention-aware data scheduling for long-sequence training,hierarchical and overlapped communication for expert parallelism, and DVM-based operator fusion. A systematic parallelization framework, leveraging analytical estimation and integer linear programming, is also proposed to optimize multi-dimensional parallelism configurations. Additionally, we present methodological approaches to cluster-level optimizations, addressing host- and device-bound bottlenecks during large-scale training tasks. These infrastructure advancements yield significant throughput improvements and near-linear scaling on clusters comprising thousands of devices, providing a robust foundation for large-scale language model development on hardware ecosystems.

cs.CL↗

arXiv2025

PCMind-2.1-Kaiyuan-2B Technical Report

The rapid advancement of Large Language Models (LLMs) has resulted in a significant knowledge gap between the open-source community and industry, primarily because the latter relies on closed-source, high-quality data and training recipes. To address this, we introduce PCMind-2.1-Kaiyuan-2B, a fully open-source 2-billion-parameter model focused on improving training efficiency and effectiveness under resource constraints. Our methodology includes three key innovations: a Quantile Data Benchmarking method for systematically comparing heterogeneous open-source datasets and providing insights on data mixing strategies; a Strategic Selective Repetition scheme within a multi-phase paradigm to effectively leverage sparse, high-quality data; and a Multi-Domain Curriculum Training policy that orders samples by quality. Supported by a highly optimized data preprocessing pipeline and architectural modifications for FP16 stability, Kaiyuan-2B achieves performance competitive with state-of-the-art fully open-source models, demonstrating practical and scalable solutions for resource-limited pretraining. We release all assets (including model weights, data, and code) under Apache 2.0 license at https://huggingface.co/thu-pacman/PCMind-2.1-Kaiyuan-2B.

cs.CL↗

arXiv2025

How Learning Rate Decay Wastes Your Best Data in Curriculum-Based LLM Pretraining

Due to the scarcity of high-quality data, large language models (LLMs) are often trained on mixtures of data with varying quality levels, even after sophisticated data curation. A natural approach to better leverage high-quality data is curriculum-based pretraining, where the model is trained on data sorted in ascending order of quality as determined by a quality metric. However, prior studies have reported limited improvements from such curriculum-based pretraining strategies. This work identifies a critical factor constraining these methods: the incompatibility between the ascending data quality order and the decaying learning rate (LR) schedule. We find that while curriculum-based training substantially outperforms random shuffling when using a constant LR, its advantage diminishes under standard LR decay schedules. Our experiments show this incompatibility can be mitigated by two simple strategies: (1) employing a more moderate LR decay schedule, where the final LR is only moderately smaller than the peak LR, and (2) replacing LR decay with model averaging, i.e., computing a weighted average of the final few checkpoints. By combining these strategies, we improve the average score on a suite of standard benchmarks by 1.64% over random shuffling, without additional data refinement. Validated on 1.5B-parameter models trained over 30B tokens with various data-quality metrics, our findings call for a re-evaluation of curriculum-based LLM pretraining and underscore the potential of co-designing data curricula with optimization methods.

cs.LG↗

arXiv2025

PIMfused: Near-Bank DRAM-PIM with Fused-layer Dataflow for CNN Data Transfer Optimization

Near-bank Processing-in-Memory (PIM) architectures integrate processing cores (PIMcores) close to DRAM banks to mitigate the high cost of off-chip memory accesses. When accelerating convolutional neural network (CNN) on DRAM-PIM, performance is often constrained by cross-bank (or cross-PIMcore) data transfers, which are induced by the conventional layer-by-layer dataflow that enforces inter-bank (or inter-PIMcore) dependencies across successive CNN layers. To address this challenge, we propose PIMfused, a hardware-software co-design that enables fused-layer dataflow for end-to-end CNN execution in near-bank DRAM-PIM. By adopting fused-layer dataflow, PIMfused improves data reuse and, more importantly, breaks inter-bank data dependencies, thereby optimizing cross-bank data transfers without sacrificing bank-level parallelism. We study the impact of buffer sizes and PIMcore parallelism (1-bank vs. 4-bank) on PIMfused using end-to-end ResNet18. We present three key takeaways and show that with 4-bank PIMcores, PIMfused achieves overall PPA gains over a GDDR6-AiM-like baseline, cutting memory cycles to 30.6%, energy to 83.4%, and area to 76.5%.

cs.AR↗

arXiv2025

CaRe-BN: Precise Moving Statistics for Stabilizing Spiking Neural Networks in Reinforcement Learning

Spiking Neural Networks (SNNs) offer low-latency and energy-efficient decision-making on neuromorphic hardware by mimicking the event-driven dynamics of biological neurons. However, the discrete and non-differentiable nature of spikes leads to unstable gradient propagation in directly trained SNNs, making Batch Normalization (BN) an important component for stabilizing training. In online Reinforcement Learning (RL), imprecise BN statistics hinder exploitation, resulting in slower convergence and suboptimal policies. While Artificial Neural Networks (ANNs) can often omit BN, SNNs critically depend on it, limiting the adoption of SNNs for energy-efficient control on resource-constrained devices. To overcome this, we propose Confidence-adaptive and Re-calibration Batch Normalization (CaRe-BN), which introduces (i) a confidence-guided adaptive update strategy for BN statistics and (ii) a re-calibration mechanism to align distributions. By providing more accurate normalization, CaRe-BN stabilizes SNN optimization without disrupting the RL training process. Importantly, CaRe-BN does not alter inference, thus preserving the energy efficiency of SNNs in deployment. Extensive experiments on both discrete and continuous control benchmarks demonstrate that CaRe-BN improves SNN performance by up to $22.6\%$ across different spiking neuron models and RL algorithms. Remarkably, SNNs equipped with CaRe-BN even surpass their ANN counterparts by $5.9\%$. These results highlight a new direction for BN techniques tailored to RL, paving the way for neuromorphic agents that are both efficient and high-performing. Code is available at https://github.com/xuzijie32/CaRe-BN.

cs.NE↗

arXiv2025

Enhancing LLM-based Fault Localization with a Functionality-Aware Retrieval-Augmented Generation Framework

Fault localization (FL) is a critical but time-consuming task in software debugging, aiming to identify faulty code elements. While recent advances in large language models (LLMs) have shown promise for FL, they often struggle with complex systems due to the lack of project-specific knowledge and the difficulty of navigating large projects. To address these limitations, we propose FaR-Loc, a novel framework that enhances method-level FL by integrating LLMs with retrieval-augmented generation (RAG). FaR-Loc consists of three key components: LLM Functionality Extraction, Semantic Dense Retrieval, and LLM Re-ranking. First, given a failed test and its associated stack trace, the LLM Functionality Extraction module generates a concise natural language description that captures the failing behavior. Next, the Semantic Dense Retrieval component leverages a pre-trained code-understanding encoder to embed both the functionality description (natural language) and the covered methods (code) into a shared semantic space, enabling the retrieval of methods with similar functional behavior. Finally, the LLM Re-ranking module reorders the retrieved methods based on their contextual relevance. Our experiments on the widely used Defects4J benchmark show that FaR-Loc outperforms state-of-the-art LLM-based baselines SoapFL and AutoFL, by 14.6% and 9.1% in Top-1 accuracy, by 19.2% and 22.1% in Top-5 accuracy, respectively. It also surpasses all learning-based and spectrum-based baselines across all Top-N metrics without requiring re-training. Furthermore, we find that pre-trained code embedding models that incorporate code structure, such as UniXcoder, can significantly improve fault localization performance by up to 49.0% in Top-1 accuracy. Finally, we conduct a case study to illustrate the effectiveness of FaR-Loc and to provide insights for its practical application.

cs.SE↗

arXiv2025

SCAN: Self-Denoising Monte Carlo Annotation for Robust Process Reward Learning

Process reward models (PRMs) offer fine-grained, step-level evaluations that facilitate deeper reasoning processes in large language models (LLMs), proving effective in complex tasks like mathematical reasoning. However, developing PRMs is challenging due to the high cost and limited scalability of human-annotated data. Synthetic data from Monte Carlo (MC) estimation is a promising alternative but suffers from a high noise ratio, which can cause overfitting and hinder large-scale training. In this work, we conduct a preliminary study on the noise distribution in synthetic data from MC estimation, identifying that annotation models tend to both underestimate and overestimate step correctness due to limitations in their annotation capabilities. Building on these insights, we propose Self-Denoising Monte Carlo Annotation (SCAN), an efficient data synthesis and noise-tolerant learning framework. Our key findings indicate that: (1) Even lightweight models (e.g., 1.5B parameters) can produce high-quality annotations through a self-denoising strategy, enabling PRMs to achieve superior performance with only 6% the inference cost required by vanilla MC estimation. (2) With our robust learning strategy, PRMs can effectively learn from this weak supervision, achieving a 39.2 F1 score improvement (from 19.9 to 59.1) in ProcessBench. Despite using only a compact synthetic dataset, our models surpass strong baselines, including those trained on large-scale human-annotated datasets such as PRM800K. Furthermore, performance continues to improve as we scale up the synthetic data, highlighting the potential of SCAN for scalable, cost-efficient, and robust PRM training.

cs.LG↗

arXiv2025

BrowseComp-Plus: A More Fair and Transparent Evaluation Benchmark of Deep-Research Agent

Deep-Research agents, which integrate large language models (LLMs) with search tools, have shown success in improving the effectiveness of handling complex queries that require iterative search planning and reasoning over search results. Evaluations on current benchmarks like BrowseComp relies on black-box live web search APIs, have notable limitations in (1) fairness: dynamic and opaque web APIs hinder fair comparisons and reproducibility of deep research methods; (2) transparency: lack of control over the document corpus makes it difficult to isolate retriever contributions. In other words, the current evaluations may compare a complete deep research system at a given time, but they do not foster well-controlled experiments to provide insights into the capability of underlying deep research LLMs. To address these challenges, we introduce BrowseComp-Plus, a benchmark derived from BrowseComp, employing a fixed, carefully curated corpus. Each query in BrowseComp-Plus includes human-verified supporting documents and mined challenging negatives, enabling controlled experimentation. The benchmark is shown to be effective in distinguishing the performance of deep research systems. For instance, the open-source model Search-R1, when paired with the BM25 retriever, achieves 3.86% accuracy, whereas the GPT-5 achieves 55.9%. Integrating the GPT-5 with the Qwen3-Embedding-8B retriever further enhances its accuracy to 70.1% with fewer search calls. This benchmark allows comprehensive evaluation and disentangled analysis of deep research agents and retrieval methods, fostering insights into retrieval effectiveness, citation accuracy, and context engineering in Deep-Research system.

cs.CL↗

arXiv2025

NMR study of supersolid phases in the triangular-lattice antiferromagnet Na2BaCo(PO4)2

We report ultra-low-temperature $^{23}$Na NMR measurements on the Ising triangular lattice antiferromagnet Na$_2$BaCo(PO$_4$)$_2$, which precisely resolve the phase diagram under magnetic field applied along the crystalline $c$ axis. With increasing field, the NMR spectra resolve three ordered phases with distinct spin configurations: the Y, up-up-down (UUD), and V phases. The spin-lattice relaxation rate $1/T_1$ data demonstrate gapless excitations in the Y and V phases, strongly supporting their supersolid nature. However, the phase transitions from the UUD phase to the two supersolid phases exhibit dramatically different behaviors upon cooling. Prior to entering the Y phase, $1/T_1$ identifies a gapless regime within the UUD phase, suggesting a Berezinskii-Kosterlitz-Thouless phase above a second-order phase transition. In contrast, the coexistence of the UUD and V phases observed in our experiments provides direct evidence of a first-order phase transition between these phases.

cond-mat.str-el↗

arXiv2025

Towards High-performance Spiking Transformers from ANN to SNN Conversion

Spiking neural networks (SNNs) show great potential due to their energy efficiency, fast processing capabilities, and robustness. There are two main approaches to constructing SNNs. Direct training methods require much memory, while conversion methods offer a simpler and more efficient option. However, current conversion methods mainly focus on converting convolutional neural networks (CNNs) to SNNs. Converting Transformers to SNN is challenging because of the presence of non-linear modules. In this paper, we propose an Expectation Compensation Module to preserve the accuracy of the conversion. The core idea is to use information from the previous T time-steps to calculate the expected output at time-step T. We also propose a Multi-Threshold Neuron and the corresponding Parallel Parameter normalization to address the challenge of large time steps needed for high accuracy, aiming to reduce network latency and power consumption. Our experimental results demonstrate that our approach achieves state-of-the-art performance. For example, we achieve a top-1 accuracy of 88.60\% with only a 1\% loss in accuracy using 4 time steps while consuming only 35\% of the original power of the Transformer. To our knowledge, this is the first successful Artificial Neural Network (ANN) to SNN conversion for Spiking Transformers that achieves high accuracy, low latency, and low power consumption on complex datasets. The source codes of the proposed method are available at https://github.com/h-z-h-cell/Transformer-to-SNN-ECMT.

cs.CV↗

arXiv2025

Brickify: Enabling Expressive Design Intent Specification through Direct Manipulation on Design Tokens

Expressing design intent using natural language prompts requires designers to verbalize the ambiguous visual details concisely, which can be challenging or even impossible. To address this, we introduce Brickify, a visual-centric interaction paradigm -- expressing design intent through direct manipulation on design tokens. Brickify extracts visual elements (e.g., subject, style, and color) from reference images and converts them into interactive and reusable design tokens that can be directly manipulated (e.g., resize, group, link, etc.) to form the visual lexicon. The lexicon reflects users' intent for both what visual elements are desired and how to construct them into a whole. We developed Brickify to demonstrate how AI models can interpret and execute the visual lexicon through an end-to-end pipeline. In a user study, experienced designers found Brickify more efficient and intuitive than text-based prompts, allowing them to describe visual details, explore alternatives, and refine complex designs with greater ease and control.

cs.HC↗