arXiv subjects

Zhongxin Liu

Publications and source records attributed to Zhongxin Liu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Inside the Skill Market: From Software Engineering Activities to Reusable Agent Skills

Software engineering (abbrev. SE) has continuously evolved through increasingly powerful forms of reuse, from source code and libraries to components and services. Recent advances in AI agents have introduced a potentially new reusable artifact: skills. Emerging agent skill repositories and marketplaces enable developers to package, share, and reuse SE expertise as reusable skills. This trend raises a fundamental question: what SE activities are being encapsulated into reusable skills? Existing studies primarily focus on a broad range of skills acquisition, safety, or benchmarking, while lacking a systematic understanding of SE-specific skills and their coverage across the software development lifecycle. To address this gap, we conduct the first large-scale empirical study of SE skills in public repositories and marketplaces. We collect and analyze a large corpus of SE skills, examining the activities they encapsulate, lifecycle coverage, evolution characteristics, and evaluation mechanisms. Our findings reveal that SE activities are increasingly becoming reusable artifacts via skills and suggest promising research opportunities for skill recommendation and engineering-oriented structuring, as well as the need for mechanisms to encapsulate high-context SE activities into reusable skills. Overall, our study provides the first activity-centric characterization of SE skills and reveals how SE activities are increasingly being transformed into reusable skills. These findings offer new insights into skill reuse, ecosystem development, and the future of agent-centric SE.

cs.SE↗

arXiv2026

An Iterative Test-and-Repair Framework for Competitive Code Generation

Large language models (LLMs) have made remarkable progress in code generation, but competitive programming remains a challenge. Recent training-based methods have improved code generation by using reinforcement learning (RL) with execution feedback. The more recent framework CURE further incorporates test generation into the training process, jointly training a Coder and a Tester within a single model. At inference time, the Coder generates many candidate programs, and the Tester generates tests from the problem description. The candidate who passes the most of the generated tests is selected as the final answer. However, CURE has two critical limitations. First, the Tester never reads any candidate code, so its tests often fail to expose implementation-specific bugs. Second, the Coder generates every candidate from scratch and never learns to fix a buggy program based on a failed test. To address these limitations, we propose FixAudit, which approaches competitive code generation from a new perspective: starting from a single initial candidate, it iteratively improves the candidate through a targeted test-and-repair debugging cycle. The framework trains one shared model with two specialized roles through four stages: the Fixer, which repairs the current candidate based on a failing test, and the Auditor, which reads the candidate code to generate new tests that expose its remaining bugs. We evaluate FixAudit on three benchmarks: APPS, CodeContests, and xCodeEval. Applied to a 7B model, the framework surpasses the average performance of the larger 32B baseline within the same model family under the zero-shot setting. Compared to strong baselines built on the same 7B base model, FixAudit improves average Pass@1 by 35.1% to 36.8% and average AvgPassRatio by 7.1% to 24.5%.

cs.SE↗

arXiv2026

ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?

Code generation is important in software engineering, and Reinforcement Learning with Verifiable Rewards (RLVR) is a powerful paradigm to improve it through execution-based feedback. However, most RLVR pipelines rely on human-curated tests, making progress bottlenecked by scarce and costly supervision. Existing work tried to use self-generated tests to ground rewards, but the lack of discriminative tests constrains the effect due to the sub-optimal performance of the model on test generation. We aim to improve code generation without ground-truth supervision by co-evolving code and test generation, so that their interactions yield progressively more informative supervision. To this end, we present ZeroCoder, a fully label-free co-evolutionary framework that jointly trains a Coder and a Tester using execution feedback from self-generated code-test interactions. For each problem, ZeroCoder executes sampled solutions against sampled tests to form a passing matrix, identifies a consensus subset of likely-correct solutions and consistent tests via a pluggable selection algorithm, and derives role-specific rewards. To ensure reward quality, ZeroCoder filters low-information instances via rank-based pre-filtering and trains the Tester with a curriculum balancing validity and mutation-driven discriminativeness. We further identify selector drift, the progressive miscalibration of fixed selection rules during co-evolution, and introduce DyB4, a Bayesian selector that uses as few as 10 labeled instances to recalibrate its priors dynamically. Across three models and six benchmarks, ZeroCoder consistently improves code generation and test generation. In the fully label-free setting, it improves code generation by up to 14.5% over the base model on Qwen2.5-Coder-7B-Instruct. With DyB4, the gain reaches 21.6%, while test generation improves by 24.3%, approaching oracle-supervised performance.

cs.SE↗

arXiv2026

iCoRe: An Iterative Correlation-Aware Retriever for Bug Reproduction Test Generation

Automatically generating bug reproduction tests (BRT) from issue descriptions is crucial for software maintenance. LLM-based approaches have shown great potential for this task. Their effectiveness heavily relies on retrieving high-quality context from the codebase. The retrieval phase of existing approaches relies on either traditional methods like BM25 or LLM-driven strategies. LLM-based retrieval strategies typically equip an LLM with tools to autonomously explore the repository or select the most relevant files and code snippets from a provided list as context. However, these retrieval methods suffer from three key limitations: 1) They often employ a unified strategy for retrieving both source code and test cases, overlooking their distinct retrieval requirements. 2) They focus solely on semantic similarity while ignoring function call relationships, leading to irrelevant context. 3) The retrieval lacks a feedback loop from the generation phase, preventing it from refining the context based on execution results. These limitations collectively result in low-quality context, thereby hindering the accuracy of bug reproduction. To address these challenges, we propose iCoRe, an iterative, correlation-aware context retrieval approach explicitly aware of three key correlations: 1) between source code and test cases, which requires differentiated retrieval, 2) between textual semantics and function call structures for accurate relevance assessment, and 3) between the retrieval and generation phases, which enables iterative feedback and refinement. To evaluate iCoRe, we integrate it with an LLM-based BRT generator and conduct a comprehensive evaluation on the SWT-bench Lite and TDD-bench Verified benchmarks. Experimental results show that our method achieves a Fail-to-Pass rate of 42.0% and 52.8% respectively, representing 19.7%-31.7% relative improvements over existing retrieval methods.

cs.SE↗

arXiv2026

Natural Language-Focused Software Engineering via Code-Documentation Equivalence

Source code documentation is an integral part of software development and maintenance, as it helps in understanding the code and facilitates communication among developers. However, existing documentation is often incomplete, outdated, or inaccurate, which can lead to misunderstandings and errors. In the era of large language models (LLMs), which are being extensively used for software engineering tasks, the quality of documentation becomes even more critical, as documentation provides important context for the models. In this paper, we introduce the notion of documentation-to-code equivalence, a novel property that captures whether documentation accurately and completely describes the code it documents. We present a novel approach, called Documentary, to automatically generate equivalent documentation for a given code snippet. Our evaluation shows that Documentary can generate equivalent documentation for 53.4% of the evaluated function-level code snippets. To show the benefits of documentation-to-code equivalence, we describe and evaluate two software engineering tasks: code understanding and code editing. Our results show that documentation-to-code equivalence allows an LLM to predict the output of a function with 12.8--24.5% higher accuracy, when compared to human-written documentation and documentation generated by a baseline approach. Furthermore, human developers consider documentation generated by Documentary to be more useful for understanding and editing code than the original human-written documentation.

cs.SE↗

arXiv2026

PracRepair: LLM-Empowered Automated Program Repair Inspired by Human-Like Debugging Practices

As software systems grow in scale and complexity, debugging and repair remain costly and time-consuming. Large language models (LLMs) have advanced automated program repair (APR), but existing LLM-based APR approaches still largely rely on static or retrieved context, error messages, and coarse-grained validation outcomes. As a result, they underutilize dynamic information for failure understanding and repair, including failure-execution dynamics and patch-validation dynamics. Effectively leveraging such information, however, is challenging: failure-execution traces are large and noisy, raw static-dynamic context is not self-explanatory, and patch-validation dynamics are often reduced to coarse feedback. To address these challenges, we propose \textsc{PracRepair}, a fully automated LLM-based APR framework inspired by human-like debugging practices. \textsc{PracRepair} constructs an on-demand static-dynamic context from buggy programs and failure executions, performs question-driven failure diagnosis to formulate explicit repair hypotheses, and iteratively refines candidate patches using validation diagnostics and trace-level behavioral changes. Experimental results on Defects4J V1.2 and V2.0 show that \textsc{PracRepair} consistently outperforms state-of-the-art baselines. Specifically, under GPT-3.5, \textsc{PracRepair} correctly fixes 139/136 bugs on Defects4J V1.2/V2.0, while under GPT-4o it further improves to 162/171. Moreover, \textsc{PracRepair} generalizes effectively to RWB (Real-World Bugs), achieving the best performance across multiple foundation models.

cs.SE↗

arXiv2026

Names Are All You Need: Effective and Safe Regression Test Selection for Python

Regression test selection reduces the cost of regression testing by executing only those tests affected by a code change. Despite extensive study of RTS in statically typed languages, achieving effective and safe RTS in Python is challenging. Python's dynamic typing makes precise call-graph construction difficult, which can cause call-graph-based RTS to miss affected tests. Python's eager importing mechanism, in contrast, renders file-level dependency analysis overly conservative. This paper presents NameRTS, the first Python RTS approach based on fine-grained dependency analysis. NameRTS models a Python program as a bipartite graph of code element nodes and name nodes, with edges capturing definitions and references. RTS is formulated as a reachability problem on this graph: a test is selected if any modified code element is reachable from the names used in that test. This design avoids call-graph construction, enabling a conservative analysis amenable to safety. To control dependency cascades introduced by coarse name matching, NameRTS applies two pruning strategies that leverage prior test executions and context information to refine name matching. To evaluate NameRTS, we construct the first Python RTS dataset with a ground truth indicating which test files are affected by each commit. We compare NameRTS with the best-performing baseline, BabelRTS, an RTS technique based on coarse file-level dependencies. On this benchmark, NameRTS skips 69.90% of test files on average, outperforming BabelRTS by 146.5%. It also reduces end-to-end testing time by 45.59%, yielding a 107.7% improvement over BabelRTS. In terms of safety, NameRTS selects all affected tests for 99.6% of commits, with only rare misses in exceptional cases. In contrast, BabelRTS is safe for 76.6% of commits. These results demonstrate the effectiveness of NameRTS, paving the way for more efficient regression testing in Python.

cs.SE↗

arXiv2026

ReCode: Reinforcing Code Generation with Reasoning-Process Rewards

In practice, rigorous reasoning is often a key driver of correct code, while Reinforcement Learning (RL) for code generation often neglects optimizing reasoning quality. Bringing process-level supervision into RL is appealing, but it faces two challenges. First, training reliable reward models to assess reasoning quality is bottlenecked by the scarcity of fine-grained preference data. Second, naively incorporating such neural rewards may suffer from reward hacking. This work proposes ReCode (Reasoning-Reinforced Code Generation), a novel RL training framework comprising: (1) Contrastive Reasoning-Process Reward Learning (CRPL), which trains a reward model with synthesized optimized and degraded reasoning variants to assess the quality of reasoning process; and (2) Consistency-Gated GRPO (CG-GRPO), which integrates the reasoning-process reward model into RL by gating neural reasoning-process rewards with strict execution outcomes, using execution correctness as a hard gate to mitigate reward hacking. Additionally, to assess the reward model's discriminative capability in assessing reasoning-process quality, we introduce LiveCodeBench-RewardBench (LCB-RB), a new benchmark comprising preference pairs of superior and inferior reasoning processes tailored for code generation. Experimental results across HumanEval(+), MBPP(+), LiveCodeBench, and BigCodeBench show that a 7B model trained with ReCode outperforms the base version by 16.1% and reaches performance comparable to GPT-4-Turbo. We further demonstrate the generalizability of ReCode by extending it to the math domain.

cs.SE↗

arXiv2026

DeepGuard: Secure Code Generation via Multi-Layer Semantic Aggregation

Large Language Models (LLMs) for code generation can replicate insecure patterns from their training data. To mitigate this, a common strategy for security hardening is to fine-tune models using supervision derived from the final transformer layer. However, this design may suffer from a final-layer bottleneck: vulnerability-discriminative cues can be distributed across layers and become less detectable near the output representations optimized for next-token prediction. To diagnose this issue, we perform layer-wise linear probing. We observe that vulnerability-related signals are most detectable in a band of intermediate-to-upper layers yet attenuate toward the final layers. Motivated by this observation, we introduce DeepGuard, a framework that leverages distributed security-relevant cues by aggregating representations from multiple upper layers via an attention-based module. The aggregated signal powers a dedicated security analyzer within a multi-objective training objective that balances security enhancement and functional correctness, and further supports a lightweight inference-time steering strategy. Extensive experiments across five code LLMs demonstrate that DeepGuard improves the secure-and-correct generation rate by an average of 11.9% over strong baselines such as SVEN. It also preserves functional correctness while exhibiting generalization to held-out vulnerability types. Our code is public at https://github.com/unknownhl/DeepGuard.

cs.SE↗

arXiv2026

Mitigating Implicit Inconsistencies in Patch Porting

Promptly porting patches from a source codebase to its variants (e.g., forks and branches) is essential for mitigating propagated defects and vulnerabilities. Recent studies have explored automated patch porting to reduce manual effort and delay, but existing approaches mainly handle inconsistencies visible in a patch's local context and struggle with those requiring global mapping knowledge between codebases. We refer to such non-local inconsistencies as implicit inconsistencies. Implicit inconsistencies pose greater challenges for developers to resolve due to their non-local nature. To address them, we propose MIP, which enables collaboration among an LLM, a compiler, and code analysis utilities. MIP adopts different strategies for different cases: when source identifiers exist in the target codebase, it leverages compiler diagnostics; otherwise, it retrieves matched code segment pairs from the two codebases as mapping knowledge for mitigation. Experiments on two representative scenarios, cross-fork and cross-branch patch porting, show that MIP successfully resolves more than twice as many patches as the best-performing baseline in both settings. A user study with our industry partner further demonstrates its practical effectiveness.

cs.SE↗

arXiv2026

ExecVerify: White-Box RL with Verifiable Stepwise Rewards for Code Execution Reasoning

Code LLMs still struggle with code execution reasoning, especially in smaller models. Existing methods rely on supervised fine-tuning (SFT) with teacher-generated explanations, primarily in two forms: (1) input-output (I/O) prediction chains and (2) natural-language descriptions of execution traces. However, intermediate execution steps cannot be explicitly verified during SFT, so the training objective can reduce to merely matching teacher explanations. Moreover, training data is typically collected without explicit control over task difficulty. We introduce ExecVerify, which goes beyond text imitation by incorporating verifiable white-box rewards derived from execution traces, including next-statement prediction and variable value/type prediction. Our work first builds a dataset with multiple difficulty levels via constraint-based program synthesis. Then, we apply reinforcement learning (RL) to reward correct answers about both intermediate execution steps and final outputs, aligning the training objective with semantic correctness at each execution step. Finally, we adopt a two-stage training pipeline that first enhances execution reasoning and then transfers to code generation. Experiments demonstrate that a 7B model trained with ExecVerify achieves performance comparable to 32B models on code reasoning benchmarks and improves pass@1 by up to 5.9\% on code generation tasks over strong post-training baselines.

cs.SE↗

arXiv2026

Balancing Latency and Accuracy of Code Completion via Local-Cloud Model Cascading

Line-level code completion requires a critical balance between high accuracy and low latency. Existing methods suffer from a trade-off: large language models (LLMs) provide high-quality suggestions but incur high latency, while small language models (SLMs) are fast but often suboptimal. We propose MCCom (Model-Cascading-based code Completion), a framework that cascades a local SLM with a cloud-based LLM. To achieve effective cascading, MCCom leverages user actions as a novel signal to trigger the LLM only when the SLM fails, significantly reducing cloud computation costs. Furthermore, we introduce a two-stage speculative decoding strategy and an iterative retrieval mechanism to enhance collaboration between the models. We also train a 121M-parameter lightweight model, which achieves 73.8% of the performance of a 7B state-of-the-art model. Evaluated on RepoEval and a new real-world benchmark StmtEval, MCCom reduces inference latency by up to 47.9% and LLM usage by 46.3%, while improving the LLM's exact match rate by 8.9% through effective collaboration.

cs.SE↗

arXiv2026

UniCoR: Modality Collaboration for Robust Cross-Language Hybrid Code Retrieval

Effective code retrieval is indispensable and it has become an important paradigm to search code in hybrid mode using both natural language and code snippets. Nevertheless, it remains unclear whether existing approaches can effectively leverage such hybrid queries, particularly in cross-language contexts. We conduct a comprehensive empirical study of representative code models and reveal three challenges: (1)insufficient semantic understanding; (2) inefficient fusion in hybrid code retrieval; and (3) weak generalization in cross-language scenarios. To address these challenges, we propose UniCoR, a novel self-supervised framework designed to learn Unified Code Representations that are semantically robust, modally collaborative, and language-agnostic. Firstly, we design a multi-perspective supervised contrastive learning module to enhance semantic understanding and modality fusion. It aligns representations from multiple perspectives, including code-to-code, natural language-to-code, and natural language-to-natural language, enforcing the model to capture a semantic essence among modalities. Secondly, we introduce a representation distribution consistency learning module to improve cross-language generalization, which explicitly aligns the feature distributions of different programming languages, enabling language-agnostic representation learning. Extensive experiments on both an empirical benchmark and a large-scale benchmark show that UniCoR outperforms all baseline models, achieving an average improvement of 8.64% in MRR and 11.54% in MAP over the best-performing baseline. Furthermore, UniCoR exhibits stability in hybrid code retrieval and generalization capability in cross-language scenarios.

cs.SE↗

arXiv2026

Forecasting Developer Environments with GenAI: A Research Perspective

Generative Artificial Intelligence (GenAI) models are achieving remarkable performance in various tasks, including code generation, testing, code review, and program repair. The ability to increase the level of abstraction away from writing code has the potential to change the Human-AI interaction within the integrated development environment (IDE). To explore the impact of GenAI on IDEs, 33 experts from the Software Engineering, Artificial Intelligence, and Human-Computer Interaction domains gathered to discuss challenges and opportunities at Shonan Meeting 222, a four-day intensive research meeting. Four themes emerged as areas of interest for researchers and practitioners.

cs.SE↗

arXiv2026

FGIT: Fault-Guided Fine-Tuning for Code Generation

Modern instruction-tuned large language models (LLMs) have made remarkable progress in code generation. However, these LLMs fine-tuned with standard supervised fine-tuning (SFT) sometimes generate plausible-looking but functionally incorrect code variants. This issue likely stems from the limitation of standard SFT, which treats all tokens equally during optimization and fails to emphasize the error-sensitive segments-specific code differences between correct implementations and similar incorrect variants. To address this problem, we propose Fault-Guided Fine-Tuning (FGIT), a novel fine-tuning technique that enhances LLMs' code generation by (1) extracting multi-granularity (line/token-level) differences between correct and incorrect yet similar implementations to identify error-sensitive segments, and (2) dynamically prioritizing those segments during training via dynamic loss weighting. Through extensive experiments on seven LLMs across three widely-used benchmarks, our method achieves an average relative improvement of 6.9% on pass@1 with some enhanced 6.7B LLMs outperforming closed-source models, e.g., GPT-3.5-Turbo. Furthermore, our fine-tuning technique demonstrates strong generalization with performance improvements ranging from 3.8% to 19.1% across diverse instruction-tuned LLMs, and our ablation studies confirm the contributions of different granularities of differences and hyperparameters.

cs.SE↗

arXiv2026

SolContractEval: A Benchmark for Evaluating Contract-Level Solidity Code Generation

The rise of blockchain has brought smart contracts into mainstream use, creating a demand for smart contract generation tools. While large language models (LLMs) excel at generating code in general-purpose languages, their effectiveness on Solidity, the primary language for smart contracts, remains underexplored. Solidity constitutes only a small portion of typical LLM training data and differs from general-purpose languages in its version-sensitive syntax and limited flexibility. These factors raise concerns about the reliability of existing LLMs for Solidity code generation. Critically, existing evaluations, focused on isolated functions and synthetic inputs, fall short of assessing models' capabilities in real-world contract development. To bridge this gap, we introduce SolContractEval, the first contract-level benchmark for Solidity code generation. It comprises 124 tasks drawn from real on-chain contracts across nine major domains. Each task input, consisting of complete context dependencies, a structured contract framework, and a concise task prompt, is independently annotated and cross-validated by experienced developers. To enable precise and automated evaluation of functional correctness, we also develop a dynamic evaluation framework based on historical transaction replay. Building on SolContractEval, we perform a systematic evaluation of six mainstream LLMs. We find that Claude-3.7-Sonnet achieves the highest overall performance, though evaluated models underperform relative to their capabilities on class-level generation tasks in general-purpose programming languages. Second, current models perform better on tasks that follow standard patterns but struggle with complex logic and inter-contract dependencies. Finally, they exhibit limited understanding of Solidity-specific features and contextual dependencies.

cs.SE↗

arXiv2025

Intention Chain-of-Thought Prompting with Dynamic Routing for Code Generation

Large language models (LLMs) exhibit strong generative capabilities and have shown great potential in code generation. Existing chain-of-thought (CoT) prompting methods enhance model reasoning by eliciting intermediate steps, but suffer from two major limitations: First, their uniform application tends to induce overthinking on simple tasks. Second, they lack intention abstraction in code generation, such as explicitly modeling core algorithmic design and efficiency, leading models to focus on surface-level structures while neglecting the global problem objective. Inspired by the cognitive economy principle of engaging structured reasoning only when necessary to conserve cognitive resources, we propose RoutingGen, a novel difficulty-aware routing framework that dynamically adapts prompting strategies for code generation. For simple tasks, it adopts few-shot prompting; for more complex ones, it invokes a structured reasoning strategy, termed Intention Chain-of-Thought (ICoT), which we introduce to guide the model in capturing task intention, such as the core algorithmic logic and its time complexity. Experiments across three models and six standard code generation benchmarks show that RoutingGen achieves state-of-the-art performance in most settings, while reducing total token usage by 46.37% on average across settings. Furthermore, ICoT outperforms six existing prompting baselines on challenging benchmarks.

cs.AI↗

arXiv2025

KSHSeek: Data-Driven Approaches to Mitigating and Detecting Knowledge-Shortcut Hallucinations in Generative Models

The emergence of large language models (LLMs) has significantly advanced the development of natural language processing (NLP), especially in text generation tasks like question answering. However, model hallucinations remain a major challenge in natural language generation (NLG) tasks due to their complex causes. We systematically expand on the causes of factual hallucinations from the perspective of knowledge shortcuts, analyzing hallucinations arising from correct and defect-free data and demonstrating that knowledge-shortcut hallucinations are prevalent in generative models. To mitigate this issue, we propose a high similarity pruning algorithm at the data preprocessing level to reduce spurious correlations in the data. Additionally, we design a specific detection method for knowledge-shortcut hallucinations to evaluate the effectiveness of our mitigation strategy. Experimental results show that our approach effectively reduces knowledge-shortcut hallucinations, particularly in fine-tuning tasks, without negatively impacting model performance in question answering. This work introduces a new paradigm for mitigating specific hallucination issues in generative models, enhancing their robustness and reliability in real-world applications.

cs.CL↗