arXiv subjects

Tingting Xu

Publications and source records attributed to Tingting Xu.

9 recordsLinked to original sources

arXiv2026

Understanding before Naming! Enhancing LLM-based Method Name Prediction with Code Summarization

Method names are critical to software quality, affecting code comprehensibility, maintainability, and developer collaboration. However, manually designing meaningful method names is challenging. Method Name Prediction (MNP), which automatically generates method names from code snippets, has recently attracted attention. Although large language models (LLMs) show promising performance for MNP, two challenges remain. First, existing evaluations mainly rely on token similarity metrics, which often fail to reflect human judgments of semantic quality. Second, current LLM-based MNP methods usually generate names through direct code-to-name mapping, which differs from the human process of understanding functionality before naming. To address these challenges, we conduct empirical studies on LLM-based evaluation and MNP strategies. We compare 6 metric-based evaluators, 5 LLM-based evaluators, and 6 human evaluators. Results show that LLM-based evaluators, especially DeepSeek-based evaluators, are more consistent with human judgments than traditional metrics. We further compare direct generation and summarization-and-refinement strategies. Results indicate that summarization and refinement generally improve the semantic quality of generated names. Case studies reveal three limitations: inaccurate summaries, semantic misalignment, and close semantic scores. Based on these findings, we propose SMNP, an MNP approach combining MNP-oriented summarization and chain-of-thought enhanced refinement. Experiments on 5 LLMs and 2 datasets demonstrate the effectiveness and robustness of SMNP.

cs.SE↗

arXiv2026

Securing Code Understanding: Detecting Natural Backdoor Vulnerability in Code Language Models

Code Language Models (CodeLMs) have become integral to software engineering, significantly advancing code intelligence tasks. However, their widespread adoption has raised critical security concerns, particularly regarding susceptibility to backdoor attacks. Recent studies have uncovered naturally occurring backdoors, referred to as natural backdoors, in normally trained deep learning models. Despite posing threats as serious as those introduced through data poisoning, security implications of natural backdoor vulnerabilities in CodeLMs remain poorly understood. In this paper, we conduct a thorough empirical study of natural backdoor vulnerabilities in CodeLMs across various model architectures and code intelligence tasks. Specifically, we examine potential natural backdoor vulnerabilities across 44 scenarios, demonstrating that natural backdoors are prevalent and intrinsic to CodeLMs. We reveal differences between injected and natural backdoor vulnerabilities at both the model and parameter levels. We then analyze the transferability of natural backdoor vulnerabilities from three perspectives: datasets, model architectures, and shared knowledge. We further investigate the causes of natural backdoors from two aspects: training datasets and the model training procedure. We evaluate existing backdoor defense techniques, including pre-training, in-training, and post-training defenses, in mitigating natural backdoors. Finally, we propose ScanNBT, a novel detection method designed to improve comprehensive detection of natural backdoor vulnerabilities in CodeLMs. We aim for our findings to enhance understanding of these vulnerabilities and provide insights for strengthening CodeLM security against backdoor threats.

cs.CR↗

arXiv2025

Data Quality Issues in Multilingual Speech Datasets: The Need for Sociolinguistic Awareness and Proactive Language Planning

Our quality audit for three widely used public multilingual speech datasets - Mozilla Common Voice 17.0, FLEURS, and Vox Populi - shows that in some languages, these datasets suffer from significant quality issues, which may obfuscate downstream evaluation results while creating an illusion of success. We divide these quality issues into two categories: micro-level and macro-level. We find that macro-level issues are more prevalent in less institutionalized, often under-resourced languages. We provide a case analysis of Taiwanese Southern Min (nan_tw) that highlights the need for proactive language planning (e.g. orthography prescriptions, dialect boundary definition) and enhanced data quality control in the dataset creation process. We conclude by proposing guidelines and recommendations to mitigate these issues in future dataset development, emphasizing the importance of sociolinguistic awareness and language planning principles. Furthermore, we encourage research into how this creation process itself can be leveraged as a tool for community-led language planning and revitalization.

cs.CL↗

arXiv2025

Hemispheric Distribution of Solar Active Regions During Solar Cycles 23-25

Solar active regions (ARs) are crucial for understanding the long-term evolution of solar activities and predicting eruptive phenomena, including solar flares and coronal mass ejections. However, the cycle-dependent properties in the north-south asymmetry of ARs have not been fully understood. In this study, we investigate the hemispheric distribution of ARs from Carrington Rotation 1909 to 2278 (between 1996 May and 2023 November) by using three parameters that describe the magnetic field distribution of ARs: number, area, and flux. The main findings are as follows: (1) The three AR parameters show significant hemispheric asymmetry in cycles 23-25. The strong correlation between AR area and flux indicates that they can better reflect the intrinsic properties of solar magnetic field. (2) The correlation between sunspot activity and AR parameters varies in the two hemispheres across the different cycles. The AR parameters provide additional information for the variations in sunspot activity, which can better predict the intensity and cyclical changes of solar activity. (3) The variation in the fitting slope sign of the asymmetry index for AR parameters reflects periodic changes in hemispheric ARs, providing valuable insights into the activity of other stars. (4) Both the dominant hemisphere and the cumulative trend of AR parameters display a cycle-dependent behavior. Moreover, the trend variations of AR area and flux are similar, reflecting the long-term evolutionary characteristics of solar magnetic field. Our analysis results are relevant for understanding the hemispheric coupling of solar magnetic activity and its cyclic evolutionary patterns.

astro-ph.SR↗

arXiv2024

A Prompt Learning Framework for Source Code Summarization

(Source) code summarization is the task of automatically generating natural language summaries (also called comments) for given code snippets. Recently, with the successful application of large language models (LLMs) in numerous fields, software engineering researchers have also attempted to adapt LLMs to solve code summarization tasks. The main adaptation schemes include instruction prompting, task-oriented (full-parameter) fine-tuning, and parameter-efficient fine-tuning (PEFT). However, instruction prompting involves designing crafted prompts and requires users to have professional domain knowledge, while task-oriented fine-tuning requires high training costs, and effective, tailored PEFT methods for code summarization are still lacking. This paper proposes an effective prompt learning framework for code summarization called PromptCS. It no longer requires users to rack their brains to design effective prompts. Instead, PromptCS trains a prompt agent that can generate continuous prompts to unleash the potential for LLMs in code summarization. Compared to the human-written discrete prompt, the continuous prompts are produced under the guidance of LLMs and are therefore easier to understand by LLMs. PromptCS is non-invasive to LLMs and freezes the parameters of LLMs when training the prompt agent, which can greatly reduce the requirements for training resources. Our comprehensive experimental results show that PromptCS significantly outperforms instruction prompting schemes (including zero-shot learning and few-shot learning) on all four widely used metrics, and is comparable to the task-oriented fine-tuning scheme. In some base LLMs, e.g., StarCoderBase-1B and -3B, PromptCS even outperforms the task-oriented fine-tuning scheme. More importantly, the training efficiency of PromptCS is faster than the task-oriented fine-tuning scheme, with a more pronounced advantage on larger LLMs.

cs.SE↗

arXiv2023

Galaxy Morphology Classification Using Multi-Scale Convolution Capsule Network

The classification of galaxy morphology is a hot issue in astronomical research. Although significant progress has been made in the last decade in classifying galaxy morphology using deep learning technology, there are still some deficiencies in spatial feature representation and classification accuracy. In this study, we present a multi-scale convolutional capsule network (MSCCN) model for the classification of galaxy morphology. First, this model improves the convolutional layers through using a multi-branch structure to extract multi-scale hidden features of galaxy images. In order to further explore the hidden information in the features, the multi-scale features are encapsulated and fed into the capsule layer. Second, we use a sigmoid function to replace the softmax function in dynamic routing, which can enhance the robustness of MSCCN. Finally, the classification model achieving 97% accuracy, 96% precision, 98% recall, and 97% F1-score under macroscopic averaging. In addition, a more comprehensive model evaluation were accomplished in this study. We visualized the morphological features for the part of sample set, which using the t-distributed stochastic neighbor embedding (t-SNE) algorithm. The results shows that the model has the better generalization ability and robustness, it can be effectively used in the galaxy morphological classification.

astro-ph.GA↗

arXiv2022

RFI Identification Based On Deep-Learning]{A Robust RFI Identification For Radio Interferometry based on a Convolutional Neural Network

The rapid development of new generation radio interferometers such as the Square Kilometer Array (SKA) has opened up unprecedented opportunities for astronomical research. However, anthropogenic Radio Frequency Interference (RFI) from communication technologies and other human activities severely affects the fidelity of observational data. It also significantly reduces the sensitivity of the telescopes. We proposed a robust Convolutional Neural Network (CNN) model to identify RFI based on machine learning methods. We overlaid RFI on the simulation data of SKA1-LOW to construct three visibility function datasets. One dataset was used for modeling, and the other two were used for validating the model's usability. The experimental results show that the Area Under the Curve (AUC) reaches 0.93, with satisfactory accuracy and precision. We then further investigated the effectiveness of the model by identifying the RFI in the actual observational data from LOFAR and MeerKAT. The results show that the model performs well. The overall effectiveness is comparable to AOFlagger software and provides an improvement over existing methods in some instances.

astro-ph.IM↗

arXiv2021

A Catalog of LAMOST Variable Sources Based on Time-domain Photometry of ZTF

The identification and analysis of different variable sources is a hot issue in astrophysical research. The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) spectroscopic survey has accumulated massive spectral data but contains no information about variable sources. Although a few related studies present variable source catalogs for the LAMOST, the studies still have a few deficiencies regarding the type and number of variable sources identified. In this study, we presented a statistical modeling approach to identify variable source candidates. We first crossed the Kepler, Sloan Digital Sky Survey (SDSS), and Zwicky Transient Facility (ZTF) catalogs to obtain light curves data of variable and non-variable sources. The data are then modeled statistically using commonly used variability parameters, respectively. And then, an optimal variable source identification model is determined using the Receiver Operating Characteristic (ROC) curve and four credible evaluation indices such as precision, accuracy, recall, and F1score. Based on this identification model, a catalog of LAMOST variable sources (including 631,769 variable source candidates with a probability greater than 95% and so on) is obtained. To validate the correctness of the catalog, we performed a two-by-two cross-comparison with the GAIA catalog and other published variable source catalogs. We achieved the correct rate ranging from 50% to 100%. Among the 123,756 sources cross-matched, our variable source catalog identifies 85,669 with a correct rate of 69%, which indicates that the variable source catalog presented in this study is credible.

astro-ph.SR↗

arXiv2018

Predicting Chronic Disease Hospitalizations from Electronic Health Records: An Interpretable Classification Approach

Urban living in modern large cities has significant adverse effects on health, increasing the risk of several chronic diseases. We focus on the two leading clusters of chronic disease, heart disease and diabetes, and develop data-driven methods to predict hospitalizations due to these conditions. We base these predictions on the patients' medical history, recent and more distant, as described in their Electronic Health Records (EHR). We formulate the prediction problem as a binary classification problem and consider a variety of machine learning methods, including kernelized and sparse Support Vector Machines (SVM), sparse logistic regression, and random forests. To strike a balance between accuracy and interpretability of the prediction, which is important in a medical setting, we propose two novel methods: K-LRT, a likelihood ratio test-based method, and a Joint Clustering and Classification (JCC) method which identifies hidden patient clusters and adapts classifiers to each cluster. We develop theoretical out-of-sample guarantees for the latter method. We validate our algorithms on large datasets from the Boston Medical Center, the largest safety-net hospital system in New England.

cs.LG↗