arXiv subjects

Weiyu Kong

Publications and source records attributed to Weiyu Kong.

3 recordsLinked to original sources

arXiv2026

Compiling Large Multi-Modal Requirement Documents into Runnable Software Systems: From an Agentic Test-Driven Perspective

Large Language Models (LLMs) have significantly improved programming efficiency by translating natural language into code, yet their performance deteriorates when handling large-scale, multi-modal requirement documents containing hundreds of scenarios, often producing incorrect implementations or missing critical constraints. To address this challenge, we propose ARC (Agentic Requirement Compilation), a framework that compiles DSL-based requirement documents into runnable software systems while automatically generating modular software architecture, comprehensive test suites, and traceability across requirements, design, and code. ARC adopts a bidirectional test-driven agentic workflow, combining a top-down architecture design phase with a bottom-up implementation phase to ensure that generated code satisfies synthesized tests. We evaluate ARC on six runnable web system benchmarks and the AppForge benchmark of 101 mobile app generation tasks. Across three independent trials, ARC consistently outperforms state-of-the-art LLM-based baselines, achieving 50.6% more GUI tests passed on average for web systems, a 100% compilation success rate, and a 68.3% test pass rate on AppForge. A user study with 21 participants further shows that users with limited programming experience can write DSL-based requirement documents containing up to 174 scenarios within an average of 5.6 hours to generate maintainable runnable systems, including a real-world ticket-booking application of approximately 10K lines of code.

cs.SE↗

arXiv2026

XSearch: Explainable Code Search via Concept-to-Code Alignment

Semantic code search has been widely adopted in both academia and industry. These approaches embed natural-language queries and code snippets into a shared embedding space and retrieve results based on vector similarity. Despit strong performance on benchmark datasets, they often suffer from poor explainability and generalization. Retrieved code may appear semantically similar yet miss critical functional requirements of the query, while providing no explanation of why the result was retrieved. Moreover, such failures become more severe under distribution shift, where models struggle to generalize to unseen benchmarks. In this work, we propose XSearch, an intrinsically explainable code search framework. Our key insight is that by relying on global embedding similarity, existing retrievers inherently take an inductive view. They learn statistical patterns rather than truly understanding the query's functional requirements. We address this problem by reformulating code search as a deductive concept alignment problem. XSearch (i) identifies functional concepts in the query and (ii) explicitly aligns them with corresponding code statements. This explain-then-predict design produces inherent concept-level explanations and mitigates shortcut learning that harms out-of-distribution generalization. We train an encoder with explicit concept-alignment objectives and perform retrieval through explicit matching between query concepts and code statements. Experiments show that, trained on CodeSearchNet using GraphCodeBERT (125M parameters), XSearch improves performance on out-of-distribution benchmarks from 0.02 to 0.33 (15x) over eight state-of-the-art retrievers, and consistently outperforms both encoder- and decoder-based baselines with up to 7B parameters. A user study demonstrates that concept-alignment explanations enable users to evaluate retrieved results faster and more accurately.

cs.SE↗

arXiv2025

LLM as an Execution Estimator: Recovering Missing Dependency for Practical Time-travelling Debugging

Determining the dynamic data dependency of a step that reads a variable $v$ is challenging. It typically requires either exhaustive instrumentation, which becomes prohibitively expensive when $v$ is defined within library calls, or repeated executions, which are impractical for non-deterministic programs. In this work, we propose RecovSlicing for computing dynamic data dependency in a single run, with only partial instrumentation. We explore the intuition that LLM can potentially infer program dynamics based on a partially recorded trace and relevant code as its context. Given (1) a partially recorded trace of a program $P$ and (2) the slicing criteria consisting of a query step $s$ and a query variable $v$ read by $s$, RecovSlicing computes the runtime definition of $v$ on the trace by estimating the miss-recorded execution of $P$. In this work, we allow the user to specify implicit query variable. Technically, built upon non-deterministic LLM, we address the challenges of (1) precise recovery of runtime variable value and structure from the recorded execution and (2) aligning the memory address of recovered variables and the recorded variables for definition analysis. We evaluate RecovSlicing on 8300 data dependencies across three slicing benchmarks, comparing it with Slicer4J, ND-Slicer, LLM Slicer, and re-execution Slicer. RecovSlicing achieves significantly higher accuracy (80.3%, 91.1%, 98.3%) and recall (up to 98.3%) than the best baseline (accuracy: 39.0%, 82.0%, 59.9%; recall: 53.4%, 79.1%, 87.1%). Integrated into a dual-slicing regression bug localizer, it identifies 16% more regressions.

cs.SE↗