arXiv subjects

Xinchen Wang

Publications and source records attributed to Xinchen Wang.

17 recordsLinked to original sources

arXiv2026

Evaluating LLM-Based 0-to-1 Software Generation in End-to-End CLI Tool Scenarios

The evolution of Large Language Models (LLMs) has catalyzed a paradigm shift towards intent-driven software development, where autonomous agents are expected to design and deliver complete, runnable software systems from scratch. However, existing benchmarks fail to adequately assess this 0-to-1 generation capability due to two fundamental limitations. First, they rely on predefined structural scaffolds, which reduces the task to mere file-filling. Second, they depend on rigid white-box unit testing, which forces generated code to conform to specific internal implementations rather than validating end-to-end user-centric behavior. To bridge this gap, we introduce CLI-Tool-Bench, a novel, structure-agnostic benchmark designed to evaluate the ground-up generation of Command-Line Interface (CLI) tools. Powered by an automated black-box differential testing framework, the benchmark comprises 94 high-quality, real-world repositories spanning diverse programming languages and complexity levels. For each task, agents are provided with an empty workspace, forcing them to autonomously handle repository planning and dependencies. We evaluate the generated software by executing it in isolated sandboxes. The system-level side effects and terminal outputs are then compared against human-written oracles using a rigorous multi-tiered equivalence metric. Extensive evaluation of seven state-of-the-art LLMs reveals that the top-tier models achieve a maximum overall success rate of only 43.8%, highlighting that 0-to-1 software generation remains a highly challenging frontier. Furthermore, we discover that agents exhibit a strong tendency to generate monolithic code structures, and that higher token consumption does not necessarily yield better task performance.

cs.SE↗

arXiv2026

Evaluating Repository-level Software Documentation via Question Answering and Feature-Driven Development

Software documentation is crucial for repository comprehension. While Large Language Models (LLMs) advance documentation generation from code snippets to entire repositories, existing benchmarks have two key limitations: (1) they lack a holistic, repository-level assessment, and (2) they rely on unreliable evaluation strategies, such as LLM-as-a-judge, which suffers from vague criteria and limited repository-level knowledge. To address these issues, we introduce SWD-Bench, a novel benchmark for evaluating repository-level software documentation. Inspired by documentation-driven development, our strategy evaluates documentation quality by assessing an LLM's ability to understand and implement functionalities using the documentation, rather than by directly scoring it. This is measured through function-driven Question Answering (QA) tasks. SWD-Bench comprises three interconnected QA tasks: (1) Functionality Detection, to determine if a functionality is described; (2) Functionality Localization, to evaluate the accuracy of locating related files; and (3) Functionality Completion, to measure the comprehensiveness of implementation details. We construct the benchmark, containing 4,170 entries, by mining high-quality Pull Requests and enriching them with repository-level context. Experiments reveal limitations in current documentation generation methods and show that source code provides complementary value. Notably, documentation from the best-performing method improves the issue-solving rate of SWE-Agent by 20.00%, which demonstrates the practical value of high-quality documentation in supporting documentation-driven development.

cs.SE↗

arXiv2025

Benchmarking LLMs for Fine-Grained Code Review with Enriched Context in Practice

Code review is a cornerstone of software quality assurance, and recent advances in Large Language Models (LLMs) have shown promise in its automation. However, existing benchmarks for LLM-based code review face three major limitations. Lack of semantic context: most benchmarks provide only code diffs without textual information such as issue descriptions, which are crucial for understanding developer intent. Data quality issues: without rigorous validation, many samples are noisy-e.g., reviews on outdated or irrelevant code-reducing evaluation reliability. Coarse granularity: most benchmarks operate at the file or commit level, overlooking the fine-grained, line-level reasoning essential for precise review. We introduce ContextCRBench, a high-quality, context-rich benchmark for fine-grained LLM evaluation in code review. Our construction pipeline comprises: Raw Data Crawling, collecting 153.7K issues and pull requests from top-tier repositories; Comprehensive Context Extraction, linking issue-PR pairs for textual context and extracting the full surrounding function or class for code context; and Multi-stage Data Filtering, combining rule-based and LLM-based validation to remove outdated, malformed, or low-value samples, resulting in 67,910 context-enriched entries. ContextCRBench supports three evaluation scenarios aligned with the review workflow: hunk-level quality assessment, line-level defect localization, and line-level comment generation. Evaluating eight leading LLMs (four closed-source and four open-source) reveals that textual context yields greater performance gains than code context alone, while current LLMs remain far from human-level review ability. Deployed at ByteDance, ContextCRBench drives a self-evolving code review system, improving performance by 61.98% and demonstrating its robustness and industrial utility. https://github.com/kinesiatricssxilm14/ContextCRBench.

cs.SE↗

arXiv2025

AXIOM: Benchmarking LLM-as-a-Judge for Code via Rule-Based Perturbation and Multisource Quality Calibration

Large language models (LLMs) have been increasingly deployed in real-world software engineering, fostering the development of code evaluation metrics to study the quality of LLM-generated code. Conventional rule-based metrics merely score programs based on their surface-level similarities with reference programs instead of analyzing functionality and code quality in depth. To address this limitation, researchers have developed LLM-as-a-judge metrics, prompting LLMs to evaluate and score code, and curated various code evaluation benchmarks to validate their effectiveness. However, these benchmarks suffer from critical limitations, hindering reliable assessments of evaluation capability: Some feature coarse-grained binary labels, which reduce rich code behavior to a single bit of information, obscuring subtle errors. Others propose fine-grained but subjective, vaguely-defined evaluation criteria, introducing unreliability in manually-annotated scores, which is the ground-truth they rely on. Furthermore, they often use uncontrolled data synthesis methods, leading to unbalanced score distributions that poorly represent real-world code generation scenarios. To curate a diverse benchmark with programs of well-balanced distributions across various quality levels and streamline the manual annotation procedure, we propose AXIOM, a novel perturbation-based framework for synthesizing code evaluation benchmarks at scale. It reframes program scores as the refinement effort needed for deployment, consisting of two stages: (1) Rule-guided perturbation, which prompts LLMs to apply sequences of predefined perturbation rules to existing high-quality programs to modify their functionality and code quality, enabling us to precisely control each program's target score to achieve balanced score distributions. (2) Multisource quality calibration, which first selects a subset of...

cs.SE↗

arXiv2025

CodeVisionary: An Agent-based Framework for Evaluating Large Language Models in Code Generation

Large language models (LLMs) have demonstrated strong capabilities in code generation, underscoring the critical need for rigorous and comprehensive evaluation. Existing evaluation approaches fall into three categories, including human-centered, metric-based, and LLM-based. Considering that human-centered approaches are labour-intensive and metric-based ones overly rely on reference answers, LLM-based approaches are gaining increasing attention due to their stronger contextual understanding capabilities. However, they generally evaluate the generated code based on static prompts, and tend to fail for complex code scenarios which typically involve multiple requirements and require more contextual information. In addition, these approaches lack fine-grained evaluation for complex code, resulting in limited explainability. To mitigate the limitations, we propose CodeVisionary, the first agent-based evaluation framework for complex code generation. CodeVisionary consists of two stages: (1) Requirement-guided multi-dimensional context distillation stage and (2) Fine-grained scoring and summarization stage. A comprehensive evaluation report is also generated for enhanced explainability. For validation, we construct a new benchmark consisting of 363 samples spanning 37 coding scenarios and 23 programming languages. Extensive experiments demonstrate that CodeVisionary achieves the best performance among three baselines for evaluating complex code generation, outperforming the best baseline with average improvements of 0.217, 0.163, and 0.141 in Pearson, Spearman, and Kendall-Tau coefficients, respectively. The resources of CodeVisionary are available at https://github.com/Eshe0922/CodeVisionary.

cs.SE↗

arXiv2025

Repo2Run: Automated Building Executable Environment for Code Repository at Scale

Scaling up executable code data is significant for improving language models' software engineering capability. The intricate nature of the process makes it labor-intensive, time-consuming and expert-knowledge-dependent to build a large number of executable code repositories, limiting the scalability of existing work based on running tests. The primary bottleneck lies in the automated building of test environments for different repositories, which is an essential yet underexplored task. To mitigate the gap, we introduce Repo2Run, the first LLM-based agent aiming at automating the building of executable test environments for any repositories at scale. Specifically, given a code repository, Repo2Run iteratively builds the Docker image, runs unit tests based on the feedback of the building, and synthesizes the Dockerfile until the entire pipeline is executed successfully. The resulting Dockerfile can then be used to create Docker container environments for running code and tests. We created a benchmark containing 420 Python repositories with unit tests for evaluation. The results illustrate that Repo2Run achieves an 86.0% success rate, outperforming SWE-agent by 77.0%. The resources of Repo2Run are available at https://github.com/bytedance/Repo2Run.

cs.SE↗

arXiv2025

Trae Agent: An LLM-based Agent for Software Engineering with Test-time Scaling

Software issue resolution is a critical challenge in software engineering and has garnered increasing attention in recent years. With the rapid advancement of large language models (LLMs), substantial progress has been made in addressing real-world software engineering tasks. Recent studies have introduced ensemble reasoning techniques to enhance the performance of LLM-based issue resolution. However, existing prompting-based methods still face limitations in effectively exploring large ensemble spaces and lack the capacity for repository-level understanding, both of which constrain their overall effectiveness. In this paper, we propose Trae Agent, the first agent-based ensemble reasoning approach for repository-level issue resolution. Trae Agent formulates our goal as an optimal solution search problem and addresses two key challenges, i.e., large ensemble spaces and repository-level understanding, through modular agents for generation, pruning, and selection. We conduct extensive experiments using three leading LLMs on the widely-adopted SWE-bench benchmark, comparing Trae Agent against four state-of-the-art ensemble reasoning techniques. Experimental results demonstrate that Trae Agent consistently achieves superior performance, with an average improvement of 10.22% over all baselines in terms of Pass@1. Trae Agent has achieved first place on the SWE-bench Verified leaderboard, with a notable Pass@1 score of 75.20%. We are pleased to release Trae Agent as an open-source project to support the research community, with all resources available at https://github.com/bytedance/trae-agent.

cs.SE↗

arXiv2024

CodeRepoQA: A Large-scale Benchmark for Software Engineering Question Answering

In this work, we introduce CodeRepoQA, a large-scale benchmark specifically designed for evaluating repository-level question-answering capabilities in the field of software engineering. CodeRepoQA encompasses five programming languages and covers a wide range of scenarios, enabling comprehensive evaluation of language models. To construct this dataset, we crawl data from 30 well-known repositories in GitHub, the largest platform for hosting and collaborating on code, and carefully filter raw data. In total, CodeRepoQA is a multi-turn question-answering benchmark with 585,687 entries, covering a diverse array of software engineering scenarios, with an average of 6.62 dialogue turns per entry. We evaluate ten popular large language models on our dataset and provide in-depth analysis. We find that LLMs still have limitations in question-answering capabilities in the field of software engineering, and medium-length contexts are more conducive to LLMs' performance. The entire benchmark is publicly available at https://github.com/kinesiatricssxilm14/CodeRepoQA.

cs.SE↗

arXiv2024

AEGIS: An Agent-based Framework for General Bug Reproduction from Issue Descriptions

In software maintenance, bug reproduction is essential for effective fault localization and repair. Manually writing reproduction scripts is a time-consuming task with high requirements for developers. Hence, automation of bug reproduction has increasingly attracted attention from researchers and practitioners. However, the existing studies on bug reproduction are generally limited to specific bug types such as program crashes, and hard to be applied to general bug reproduction. In this paper, considering the superior performance of agent-based methods in code intelligence tasks, we focus on designing an agent-based framework for the task. Directly employing agents would lead to limited bug reproduction performance, due to entangled subtasks, lengthy retrieved context, and unregulated actions. To mitigate the challenges, we propose an Automated gEneral buG reproductIon Scripts generation framework, named AEGIS, which is the first agent-based framework for the task. AEGIS mainly contains two modules: (1) A concise context construction module, which aims to guide the code agent in extracting structured information from issue descriptions, identifying issue-related code with detailed explanations, and integrating these elements to construct the concise context; (2) A FSM-based multi-feedback optimization module to further regulate the behavior of the code agent within the finite state machine (FSM), ensuring a controlled and efficient script generation process based on multi-dimensional feedback. Extensive experiments on the public benchmark dataset show that AEGIS outperforms the state-of-the-art baseline by 23.0% in F->P metric. In addition, the bug reproduction scripts generated by AEGIS can improve the relative resolved rate of Agentless by 12.5%.

cs.SE↗

arXiv2024

A Real-World Benchmark for Evaluating Fine-Grained Issue Solving Capabilities of Large Language Models

Automatically resolving software issues is crucial for software development in practice, impacting the software quality and user experience. The process of resolving real-world issues encompasses tasks such as question-answering (QA), fault localization, and code editing. Existing benchmarks such as HumanEval fall short in their ability to assess LLMs' proficiency in solving issues within a codebase. Although benchmarks like SWE-Bench are designed to evaluate the LLMs' capability to handle real-world GitHub issues, the end-to-end evaluation method cannot provide granular insights on the performance of subtasks involved in issue solving. To address existing deficiencies in benchmarking LLMs for practical software engineering tasks, we introduce FAUN-Eval, a benchmark specifically designed to evaluate the Fine-grAined issUe solviNg capabilities of LLMs. FAUN-Eval systematically assesses LLMs across three distinct tasks: QA, fault localization, and code editing. This benchmark is constructed using a dataset curated from 30 well-known GitHub repositories. For each entry, issue and pull request (PR) pairs are meticulously compiled and validated using cross-referencing and keyword verification methods. FAUN-Eval includes 300 entries and employs both LLM and manual checks to ensure data quality. We evaluate ten LLMs with FAUN-Eval, including four closed-source and six open-source models. Our experimental results reveal several key findings. We find that the top-performing LLMs differ across the different tasks. Additionally, features in issues may lead LLMs to generate incorrect information. Moreover, models may vary in their proficiency with texts of different lengths.

cs.SE↗

arXiv2024

Pyramid-Driven Alignment: Pyramid Principle Guided Integration of Large Language Models and Knowledge Graphs

Large Language Models (LLMs) possess impressive reasoning abilities but are prone to generating incorrect information, often referred to as hallucinations. While incorporating external Knowledge Graphs (KGs) can partially mitigate this issue, existing methods primarily treat KGs as static knowledge repositories, overlooking the critical disparity between KG and LLM knowledge, and failing to fully exploit the reasoning capabilities inherent in KGs. To address these limitations, we propose Pyramid-Driven Alignment (PDA), a novel framework for seamlessly integrating LLMs with KGs. PDA utilizes Pyramid Principle analysis to construct a hierarchical pyramid structure. This structure is designed to reflect the input question and generate more validated deductive knowledge, thereby enhancing the alignment of LLMs and KGs and ensuring more cohesive integration. Furthermore, PDA employs a recursive mechanism to harness the underlying reasoning abilities of KGs, resulting in more accurate knowledge retrieval for question-answering tasks. Our experimental results reveal a substantial performance advantage of PDA over state-of-the-art baselines, with improvements reaching 26.70% and 26.78%.

cs.CL↗

arXiv2024

MarsCode Agent: AI-native Automated Bug Fixing

Recent advances in large language models (LLMs) have shown significant potential to automate various software development tasks, including code completion, test generation, and bug fixing. However, the application of LLMs for automated bug fixing remains challenging due to the complexity and diversity of real-world software systems. In this paper, we introduce MarsCode Agent, a novel framework that leverages LLMs to automatically identify and repair bugs in software code. MarsCode Agent combines the power of LLMs with advanced code analysis techniques to accurately localize faults and generate patches. Our approach follows a systematic process of planning, bug reproduction, fault localization, candidate patch generation, and validation to ensure high-quality bug fixes. We evaluated MarsCode Agent on SWE-bench, a comprehensive benchmark of real-world software projects, and our results show that MarsCode Agent achieves a high success rate in bug fixing compared to most of the existing automated approaches.

cs.SE↗

arXiv2024

VulEval: Towards Repository-Level Evaluation of Software Vulnerability Detection

Deep Learning (DL)-based methods have proven to be effective for software vulnerability detection, with a potential for substantial productivity enhancements for detecting vulnerabilities. Current methods mainly focus on detecting single functions (i.e., intra-procedural vulnerabilities), ignoring the more complex inter-procedural vulnerability detection scenarios in practice. For example, developers routinely engage with program analysis to detect vulnerabilities that span multiple functions within repositories. In addition, the widely-used benchmark datasets generally contain only intra-procedural vulnerabilities, leaving the assessment of inter-procedural vulnerability detection capabilities unexplored. To mitigate the issues, we propose a repository-level evaluation system, named \textbf{VulEval}, aiming at evaluating the detection performance of inter- and intra-procedural vulnerabilities simultaneously. Specifically, VulEval consists of three interconnected evaluation tasks: \textbf{(1) Function-Level Vulnerability Detection}, aiming at detecting intra-procedural vulnerability given a code snippet; \textbf{(2) Vulnerability-Related Dependency Prediction}, aiming at retrieving the most relevant dependencies from call graphs for providing developers with explanations about the vulnerabilities; and \textbf{(3) Repository-Level Vulnerability Detection}, aiming at detecting inter-procedural vulnerabilities by combining with the dependencies identified in the second task. VulEval also consists of a large-scale dataset, with a total of 4,196 CVE entries, 232,239 functions, and corresponding 4,699 repository-level source code in C/C++ programming languages. Our analysis highlights the current progress and future directions for software vulnerability detection.

cs.SE↗

arXiv2024

ReposVul: A Repository-Level High-Quality Vulnerability Dataset

Open-Source Software (OSS) vulnerabilities bring great challenges to the software security and pose potential risks to our society. Enormous efforts have been devoted into automated vulnerability detection, among which deep learning (DL)-based approaches have proven to be the most effective. However, the current labeled data present the following limitations: (1) Tangled Patches: Developers may submit code changes unrelated to vulnerability fixes within patches, leading to tangled patches. (2) Lacking Inter-procedural Vulnerabilities: The existing vulnerability datasets typically contain function-level and file-level vulnerabilities, ignoring the relations between functions, thus rendering the approaches unable to detect the inter-procedural vulnerabilities. (3) Outdated Patches: The existing datasets usually contain outdated patches, which may bias the model during training. To address the above limitations, in this paper, we propose an automated data collection framework and construct the first repository-level high-quality vulnerability dataset named ReposVul. The proposed framework mainly contains three modules: (1) A vulnerability untangling module, aiming at distinguishing vulnerability-fixing related code changes from tangled patches, in which the Large Language Models (LLMs) and static analysis tools are jointly employed. (2) A multi-granularity dependency extraction module, aiming at capturing the inter-procedural call relationships of vulnerabilities, in which we construct multiple-granularity information for each vulnerability patch, including repository-level, file-level, function-level, and line-level. (3) A trace-based filtering module, aiming at filtering the outdated patches, which leverages the file path trace-based filter and commit time trace-based filter to construct an up-to-date dataset.

cs.CR↗

arXiv2024

Game Rewards Vulnerabilities: Software Vulnerability Detection with Zero-Sum Game and Prototype Learning

Recent years have witnessed a growing focus on automated software vulnerability detection. Notably, deep learning (DL)-based methods, which employ source code for the implicit acquisition of vulnerability patterns, have demonstrated superior performance compared to other approaches. However, the DL-based approaches are still hard to capture the vulnerability-related information from the whole code snippet, since the vulnerable parts usually account for only a small proportion. As evidenced by our experiments, the approaches tend to excessively emphasize semantic information, potentially leading to limited vulnerability detection performance in practical scenarios. First, they cannot well distinguish between the code snippets before (i.e., vulnerable code) and after (i.e., non-vulnerable code) developers' fixes due to the minimal code changes. Besides, substituting user-defined identifiers with placeholders (e.g., "VAR1" and "FUN1") in obvious performance degradation at up to 14.53% with respect to the F1 score. To mitigate these issues, we propose to leverage the vulnerable and corresponding fixed code snippets, in which the minimal changes can provide hints about semantic-agnostic features for vulnerability detection. In this paper, we propose a software vulneRability dEteCtion framework with zerO-sum game and prototype learNing, named RECON. In RECON, we propose a zero-sum game construction module. Distinguishing the vulnerable code from the corresponding fixed code is regarded as one player (i.e. Calibrator), while the conventional vulnerability detection is another player (i.e. Detector) in the zero-sum game. The goal is to capture the semantic-agnostic features of the first player for enhancing the second player's performance for vulnerability detection. Experiments on the public benchmark dataset show that RECON outperforms the state-of-the-art baseline by 6.29% in F1 score.

cs.SE↗

arXiv2023

When Less is Enough: Positive and Unlabeled Learning Model for Vulnerability Detection

Automated code vulnerability detection has gained increasing attention in recent years. The deep learning (DL)-based methods, which implicitly learn vulnerable code patterns, have proven effective in vulnerability detection. The performance of DL-based methods usually relies on the quantity and quality of labeled data. However, the current labeled data are generally automatically collected, such as crawled from human-generated commits, making it hard to ensure the quality of the labels. Prior studies have demonstrated that the non-vulnerable code (i.e., negative labels) tends to be unreliable in commonly-used datasets, while vulnerable code (i.e., positive labels) is more determined. Considering the large numbers of unlabeled data in practice, it is necessary and worth exploring to leverage the positive data and large numbers of unlabeled data for more accurate vulnerability detection. In this paper, we focus on the Positive and Unlabeled (PU) learning problem for vulnerability detection and propose a novel model named PILOT, i.e., PositIve and unlabeled Learning mOdel for vulnerability deTection. PILOT only learns from positive and unlabeled data for vulnerability detection. It mainly contains two modules: (1) A distance-aware label selection module, aiming at generating pseudo-labels for selected unlabeled data, which involves the inter-class distance prototype and progressive fine-tuning; (2) A mixed-supervision representation learning module to further alleviate the influence of noise and enhance the discrimination of representations.

cs.SE↗

arXiv2018

TiEV: The Tongji Intelligent Electric Vehicle in the Intelligent Vehicle Future Challenge of China

TiEV is an autonomous driving platform implemented by Tongji University of China. The vehicle is drive-by-wire and is fully powered by electricity. We devised the software system of TiEV from scratch, which is capable of driving the vehicle autonomously in urban paths as well as on fast express roads. We describe our whole system, especially novel modules of probabilistic perception fusion, incremental mapping, the 1st and the 2nd planning and the overall safety concern. TiEV finished 2016 and 2017 Intelligent Vehicle Future Challenge of China held at Changshu. We show our experiences on the development of autonomous vehicles and future trends.

cs.RO↗