arXiv subjects

Zhiwei Fei

Publications and source records attributed to Zhiwei Fei.

12 recordsLinked to original sources

arXiv2026

OpenCompass: A Universal Evaluation Platform for Large Language Models

In recent years, the field of artificial intelligence has undergone a paradigm shift from task-specific small-scale models to general-purpose large language models (LLMs). With the rapid iteration of LLMs, objective, quantitative, and comprehensive evaluation of their capabilities has become a critical link in advancing technological development. Currently, the mainstream static benchmark dataset-based evaluation methods face challenges such as the diversity of task types, inconsistent evaluation criteria, and fragmentation of data and processing workflows, making it difficult to efficiently conduct cross-domain and large-scale model evaluation. To address the aforementioned issues, this paper proposes and open-sources OpenCompass, a one-stop, scalable, and high-concurrency-supported general-purpose LLM evaluation platform. Adhering to the design philosophy of modularization and component decoupling, the platform boasts three core advantages: high compatibility, flexibility, and high concurrency. The core architecture of OpenCompass comprises five key components: the Configuration System, Task Partitioning Module, Execution and Scheduling Module, Task Execution Unit, and Result Visualization Module. Its workflow provides rule-based, LLM-as-a-Judge, and cascaded evaluators to adapt to the requirements of different task scenarios. Supporting mainstream benchmark datasets across multiple domains, including knowledge, reasoning, computation, science, language, code, etc., the platform offers a unified and efficient LLM evaluation tool for both academia and industry, facilitating the accurate identification of strengths and weaknesses of LLMs as well as their subsequent optimization.

cs.CL↗

arXiv2026

SafeTune: Search-based Harmfulness Minimisation for Large Language Models

The widespread adoption of Large Language Models (LLMs) raises concerns about the potential harmfulness of their responses. In this paper, we first investigate the harmfulness of responses from four general-purpose LLMs. Next, we propose SafeTune, a multi-objective search-based approach to mitigate harmfulness while increasing response relevance through hyperparameter tuning and system prompt engineering. Our initial evaluation shows that SafeTune significantly reduces the rate of harmful responses generated by Qwen3.5 0.8B and increases prompt-response relevance (both with a large effect size). Among the parameters we explore, we also find that encouraging greater repetition in responses is most impactful in reducing harmfulness while increasing relevance.

cs.SE↗

arXiv2026

SkillMOO: Multi-Objective Optimization of Agent Skills for Software Engineering

Agent skills are increasingly used to configure coding agents for software engineering (SE) tasks, yet current practice treats them as static, hand-crafted assets, or evolved on pass rate alone. This is insufficient: a skill can improve task success while substantially raising token cost, or introducing misleading guidance. We argue that SE agent skill bundles can be treated as multi-objective search objects and present SkillMOO, a framework that evolves skill bundles through LLM-proposed edits and NSGA-II Pareto selection on pass rate and inference cost. Evaluated across all 16 SkillsBench SE tasks, SkillMOO achieves the top pass rate rank on 11 of 12 non-zero-pass tasks while achieving cost reductions of up to 31.7% over static bundles, with pass rate gains up to 21 percentage points. Analysis of 38 skill edits shows that pruning and substitution dominate successful operations, offering actionable principles for skill bundle design. Thereby, the current practice of deploying skills without cost-aware validation leaves better skill configurations unexplored, motivating a new class of cost-aware, search-based skill engineering.

cs.SE↗

arXiv2026

Echo: Graph-Enhanced Retrieval and Execution Feedback for Issue Reproduction Test Generation

Identifying the root cause of a bug remains difficult for many developers because bug reports often lack a bug reproducing test case that reliably triggers the failure. Manually writing such test cases is time-consuming and requires substantial effort to understand the codebase and isolate the failing behavior. To address this challenge, we propose Echo, an agent for generating issue reproducing test cases, which advances previous work in several ways. During generation, Echo strengthens context retrieval by leveraging a code graph and a novel automatic query-refinement strategy. Echo also improves upon previous tools by automatically executing generated test cases, a first-of-its-kind feature that seamlessly integrates into practical development workflows. In addition, Echo generates potential patches and uses the patched version to validate whether a candidate test meets the fail-to-pass criterion and to provide actionable feedback for refinement. Unlike prior bug-reproduction agents that sample and rank multiple candidate tests, Echo generates a single test per issue, offering a better cost-performance trade-off. Experiments on SWT-Bench Verified show that Echo establishes a new state of the art among open-source approaches, achieving a 66.28% success rate.

cs.SE↗

arXiv2026

JMigBench: A Benchmark for Evaluating LLMs on Source Code Migration (Java 8 to Java 11)

We build a benchmark to evaluate large language models (LLMs) for source code migration tasks, specifically upgrading functions from Java 8 to Java 11. We first collected a dataset of function pairs from open-source repositories, but limitations in data quality led us to construct a refined dataset covering eight categories of deprecated APIs. Using this dataset, the Mistral Codestral model was evaluated with CodeBLEU and keyword-based metrics to measure lexical and semantic similarity as well as migration correctness. Results show that the evaluated model (Mistral Codestral) can handle trivial one-to-one API substitutions with moderate success, achieving identical migrations in 11.11% of the cases, but it struggles with more complex migrations such as CORBA or JAX-WS. These findings suggest Mistral Codestral can partially reduce developer effort by automating repetitive migration tasks but cannot yet replace humans within the scope of the JMigBench benchmark. The benchmark and analysis provide a foundation for future work on expanding datasets, refining prompting strategies, and improving migration performance across different LLMs.

cs.SE↗

arXiv2026

Environment-in-the-Loop: Rethinking Code Migration with LLM-based Agents

Modern software systems continuously undergo code upgrades to enhance functionality, security, and performance, and Large Language Models (LLMs) have demonstrated remarkable capabilities in code migration tasks. However, while research on automated code migration which including refactoring, API adaptation, and dependency updates has advanced rapidly, the exploration of the automated environment interaction that must accompany it remains relatively scarce. In practice, code and its environment are intricately intertwined. Relying solely on static analysis of the environment leads to an inadequate understanding of the target setting, prolongs feedback cycles, and consequently causes significant rework and project delays, thereby reducing overall efficiency. We contend that successful software evolution demands a holistic perspective that integrates both code and environment migration. To understand the current landscape and challenges, we first provide an overview of the status of automated environment construction. We then propose a novel framework paradigm that tightly integrates automated environment setup with the code migration workflow. Finally, we explore the challenges and future directions for automated environment interaction within the code migration domain. Our findings emphasize that without automated environment interaction, the automation of code migration is only half complete.

cs.SE↗

arXiv2025

From Code Foundation Models to Agents and Applications: A Comprehensive Survey and Practical Guide to Code Intelligence

Large language models (LLMs) have fundamentally transformed automated software development by enabling direct translation of natural language descriptions into functional code, driving commercial adoption through tools like Github Copilot (Microsoft), Cursor (Anysphere), Trae (ByteDance), and Claude Code (Anthropic). While the field has evolved dramatically from rule-based systems to Transformer-based architectures, achieving performance improvements from single-digit to over 95\% success rates on benchmarks like HumanEval. In this work, we provide a comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs, systematically examining the complete model life cycle from data curation to post-training through advanced prompting paradigms, code pre-training, supervised fine-tuning, reinforcement learning, and autonomous coding agents. We analyze the code capability of the general LLMs (GPT-4, Claude, LLaMA) and code-specialized LLMs (StarCoder, Code LLaMA, DeepSeek-Coder, and QwenCoder), critically examining the techniques, design decisions, and trade-offs. Further, we articulate the research-practice gap between academic research (e.g., benchmarks and tasks) and real-world deployment (e.g., software-related code tasks), including code correctness, security, contextual awareness of large codebases, and integration with development workflows, and map promising research directions to practical needs. Last, we conduct a series of experiments to provide a comprehensive analysis of code pre-training, supervised fine-tuning, and reinforcement learning, covering scaling law, framework selection, hyperparameter sensitivity, model architectures, and dataset comparisons.

cs.SE↗

arXiv2025

Understanding Code Agent Behaviour: An Empirical Study of Success and Failure Trajectories

The increasing deployment of Large Language Model (LLM) agents for complex software engineering tasks has created a need to understand their problem-solving behaviours beyond simple success metrics. While these agents demonstrate impressive capabilities in automated issue resolution, their decision-making processes remain largely opaque. This paper presents an empirical study of agent trajectories, namely the execution traces capturing the steps agents take when attempting to resolve software issues. We analyse trajectories from three state-of-the-art code agents (OpenHands, SWE-agent, and Prometheus) on the SWE-Bench benchmark, examining both successful and failed attempts. Our investigation reveals several key insights into agent behaviour. First, we identify how distinct problem-solving strategies, such as defensive programming and context gathering, enable success in different scenarios. Second, we find that failed trajectories are consistently longer and exhibit higher variance than successful ones, with failure patterns differing significantly between agents. Third, our fault localisation analysis shows that while most trajectories correctly identify problematic files (72-81\% even in failures), success depends more on achieving approximate rather than exact code modifications. These and other findings unveiled by our study, provide a foundation for understanding agent behaviour through trajectory analysis, contributing to the development of more robust and interpretable autonomous software engineering systems.

cs.SE↗

arXiv2025

SkipGPT: Dynamic Layer Pruning Reinvented with Token Awareness and Module Decoupling

Large language models (LLMs) achieve remarkable performance across tasks but incur substantial computational costs due to their deep, multi-layered architectures. Layer pruning has emerged as a strategy to alleviate these inefficiencies, but conventional static pruning methods overlook two critical dynamics inherent to LLM inference: (1) horizontal dynamics, where token-level heterogeneity demands context-aware pruning decisions, and (2) vertical dynamics, where the distinct functional roles of MLP and self-attention layers necessitate component-specific pruning policies. We introduce SkipGPT, a dynamic layer pruning framework designed to optimize computational resource allocation through two core innovations: (1) global token-aware routing to prioritize critical tokens, and (2) decoupled pruning policies for MLP and self-attention components. To mitigate training instability, we propose a two-stage optimization paradigm: first, a disentangled training phase that learns routing strategies via soft parameterization to avoid premature pruning decisions, followed by parameter-efficient LoRA fine-tuning to restore performance impacted by layer removal. Extensive experiments demonstrate that SkipGPT reduces over 40% of model parameters while matching or exceeding the performance of the original dense model across benchmarks. By harmonizing dynamic efficiency with preserved expressivity, SkipGPT advances the practical deployment of scalable, resource-aware LLMs. Our code is publicly available at: https://github.com/EIT-NLP/SkipGPT.

cs.CL↗

arXiv2024

InternLM-Law: An Open Source Chinese Legal Large Language Model

While large language models (LLMs) have showcased impressive capabilities, they struggle with addressing legal queries due to the intricate complexities and specialized expertise required in the legal field. In this paper, we introduce InternLM-Law, a specialized LLM tailored for addressing diverse legal queries related to Chinese laws, spanning from responding to standard legal questions (e.g., legal exercises in textbooks) to analyzing complex real-world legal situations. We meticulously construct a dataset in the Chinese legal domain, encompassing over 1 million queries, and implement a data filtering and processing pipeline to ensure its diversity and quality. Our training approach involves a novel two-stage process: initially fine-tuning LLMs on both legal-specific and general-purpose content to equip the models with broad knowledge, followed by exclusive fine-tuning on high-quality legal data to enhance structured output generation. InternLM-Law achieves the highest average performance on LawBench, outperforming state-of-the-art models, including GPT-4, on 13 out of 20 subtasks. We make InternLM-Law and our dataset publicly available to facilitate future research in applying LLMs within the legal domain.

cs.CL↗

arXiv2024

MathBench: Evaluating the Theory and Application Proficiency of LLMs with a Hierarchical Mathematics Benchmark

Recent advancements in large language models (LLMs) have showcased significant improvements in mathematics. However, traditional math benchmarks like GSM8k offer a unidimensional perspective, falling short in providing a holistic assessment of the LLMs' math capabilities. To address this gap, we introduce MathBench, a new benchmark that rigorously assesses the mathematical capabilities of large language models. MathBench spans a wide range of mathematical disciplines, offering a detailed evaluation of both theoretical understanding and practical problem-solving skills. The benchmark progresses through five distinct stages, from basic arithmetic to college mathematics, and is structured to evaluate models at various depths of knowledge. Each stage includes theoretical questions and application problems, allowing us to measure a model's mathematical proficiency and its ability to apply concepts in practical scenarios. MathBench aims to enhance the evaluation of LLMs' mathematical abilities, providing a nuanced view of their knowledge understanding levels and problem solving skills in a bilingual context. The project is released at https://github.com/open-compass/MathBench .

cs.CL↗

arXiv2023

LawBench: Benchmarking Legal Knowledge of Large Language Models

Large language models (LLMs) have demonstrated strong capabilities in various aspects. However, when applying them to the highly specialized, safe-critical legal domain, it is unclear how much legal knowledge they possess and whether they can reliably perform legal-related tasks. To address this gap, we propose a comprehensive evaluation benchmark LawBench. LawBench has been meticulously crafted to have precise assessment of the LLMs' legal capabilities from three cognitive levels: (1) Legal knowledge memorization: whether LLMs can memorize needed legal concepts, articles and facts; (2) Legal knowledge understanding: whether LLMs can comprehend entities, events and relationships within legal text; (3) Legal knowledge applying: whether LLMs can properly utilize their legal knowledge and make necessary reasoning steps to solve realistic legal tasks. LawBench contains 20 diverse tasks covering 5 task types: single-label classification (SLC), multi-label classification (MLC), regression, extraction and generation. We perform extensive evaluations of 51 LLMs on LawBench, including 20 multilingual LLMs, 22 Chinese-oriented LLMs and 9 legal specific LLMs. The results show that GPT-4 remains the best-performing LLM in the legal domain, surpassing the others by a significant margin. While fine-tuning LLMs on legal specific text brings certain improvements, we are still a long way from obtaining usable and reliable LLMs in legal tasks. All data, model predictions and evaluation code are released in https://github.com/open-compass/LawBench/. We hope this benchmark provides in-depth understanding of the LLMs' domain-specified capabilities and speed up the development of LLMs in the legal domain.

cs.CL↗