arXiv subjects

Jialun Cao

Publications and source records attributed to Jialun Cao.

At least 19 recordsLinked to original sources

arXiv2026

When Is Graph Structure Worth Its Cost? The Case for Structure Pricing in Retrieval-Augmented Generation

Graph-based retrieval-augmented generation (RAG) can help answer questions that require information from many documents. However, building a graph often requires many language-model calls during ingestion. It is therefore important to ask whether its quality gains justify the additional cost. We present EffiRAG, a graph-based RAG system designed to reduce this cost. It uses the graph to locate relevant passages and generates answers from the original text. This design preserves source information while keeping graph construction and query processing lightweight. We evaluate EffiRAG on UltraDomain, which contains 120 open-ended questions from four domains. Compared with LightRAG-hybrid, EffiRAG produces the preferred answer on 93 questions. LightRAG is preferred on 7, and the remaining 20 are splits. EffiRAG also reduces total system cost by 57 percent, from USD 0.952 to USD 0.408. The cost includes language-model calls during ingestion and querying. The advantage remains as the corpus grows. At 10 and 20 documents per domain, EffiRAG uses a lightweight, non-LLM filter to skip low-salience chunks. It remains preferred over LightRAG-hybrid. It costs 4.2 times and 4.5 times less, respectively. The comparisons identify different quality-cost trade-offs. Graph-based RAG systems should therefore be evaluated by both answer quality and cost. The results favor graph structure that locates and preserves source evidence.

cs.AI↗

arXiv2026

GraphDroid: Asynchronous LLM-Based Mobile App GUI Testing via History-Aware Exploration and Hybrid Intent Fulfillment

Automated GUI testing is a widely adopted technique for ensuring mobile application quality by simulating user interactions to exercise functionalities. Despite the research breakthroughs in the past decades, covering complex functionalities that require multi-step action sequences still remains challenging. Traditional tools lack semantic understanding capability and can rarely synthesize such action sequences. Recent LLM-based tools can generate test intents describing target functionalities and leverage the LLM to fulfill the intents, but suffer from three key limitations: 1) loss of historical context for identifying uncovered functionalities, 2) synchronous intent generation that blocks exploration, and 3) per-step LLM-driven fulfillment incurring high cost and latency. To address these limitations, we propose GraphDroid, an intent-driven GUI testing framework that integrates a cluster-based memory mechanism to effectively identify uncovered functionalities from historically visited states for comprehensive application testing. For improving testing efficiency, GraphDroid adopts an asynchronous intent generation paradigm that eliminates the latency bottleneck and a hybrid intent fulfillment strategy that reserves the LLM for fulfilling complex intents while delegating simple intents to a lightweight heuristic algorithm. We evaluate GraphDroid on 41 real-world Android apps against six state-of-the-art baselines. Results show that GraphDroid outperforms all baselines, achieving up to 36.4% higher code coverage while incurring less than one eighth of the cost of the best pure LLM-based baseline. GraphDroid also exposes 19 bugs in the 41 apps and detects 13 of 52 crashes in the Themis bug benchmark, surpassing all the six baselines. Seven of the 19 bugs were previously unknown and we reported them to the developers. So far, four bugs have been confirmed and fixed.

cs.SE↗

arXiv2026

SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring

As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.

cs.CL↗

arXiv2026

Doc2Feat-Bench: Evaluating Documentation-Driven Feature Addition

Documentation changes in mature software projects often describe newly introduced or modified behavior. This makes them a natural basis for documentation-driven feature addition, where software engineering agents implement features from public-facing documentation updates. However, existing benchmarks for agentic software development primarily rely on issue reports written for maintainer coordination and often contain implementation-level details. Motivated by this observation, this work introduces Doc2Feat-bench, a benchmark for documentation-driven feature addition. Doc2Feat-bench consists of 634 tasks across 10 mature open-source software projects, involving about 114k code changes in total. Each task pairs a documentation change with the corresponding implementation and developer-written tests, enabling evaluation of whether a software engineering agent can implement the feature described by the documentation diff. The benchmark is constructed through a five-phase pipeline, starting from release notes, and offers broad coverage across libraries, developer tools, and frameworks while preserving real-world development settings. To facilitate lightweight and reliable evaluation under limited resources, we further curate a human-validated subset named Doc2Feat-bench Verified. It contains 114 high-quality instances whose task clarity and evaluation validity are manually verified. We use Doc2Feat-bench to assess a range of state-of-the-art software engineering agents with two widely used scaffolds. Experimental results show that despite significant token consumption, the best task success rate remains as low as 37.72% (OpenHands with Qwen3-Coder-480B). Our analysis reveals that software engineering agents face key challenges in performing cross-file edits, understanding existing code modules, and correctly interpreting documentation semantics.

cs.SE↗

arXiv2026

Understanding Agent-Reactive Bugs at the Model-Harness Boundary: An Empirical Study of LLM Agent Issue Reports

LLM agents span command-line interfaces (e.g., Codex) and agent frameworks (e.g., LangChain), integrating backend LLMs with harness code that parses model outputs, controls agent loops, and manages context. Both the harness and LLM-generated responses jointly shape an agent's execution. This architecture gives rise to bugs that cannot be readily understood by inspecting either component alone, because some bugs occur only when a particular LLM response elicits an abnormal reaction from the agent. Prior empirical studies of agent bugs have largely attributed failures either to limited model capabilities or to harness-side defects, such as outdated APIs and configuration misalignment, without characterizing these AR bugs. We conduct the first empirical study focused on agent-reactive (AR) bugs. Through manual analysis of 255 bug reports from Codex, Gemini-CLI, LangChain, and CrewAI, we construct a two-axis taxonomy covering observable symptoms and the LLM behaviors that trigger them. Our findings show that many AR bugs manifest as silent errors without well-defined test oracles, which makes detection difficult. The stochasticity of LLM responses further complicates bug reproduction. We additionally examine fixes proposed by users and implemented by developers. This analysis exposes a mismatch: users frequently advocate harness-side guardrails, whereas developers may attribute the issue to the LLM or respond slowly to user-proposed fixes. These findings point to the need for mechanisms that help users and developers understand the root causes and resolutions of AR bugs. Overall, the study highlights challenges specific to LLM agents and motivates the design of test oracles, reproduction support, and fault-localization techniques for AR bugs.

cs.SE↗

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

Code Benchmarks Should Prioritize Rigor, Reliability, and Reproducibility

Code-related benchmarks play a critical role in evaluating large language models (LLMs), yet their quality fundamentally shapes how the community interprets model capabilities. In the past few years, awareness of benchmark quality has grown. Yet, after a decade-scale (2014-2025) survey over 672 code benchmarks, we observed a lag between growing awareness and actual practice. For example, in 2025 alone, the number of benchmarks that ignore code coverage when providing test cases nearly matches the total count accumulated across the previous ten years. In response, we take a clear position: Code benchmarks must prioritize rigor in benchmark construction, reliability in evaluation, and reproducibility in release. To operationalize this position, we introduce a code benchmark guideline HOW2BENCH with 55 checklists. Finally, our further human study also exposed that the current issues not only stem from the significant effort required, but also from a lack of awareness regarding their importance.

cs.SE↗

arXiv2026

Entropy Regularization Improves Policy Robustness in Continuous-Time Reinforcement Learning

Entropy regularization is widely used in continuous-time reinforcement learning (RL) to reduce sensitivity to environmental perturbations, yet its robustness benefits lack a rigorous theoretical foundation. This paper establishes the first robustness guarantees for entropy-regularized continuous-time Markov decision processes. We show that maximizing an entropy-regularized objective yields a lower bound on a worst-case robust RL problem with joint reward and transition perturbations. We analytically characterize the induced robust sets and prove that they expand monotonically with the regularization strength, justifying the empirical observation that stronger entropy improves robustness. In contrast to prior discrete-time analyses, our results remove the intractable state-distribution entropy term and provide guarantees invariant to action frequency. Experiments on queueing network control and market making confirm our theory, showing that entropy-regularized policies outperform greedy and $ε$-greedy baselines under dynamics perturbations.

math.OC↗

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

Combinatorial Synthesis: Scaling Code RLVR via Atomic Decomposition and Recombination

Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as the cornerstone for shaping the remarkable coding abilities of Large Language Models (LLMs). However, the scalability of RLVR is severely constrained by the scarcity of sufficiently challenging verifiable code tasks that target near the model's edge of competence. Prior studies often rely on heuristic seed expansions for data synthesis, which severely limits both novelty and difficulty. Consequently, the training value of such data fails to scale proportionally with the size of its synthesis. To this end, we propose Atomic Decomposition and Recombination (ADR), a novel framework that generates verifiable code tasks via decomposition into atomic elements and controlled recombination, thereby enabling the generation of genuinely novel and challenging verifiable code tasks. Experiments and analysis demonstrate that ADR achieves superior originality, difficulty, diversity, and test quality over existing baselines, and consistently delivers greater improvements in code ability across RLVR in diverse downstream domains, including algorithmic programming, tool usage, and data science. Our work sheds light on a new paradigm for novel code task synthesis and scalable RLVR training.

cs.CL↗

arXiv2026

Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training

Continual post-training aims to extend large language models (LLMs) with new knowledge, skills, and behaviors, yet it remains unclear when sequential updates enable capability transfer and when they cause catastrophic forgetting. Existing methods mitigate forgetting through sequential fine-tuning, replay, regularization, or model merging, but offer limited criteria for determining when incorporating new updates is beneficial or harmful. In this work, we study LLM continual post-training through three questions: What drives forgetting? When do sequentially acquired capabilities transfer or interfere? How can compatibility be used to control update integration? We address these questions through task geometry: we represent each post-training task by its parameter update and study the covariance geometry induced by the update. Our central finding is that: forgetting can be considered as a state-relative update-integration failure, it arises when the covariance geometries induced by tasks misalign with the geometry of the evolving model state. Sequential updates transfer when they remain compatible with the model state shaped by previous updates, and interfere when state-relative geometry conflict becomes high. Motivated by this finding, we propose Geometry-Conflict Wasserstein Merging (GCWM), a data-free update-integration method that constructs a shared Wasserstein metric via Gaussian Wasserstein barycenters and uses geometry conflict to gate geometry-aware correction. Across Qwen3 0.6B--14B on domain-continual and capability-continual settings, GCWM consistently outperforms data-free baselines, improving retention and final performance without replay data. These results identify geometry conflict as both an explanatory signal for forgetting and a practical control signal for LLM continual post-training.

cs.LG↗

arXiv2026

LiveFMBench: Unveiling the Power and Limits of Agentic Workflows in Specification Generation

Formal specification is essential for rigorous program verification, yet writing correct specifications remains costly and difficult to automate. Although large language models (LLMs) and agents have shown promising progress, their true capabilities and failure modes remain unclear. We present the first systematic and contamination-aware study of LLM- and agent-based formal specification generation for C programs. We introduce LiveFMBench, a continuously evolving benchmark of 630 ACSL (ANSI/ISO C Specification Language)-annotated C programs, including 360 newly collected cases designed to mitigate data leakage. Using this benchmark, we evaluate direct prompting with different sampling sizes, reasoning-enabled (thinking mode) inference, the agentic pipeline, and perform a fine-grained failure analysis. Experimental results reveal that naive evaluation substantially overestimates performance because models under direct prompting may exhibit unfaithful behaviors, such as deceiving automated provers or ignoring code-context constraints; after excluding such cases, the true specification generation accuracy drops by approximately 20\%. We further find that both increased sampling and thinking mode significantly improve success rates, with smaller models benefiting more from thinking mode. Agentic pipelines are particularly effective under low sampling budgets and on harder datasets. Failure analysis further shows that incorrect loop invariants are the dominant error type, while agentic pipelines notably reduce assertion errors. These results expose fundamental limitations in current LLM-based approaches and suggest they remain far from replacing human-authored formal specifications. We release LiveFMBench at https://huggingface.co/datasets/fm-universe/Live-FM-Bench and all evaluation artifacts to support future research.

cs.SE↗

arXiv2026

Across Programming Language Silos: A Study on Cross-Lingual Retrieval-augmented Code Generation

Current research on large language models (LLMs) with retrieval-augmented code generation (RACG) has largely focused on single-language settings, leaving their cross-lingual effectiveness underexplored. Multilingual RACG systems are increasingly important for migrating and reusing code across programming languages (PLs), a common yet challenging task in modern software development. To systematically study cross-lingual code knowledge transfer in RACG, we construct a dataset covering 13 PLs with nearly 14K instances. Our experiments reveal three key insights: (1) Knowledge transfer in RACG across PLs is non-trivial even using direct injection. (2) RACG exhibits unequal cross-lingual knowledge transfer, and its efficacy depends on linguistic affinity of PL pair and diversity of LLM pretraining corpus. (3) RACG shows limited reliance on natural language information embedded in code when equipped with a code-specific retriever. These findings provide practical guidance for designing effective multilingual RACG systems. https://github.com/icip-cas/Cross-Lingual-RACG

cs.SE↗

arXiv2026

MR-Adopt: Automatic Deduction of Input Transformation Function for Metamorphic Testing

While a recent study reveals that many developer-written test cases can encode a reusable Metamorphic Relation (MR), over 70% of them directly hard-code the source input and follow-up input in the encoded relation. Such encoded MRs, which do not contain an explicit input transformation to transform the source inputs to corresponding follow-up inputs, cannot be reused with new source inputs to enhance test adequacy. In this paper, we propose MR-Adopt (Automatic Deduction Of inPut Transformation) to automatically deduce the input transformation from the hard-coded source and follow-up inputs, aiming to enable the encoded MRs to be reused with new source inputs. With typically only one pair of source and follow-up inputs available in an MR-encoded test case as the example, we leveraged LLMs to understand the intention of the test case and generate additional examples of source-followup input pairs. This helps to guide the generation of input transformations generalizable to multiple source inputs. Besides, to mitigate the issue that LLMs generate erroneous code, we refine LLM-generated transformations by removing MR- irrelevant code elements with data-flow analysis. Finally, we assess candidate transformations based on encoded output relations and select the best transformation as the result. Evaluation results show that MR-Adopt can generate input transformations applicable to all experimental source inputs for 72.00% of encoded MRs, which is 33.33% more than using vanilla GPT-3.5. By incorporating MR- Adopt-generated input transformations, encoded MR-based test cases can effectively enhance the test adequacy, increasing the line coverage and mutation score by 10.62% and 18.91%, respectively.

cs.SE↗

arXiv2026

Can Large Language Models Model Programs Formally?

In the digital age, ensuring the correctness, safety, and reliability of software through formal verification is paramount, particularly as software increasingly underpins critical infrastructure. Formal verification, split into theorem proving and model checking, provides a feasible and reliable path. Unlike theorem proving, which yields notable advances, model checking has been less focused due to the difficulty of automatic program modeling. To fill this gap, we introduce Model-Bench, a benchmark and an accompanying pipeline for evaluating and improving LLMs' program modeling capability by modeling Python programs into verification-ready model checking specifications checkable by its accompanying model checker. Model-Bench comprises 400 Python programs derived from three well-known benchmarks (HumanEval, MBPP, and LiveCodeBench). Our extensive experiments reveal significant limitations in LLMs' program modeling and further provide inspiring directions.

cs.SE↗

arXiv2026

SWE-ABS: Adversarial Benchmark Strengthening Exposes Inflated Success Rates on Test-based Benchmark

The SWE-Bench Verified leaderboard is approaching saturation, with the top system achieving 78.80%. However, we show that this performance is inflated. Our re-evaluation reveals that one in five "solved" patches from the top-30 agents are semantically incorrect, passing only because weak test suites fail to expose their errors. We present SWE-ABS, an adversarial framework that strengthens test suites through a two-stage pipeline: (1) coverage-driven augmentation using program slicing to target untested code regions, and (2) mutation-driven adversarial testing that synthesizes plausible but incorrect patches to expose semantic blind spots. On SWE-Bench Verified (500 instances), SWE-ABS strengthens 50.2% of instances, a 25.1x improvement over prior work, and rejects 19.71% of previously passing patches. As a result, the top agent's score decreases from 78.80% to 62.20%, leading to significant leaderboard reshuffling, with the previous top-ranked agent dropping to fifth place.

cs.SE↗

arXiv2026

FeedbackEval: A Benchmark for Evaluating Large Language Models in Feedback-Driven Code Repair Tasks

Code repair is a fundamental task in software development, facilitating efficient bug resolution and software maintenance. Although large language models (LLMs) have demonstrated considerable potential in automated code repair, their ability to comprehend and leverage diverse types of feedback, which is crucial for iterative self-correction in authentic debugging scenarios, remains insufficiently understood. To bridge this gap, we introduce FeedbackEval, a systematic benchmark constructed from three heterogeneous sources (HumanEval, CoderEval, and SWE-Bench-verified), to evaluate LLMs' feedback comprehension and code repair performance. We conduct a comprehensive empirical study on five state-of-the-art LLMs, including GPT-4o, Claude-3.5, Deepseek-R1, GLM-4, and Qwen2.5, to evaluate their behavior under both single-iteration and iterative code repair settings. Our results show that mixed feedback yields the highest repair success (63.6%), with LLM-Expert and test feedback providing strong targeted gains (62.9% and 57.9%, respectively), while minimal (53.1%) and compiler feedback (49.2%) offer moderate benefits and LLM-Skilled proves least effective (48.8%). Iterative feedback further enhances repair performance, though the marginal benefit diminishes after two or three iterations. Moreover, prompt structure is shown to be critical: structured reasoning (RR, CoT) and dynamic example selection deliver notable improvements, whereas removing semantic cues such as docstrings or role-play causes severe degradation. This work introduces a robust benchmark and delivers practical insights to advance the understanding and development of feedback-driven code repair using LLMs.

cs.SE↗

arXiv2026

Can Emulating Semantic Translation Help LLMs with Code Translation? A Study Based on Pseudocode

Although large language models (LLMs) show promising potential in code translation, they still struggle to generate accurate translations using the commonly adopted direct code-to-code translation approach, which converts an original program into the target programming language (PL) in a single step. Inspired by the success of incorporating intermediate steps to guide LLMs in resolving challenging tasks, in this study, we explore pseudocode-based code translation. This approach emulates human semantic translation by first interpreting the original program's intent and logic into pseudocode and then implementing it in the target PL. To understand the effectiveness of this underexplored approach, we present a systematic empirical study on pseudocode-based code translation, aiming to investigate its helpfulness in enhancing the direct translation approach, illuminate its effective usage, and identify its limitations. By comparing direct and pseudocode-based translation on 9,690 translation tasks across six PLs with five popular LLMs, we found that pseudocode-based translation can effectively complement direct translation, particularly when translating from flexible to rigid PLs and handling a low-training-resource PL. Based on the findings, we suggest combining the translation results of both approaches for test-based selection to leverage their complementary strengths. We also reveal the advantages of pseudocode-based translation in decoupling the code understanding and generation burden on complicated programs and mitigating distractions from PL-specific implementations in original programs, as well as its limitations due to incorrect, incomplete, or ambiguous pseudocode. Our study sheds light on the effective use of pseudocode-based translation and provides evidence to help enhance LLMs in code translation.

cs.SE↗