代码大模型 / AI 编程

TLA+ is a formal specification language for verifying distributed systems and safety-critical protocols. Large language models (LLMs) frequently produce TLA+ specifications that fail the TLC model checker for semantic reasons. Across 25 LLMs, the best public baseline is 26.6% syntactic parse and 8.6% semantic model-check. We present TLA-Prover, a 20-billion-parameter model for TLA+ specification synthesis. Training combines supervised fine-tuning (SFT) on verified examples with repair-based group-relative policy optimization (GRPO). In the GRPO stage, the model learns to fix its own rejected specifications. We also train a direct preference optimization (DPO) variant from the same SFT checkpoint as an ablation. TLC provides the reward signal directly, with no learned reward model. Four tiers grade each output: Bronze (parses), Silver (no warnings), Gold (passes TLC), and Diamond. To reach Diamond, the model's correctness property is automatically altered in a small way; TLC must then detect a violation. If TLC still passes, the property was always-true and contributes nothing; the output fails Diamond. TLA-Prover reaches 9/30 (i.e. pass@1 = 30%) at both Gold and Diamond on a held-out 30-problem benchmark. This is roughly 3.5x the 8.6% untuned baseline. The DPO variant reaches 20% at Diamond. Gold and Diamond coincide at every checkpoint; this prevents the trivial-property failure mode.

With data-driven development now widely adopted, online A/B testing is an established method for measuring the effects of new technologies. However, deploying online experiments demands resources for design, implementation, and deployment, and may negatively impact users (e.g., unsafe or unethical outcomes) while requiring weeks of data collection. To address this, the growing research area of off-policy evaluation (OPE), or offline A/B testing, assesses new technologies offline using previously collected logged data. OPE is also a fundamental problem in reinforcement learning and is important where online testing is expensive or risky, such as healthcare, recommender systems, education, and robotics. Despite advances in code-generation large language models (LLMs) and agentic workflows, little is known about whether and how LLMs and LLM-based agents can automatically optimize OPE implementations. We propose GrowthHacker, a benchmark that evaluates baseline LLMs and LLM-based agents on large-scale public datasets. GrowthHacker autonomously and iteratively modifies code, runs OPE, and uses the metrics to guide subsequent optimization. We evaluate methods on Open Bandit Pipeline (OBP) and Scope-RL, and develop a two_agent framework that addresses limitations of existing frameworks while reducing complexity. Across both libraries, two_agent shows the highest reliability (98.1%-100% success rate) and positive-outcome rate (78%), with a median improvement of 4.4% among positive outcomes; CrewAI achieves the highest average improvement (37.9%) and is the only framework with zero extreme-value failures. AutoGen and Default each reach 65% positive-outcome rates. These results establish the feasibility of using LLM-based agents as automated "growth hackers" to continuously improve OPE systems, with implications for scaling data-driven decision-making where manual optimization is expensive.

Repairing system crashes discovered by kernel fuzzers like Syzkaller is a critical yet underexplored challenge in software engineering. While recent works have introduced Large Language Model (LLM) based agents for Linux kernel crash-resolution, their evaluation benchmarks are usually static and thus, do not capture the evolving nature of the Linux kernel, and suffer from potential data contamination due to LLM knowledge cutoffs. To address the above problem, we present (i) Live-kBench, an evaluation framework for self-evolving benchmarks that continuously scrapes and evaluates agents on freshly discovered kernel bugs, and (ii) kEnv, an agent-agnostic standardized crash-resolution environment for kernel compilation, execution, and feedback. This design decouples agent workflows from heavy-weight execution, enabling fair and scalable comparison across diverse agent frameworks under identical conditions. To this end, we curate an inaugural dataset of 534 Linux kernel bugs and empirically demonstrate a significant performance gap, with agents achieving up to 25% higher equivalent patch rate on bugs fixed before the LLM knowledge cutoff. Using kEnv, we benchmark three state-of-the-art agents, showing that they resolve 74% of crashes on the first attempt (plausible patches); however only ~20% of generated patches closely match developer fixes. Additionally, exposing crash resolution feedback improves crash resolution rate by 29%. Live-kBench provides the community with an evaluation infrastructure for self-evolving benchmarks that is both time and attribute sensitive; complete with a public dashboard to track agent progress on Linux kernel bugs.

With the increasing complexity of large-scale software systems, identifying all necessary modifications for a specific change is challenging. Co-changed methods, which are methods frequently modified together, are crucial for understanding software dependencies. However, existing methods often produce large results with high false positives. Focusing on pull requests instead of individual commits provides a more comprehensive view of related changes, capturing essential co-change relationships. To address these challenges, we propose a learning-to-rank approach that combines source code features and change history to predict and rank co-changed methods at the pull-request level. Experiments on 150 open-source Java projects, totaling 41.5 million lines of code and 634,216 pull requests, show that the Random Forest model outperforms other models by 2.5 to 12.8 percent in NDCG@5. It also surpasses baselines such as file proximity, code clones, FCP2Vec, and StarCoder 2 by 4.7 to 537.5 percent. Models trained on longer historical data (90 to 180 days) perform consistently, while accuracy declines after 60 days, highlighting the need for bi-monthly retraining. This approach provides an effective tool for managing co-changed methods, enabling development teams to handle dependencies and maintain software quality.

State Space Models (SSMs) have emerged as an efficient alternative to the Transformer architecture. Prior work shows that, when trained under comparable conditions, SSMs can match or surpass Transformers on code understanding tasks. However, their internal mechanisms remain a black box. We present the first systematic analysis of what SSM-based code models learn along with the direct comparison between SSM and Transformer models in this domain. Our analysis shows that SSMs capture syntactic and semantic structure more effectively than Transformers during pretraining but forgets certain relations during fine-tuning on some tasks. To investigate this behavior, we introduce SSM-Interpret, a frequency-domain framework that exposes a spectral shift toward short-range dependencies during fine-tuning. Guided by these findings, we propose architectural modifications that significantly improve the performance of SSM-based code model by upto +6 MRR on NLCodeSearch. This demonstrates that our analysis not only explains model behavior but also leads directly to better designs.

Context: Schools, training platforms, and technology firms increasingly need to assess programming proficiency at scale with transparent, reproducible methods that support personalized learning pathways. Objective: This study introduces a pedagogical framework for Scratch project assessment, aligned with the Common European Framework of Reference (CEFR), providing universal competency levels for students and teachers alongside actionable insights for curriculum design. Method: We apply Fuzzy C-Means clustering to 2008246 Scratch projects evaluated via Dr.Scratch, implementing an ordinal criterion to map clusters to CEFR levels (A1-C2), and introducing enhanced classification metrics that identify transitional learners, enable continuous progress tracking, and quantify classification certainty to balance automated feedback with instructor review. Impact: The framework enables diagnosis of systemic curriculum gaps-notably a "B2 bottleneck" where only 13.3% of learners reside due to the cognitive load of integrating Logic Synchronization, and Data Representation--while providing certainty--based triggers for human intervention.

We introduce SoliDualSPHysics, a novel open-source and GPU-accelerated software that extends DualSPHysics to enable the numerical simulation of hyperelastic, finite-strain plastic, and brittle fracture behavior in deformable solids within a unified smoothed particle hydrodynamics (SPH) software framework. The software implements a total Lagrangian formulation for solid mechanics that allows direct application of external loads and boundary conditions, enabling independent solid mechanics simulations. Brittle fracture is modeled through a phase-field approach coupled with SPH, allowing crack initiation, propagation, and branching under dynamic loading without explicit crack tracking, ad hoc crack-path criteria, or local refinement. The framework also supports user-defined mathematical expressions to prescribe time- and space-dependent quantities, complementing the solid and fracture extensions and enhancing flexibility across existing and future DualSPHysics applications. Leveraging DualSPHysics' native CPU/GPU parallel architecture, the software achieves substantial computational acceleration for large-scale simulations, and the implementation is verified and validated against benchmark numerical problems and experimental data, demonstrating accuracy, robustness, and favorable scaling performance. Comprehensive implementation details and user documentation are provided to ensure reproducibility and to support further development by the community. The framework and source code are freely available through a public GitHub repository.