A novel paradigm of mobile edge generation (MEG)-enabled digital twin (DT) is proposed, which enables distributed on-device generation at mobile edge networks for real-time DT applications. First, an MEG-DT architecture is put forward to decentralize generative artificial intelligence (GAI) models onto edge servers (ESs) and user equipments (UEs), which has the advantages of low latency, privacy preservation, and individual-level customization. Then, various single-user and multi-user generation mechanisms are conceived for MEG-DT, which strike trade-offs between generation latency, hardware costs, and device coordination. Furthermore, to perform efficient distributed generation, two operating protocols are explored for transmitting interpretable and latent features between ESs and UEs, namely sketch-based generation and seed-based generation, respectively. Based on the proposed protocols, the convergence between MEG and DT are highlighted. Considering the seed-based image generation scenario, numerical case studies are provided to reveal the superiority of MEG-DT over centralized generation. Finally, promising applications and research opportunities are identified. Code is available at https://github.com/xiaoxiaxusummer/MEG_DT
Large reasoning models achieve strong performance on challenging tasks, but their long chain-of-thought often incurs substantial inference costs by continuing to deliberate after sufficient information has been obtained. Existing reasoning compression methods primarily operate on reasoning trajectories, paying less attention to the local decision of whether further reasoning is necessary given the current state. We propose LoCo, a local preference learning method for reasoning compression. LoCo detects candidate overthinking states from the model's own traces, freezes the corresponding reasoning prefixes, regenerates shorter verified continuations from the same states, and trains on the resulting state-conditioned preference pairs with DPO. This formulation teaches the model when to continue reasoning and when to conclude, rather than simply encouraging shorter trajectories overall. We construct training data entirely from self-generated mathematical traces and evaluate LoCo on five mathematical and three code generation benchmarks. LoCo reduces completion length by 41.7\%--45.4\% on mathematics and 23.4\%--36.3\% on code, while maintaining or improving accuracy and achieving the best AUC across all code benchmarks. LoCo consistently improves the accuracy-efficiency trade-off on mathematics and transfers effectively to code generation despite math-only training, demonstrating the cross-domain generality of local preference learning for reasoning compression. The code and model will be publicly released at https://github.com/MuyaoYuan/LoCo.
Among the thriving ecosystem of cloud computing and the proliferation of Large Language Model (LLM)-based code generation tools, there is a lack of benchmarking for code generation in cloud-native applications. In response to this need, we present CloudEval-YAML, a practical benchmark for cloud configuration generation. CloudEval-YAML tackles the diversity challenge by focusing on YAML, the de facto standard of numerous cloud-native tools. We develop the CloudEval-YAML benchmark with practicality in mind: the dataset consists of hand-written problems with unit tests targeting practical scenarios. We further enhanced the dataset to meet practical needs by rephrasing questions in a concise, abbreviated, and bilingual manner. The dataset consists of 1011 problems that take more than 1200 human hours to complete. To improve practicality during evaluation, we build a scalable evaluation platform for CloudEval-YAML that achieves a 20 times speedup over a single machine. To the best of our knowledge, the CloudEval-YAML dataset is the first hand-written dataset targeting cloud-native applications. We present an in-depth evaluation of 12 LLMs, leading to a deeper understanding of the problems and LLMs, as well as effective methods to improve task performance and reduce cost.
The computing education community has a rich history of pedagogical innovation designed to support students in introductory courses, and to support teachers in facilitating student learning. Very recent advances in artificial intelligence have resulted in code generation models that can produce source code from natural language problem descriptions -- with impressive accuracy in many cases. The wide availability of these models and their ease of use has raised concerns about potential impacts on many aspects of society, including the future of computing education. In this paper, we discuss the challenges and opportunities such models present to computing educators, with a focus on introductory programming classrooms. We summarize the results of two recent articles, the first evaluating the performance of code generation models on typical introductory-level programming problems, and the second exploring the quality and novelty of learning resources generated by these models. We consider likely impacts of such models upon pedagogical practice in the context of the most recent advances at the time of writing.
The advent of large language models (LLMs) like GitHub Copilot has significantly enhanced programmers' productivity, particularly in code generation. However, these models often struggle with real-world tasks without fine-tuning. As LLMs grow larger and more performant, fine-tuning for specialized tasks becomes increasingly expensive. Parameter-efficient fine-tuning (PEFT) methods, which fine-tune only a subset of model parameters, offer a promising solution by reducing the computational costs of tuning LLMs while maintaining their performance. Existing studies have explored using PEFT and LLMs for various code-related tasks and found that the effectiveness of PEFT techniques is task-dependent. The application of PEFT techniques in unit test generation remains underexplored. The state-of-the-art is limited to using LLMs with full fine-tuning to generate unit tests. This paper investigates both full fine-tuning and various PEFT methods, including LoRA, (IA)^3, and prompt tuning, across different model architectures and sizes. We use well-established benchmark datasets to evaluate their effectiveness in unit test generation. Our findings show that PEFT methods can deliver performance comparable to full fine-tuning for unit test generation, making specialized fine-tuning more accessible and cost-effective. Notably, prompt tuning is the most effective in terms of cost and resource utilization, while LoRA approaches the effectiveness of full fine-tuning in several cases.
Existing works have established multiple benchmarks to highlight the security risks associated with Code GenAI. These risks are primarily reflected in two areas: a model potential to generate insecure code (insecure coding) and its utility in cyberattacks (cyberattack helpfulness). While these benchmarks have made significant strides, there remain opportunities for further improvement. For instance, many current benchmarks tend to focus more on a model ability to provide attack suggestions rather than its capacity to generate executable attacks. Additionally, most benchmarks rely heavily on static evaluation metrics, which may not be as precise as dynamic metrics such as passing test cases. Conversely, expert-verified benchmarks, while offering high-quality data, often operate at a smaller scale. To address these gaps, we develop SecCodePLT, a unified and comprehensive evaluation platform for code GenAIs' risks. For insecure code, we introduce a new methodology for data creation that combines experts with automatic generation. Our methodology ensures the data quality while enabling large-scale generation. We also associate samples with test cases to conduct code-related dynamic evaluation. For cyberattack helpfulness, we set up a real environment and construct samples to prompt a model to generate actual attacks, along with dynamic metrics in our environment. We conduct extensive experiments and show that SecCodePLT outperforms the state-of-the-art (SOTA) benchmark CyberSecEval in security relevance. Furthermore, it better identifies the security risks of SOTA models in insecure coding and cyberattack helpfulness. Finally, we apply SecCodePLT to the SOTA code agent, Cursor, and, for the first time, identify non-trivial security risks in this advanced coding agent.
Program comprehension is a dominant process in software development and maintenance. Experts are considered to comprehend the source code efficiently by directing their gaze, or attention, to important components in it. However, reflecting the importance of components is still a remaining issue in gaze behavior analysis for source code comprehension. Here we show a conceptual framework to compare the quantified importance of source code components with the gaze behavior of programmers. We use "attention" in attention models (e.g., code2vec) as the importance indices for source code components and evaluate programmers' gaze locations based on the quantified importance. In this report, we introduce the idea of our gaze behavior analysis using the attention map, and the results of a preliminary experiment.
Machine translation for low resource language pairs is a challenging task. This task could become extremely difficult once a speaker uses code switching. We propose a method to build a machine translation model for code-switched Kazakh-Russian language pair with no labeled data. Our method is basing on generation of synthetic data. Additionally, we present the first codeswitching Kazakh-Russian parallel corpus and the evaluation results, which include a model achieving 16.48 BLEU almost reaching an existing commercial system and beating it by human evaluation.
We introduce AutoCoder, the first Large Language Model to surpass GPT-4 Turbo (April 2024) and GPT-4o in pass@1 on the Human Eval benchmark test ($\mathbf{90.9\%}$ vs. $\mathbf{90.2\%}$). In addition, AutoCoder offers a more versatile code interpreter compared to GPT-4 Turbo and GPT-4o. It's code interpreter can install external packages instead of limiting to built-in packages. AutoCoder's training data is a multi-turn dialogue dataset created by a system combining agent interaction and external code execution verification, a method we term \textbf{\textsc{AIEV-Instruct}} (Instruction Tuning with Agent-Interaction and Execution-Verified). Compared to previous large-scale code dataset generation methods, \textsc{AIEV-Instruct} reduces dependence on proprietary large models and provides execution-validated code dataset. The code and the demo video is available in \url{https://github.com/bin123apple/AutoCoder}.
The reliance of popular programming languages such as Python and JavaScript on centralized package repositories and open-source software, combined with the emergence of code-generating Large Language Models (LLMs), has created a new type of threat to the software supply chain: package hallucinations. These hallucinations, which arise from fact-conflicting errors when generating code using LLMs, represent a novel form of package confusion attack that poses a critical threat to the integrity of the software supply chain. This paper conducts a rigorous and comprehensive evaluation of package hallucinations across different programming languages, settings, and parameters, exploring how a diverse set of models and configurations affect the likelihood of generating erroneous package recommendations and identifying the root causes of this phenomenon. Using 16 popular LLMs for code generation and two unique prompt datasets, we generate 576,000 code samples in two programming languages that we analyze for package hallucinations. Our findings reveal that that the average percentage of hallucinated packages is at least 5.2% for commercial models and 21.7% for open-source models, including a staggering 205,474 unique examples of hallucinated package names, further underscoring the severity and pervasiveness of this threat. To overcome this problem, we implement several hallucination mitigation strategies and show that they are able to significantly reduce the number of package hallucinations while maintaining code quality. Our experiments and findings highlight package hallucinations as a persistent and systemic phenomenon while using state-of-the-art LLMs for code generation, and a significant challenge which deserves the research community's urgent attention.
As language models support larger and larger context sizes, evaluating their ability to make effective use of that context becomes increasingly important. We analyze the ability of several code generation models to handle long range dependencies using a suite of multi-step key retrieval tasks in context windows up to 8k tokens in length. The tasks progressively increase in difficulty and allow more nuanced evaluation of model capabilities than tests like the popular needle-in-the-haystack test. We find that performance degrades significantly (up to 2x) when a function references another function that is defined later in the prompt. We also observe that models that use sliding window attention mechanisms have difficulty handling references further than the size of a single window. We perform simple prompt modifications using call graph information to improve multi-step retrieval performance up to 3x. Our analysis highlights different facets of long-context performance and is suggestive of prompt construction strategies for code completion tools
Generating accurate and executable code using large language models (LLMs) is challenging for languages with limited public training data compared to popular languages such as Python. This paper introduces a generalizable approach that uses small-scale code versions of the Qwen 2.5 model combined with Group Relative Policy Optimization (GRPO) to enable effective code generation through explicit reasoning steps, which is particularly beneficial for languages with smaller source code databases. Using Prolog as a representative use case -- given its limited online presence -- the initial model faced challenges in generating executable code. After some training steps, the model successfully produces logically consistent and syntactically accurate code by directly integrating reasoning-driven feedback into the reinforcement learning loop. Experimental evaluations using mathematical logic problem benchmarks illustrate significant improvements in reasoning quality, code accuracy, and logical correctness, underscoring the potential of this approach to benefit a wide range of programming languages lacking extensive training resources.
This study aims to assess the performance of two advanced Large Language Models (LLMs), GPT-3.5 and GPT-4, in the task of code clone detection. The evaluation involves testing the models on a variety of code pairs of different clone types and levels of similarity, sourced from two datasets: BigCloneBench (human-made) and GPTCloneBench (LLM-generated). Findings from the study indicate that GPT-4 consistently surpasses GPT-3.5 across all clone types. A correlation was observed between the GPTs' accuracy at identifying code clones and code similarity, with both GPT models exhibiting low effectiveness in detecting the most complex Type-4 code clones. Additionally, GPT models demonstrate a higher performance identifying code clones in LLM-generated code compared to humans-generated code. However, they do not reach impressive accuracy. These results emphasize the imperative for ongoing enhancements in LLM capabilities, particularly in the recognition of code clones and in mitigating their predisposition towards self-generated code clones--which is likely to become an issue as software engineers are more numerous to leverage LLM-enabled code generation and code refactoring tools.
We propose CURE, a novel reinforcement learning framework with a dedicated reward design that co-evolves coding and unit test generation capabilities based on their interaction outcomes, without any ground-truth code as supervision. This approach enables flexible and scalable training and allows the unit tester to learn directly from the coder's mistakes. Our derived ReasonFlux-Coder-7B and 14B models improve code generation accuracy by 5.3% and Best-of-N accuracy by 9.0% after optimization on Qwen2.5-Instruct models, outperforming similarly sized Qwen-Coder, DeepSeek-Coder, and Seed-Coder. They naturally extend to downstream tasks such as test-time scaling and agentic coding-achieving a 8.1% improvement over the base model. For the long-CoT model, our ReasonFlux-Coder-4B consistently outperforms Qwen3-4B while achieving 64.8% inference efficiency in unit test generation. Notably, we also find that our model can serve as an effective reward model for reinforcement learning on base models. Project: https://github.com/Gen-Verse/CURE
With the advent of large language models (LLMs), software engineering agents (SWE agents) have emerged as a powerful paradigm for automating a range of software tasks -- from code generation and repair to test case synthesis. These agents operate autonomously by interpreting user input and responding to environmental feedback. While various agent architectures have demonstrated strong empirical performance, the internal decision-making worfklows that drive their behavior remain poorly understood. Deeper insight into these workflows hold promise for improving both agent reliability and efficiency. In this work, we present the first systematic study of SWE agent behavior through the lens of execution traces. Our contributions are as follows: (1) we propose the first taxonomy of decision-making pathways across five representative agents; (2) using this taxonomy, we identify three core components essential to agent success -- bug localization, patch generation, and reproduction test generation -- and study each in depth; (3) we study the impact of test generation on successful patch production; and analyze strategies that can lead to successful test generation; (4) we further conduct the first large-scale code clone analysis comparing agent-generated and developer-written patches and provide a qualitative study revealing structural and stylistic differences in patch content. Together, these findings offer novel insights into agent design and open avenues for building agents that are both more effective and more aligned with human development practices.
Code Large Language Models (Code LLMs), such as Code llama and DeepSeek-Coder, have demonstrated exceptional performance in the code generation tasks. However, most existing models focus on the abilities of generating correct code, but often struggle with bug repair. We introduce a suit of methods to enhance LLM's SQL bug-fixing abilities. The methods are mainly consisted of two parts: A Progressive Dataset Construction (PDC) from scratch and Dynamic Mask Supervised Fine-tuning (DM-SFT). PDC proposes two data expansion methods from the perspectives of breadth first and depth first respectively. DM-SFT introduces an efficient bug-fixing supervised learning approach, which effectively reduce the total training steps and mitigate the "disorientation" in SQL code bug-fixing training. In our evaluation, the code LLM models trained with two methods have exceeds all current best performing model which size is much larger.
Expert programmers' eye-movements during source code reading are valuable sources that are considered to be associated with their domain expertise. We advocate a vision of new intelligent systems incorporating expertise of experts for software development tasks, such as issue localization, comment generation, and code generation. We present a conceptual framework of neural autonomous agents based on imitation learning (IL), which enables agents to mimic the visual attention of an expert via his/her eye movement. In this framework, an autonomous agent is constructed as a context-based attention model that consists of encoder/decoder network and trained with state-action sequences generated by an experts' demonstration. Challenges to implement an IL-based autonomous agent specialized for software development task are discussed in this paper.
As deep learning models scale, their training cost has surged significantly. Due to both hardware advancements and limitations in current software stacks, the need for data efficiency has risen. Data efficiency refers to the effective hiding of data access latency and the avoidance of unnecessary data movements. Major challenges arise from the growing disparity between GPU memory bandwidth and computational throughput, imminent GPU memory capacity limitations, and inefficiencies in the PyTorch software stack, including a lack of device-specific PCIe transfer optimizations and high-level domain-specific abstractions. To effectively mitigate these data inefficiencies for deep learning training, this dissertation analyzes data inefficiency in representative deep training tasks, specifically in graph neural networks (GNNs) and large language models (LLMs). It then proposes novel runtime and code generation techniques to mitigate these challenges and implements these optimizations seamlessly within the PyTorch stack while maintaining strong programmability and interoperability. First, PyTorch-Direct is devised to incorporate the GPU-centric PCIe data transfer paradigm in PyTorch for GNN training. Next, Hector intermediate representation (IR) and its code generator are proposed to introduce domain-specific high-level abstraction and systematically address memory-intensive performance challenges for relational GNNs. Finally, in LLM training, the throughput has been increasingly constrained by GPU memory capacity. To mitigate this, the SSDTrain offloading framework is designed and implemented. Together, these contributions show that code generation and runtime techniques can systematically mitigate the data management bottlenecks in deep learning training, which stem from the data-intensive nature of workloads and the oversimplification inherent in the deep learning training software stack.
Although large language models have shown strong capabilities in domain-specific code generation, their performance in robotics software development, especially for the Robot Operating System (ROS), remains largely unexplored. ROS code generation presents distinctive challenges, including complex inter-node coordination, middleware configuration, and lifecycle semantics that go far beyond standard coding benchmarks, and many important correctness properties become observable only once migrated nodes are launched and communicate. We present ROSTrans, a benchmark for evaluating LLMs on ROS1-to-ROS2 code migration, a representative and practically important setting for ROS code generation. ROSTrans comprises 108 scaffolded completion tasks drawn from a diverse set of 70+ open-source ROS1 repositories and organized into a hierarchical taxonomy assessing interface-level, behavioral-level and system-level correctness. For every task we establish an executable reference: a verified ROS2 implementation, a runtime test suite that launches the nodes and exercises their topics, services, actions and parameters, and a container ensuring reproducibility. Candidates are scored by injecting only their target files into an otherwise fixed copy of the reference package, so a candidate can neither supply nor weaken the tests it is judged by. Evaluating six frontier LLMs over five stochastic generations per task, we find pass@1 below 10.9%, and that 56.5% of all samples build successfully and then fail at runtime,and 32.0% fail runtime assertions on the migrated interfaces, a class of error no compile-only oracle can detect. Execution-guided repair further improves task success, reaching 49.1% and 47.2% for Claude Opus 4.6 and GPT-5.5 after three rounds, respectively.
Collecting large amounts of real-world interaction data to train general robotic policies is often prohibitively expensive, thus motivating the use of simulation data. However, existing methods for data generation have generally focused on scene-level diversity (e.g., object instances and poses) rather than task-level diversity, due to the human effort required to come up with and verify novel tasks. This has made it challenging for policies trained on simulation data to demonstrate significant task-level generalization. In this paper, we propose to automatically generate rich simulation environments and expert demonstrations by exploiting a large language models' (LLM) grounding and coding ability. Our approach, dubbed GenSim, has two modes: goal-directed generation, wherein a target task is given to the LLM and the LLM proposes a task curriculum to solve the target task, and exploratory generation, wherein the LLM bootstraps from previous tasks and iteratively proposes novel tasks that would be helpful in solving more complex tasks. We use GPT4 to expand the existing benchmark by ten times to over 100 tasks, on which we conduct supervised finetuning and evaluate several LLMs including finetuned GPTs and Code Llama on code generation for robotic simulation tasks. Furthermore, we observe that LLMs-generated simulation programs can enhance task-level generalization significantly when used for multitask policy training. We further find that with minimal sim-to-real adaptation, the multitask policies pretrained on GPT4-generated simulation tasks exhibit stronger transfer to unseen long-horizon tasks in the real world and outperform baselines by 25%. See the project website (https://liruiw.github.io/gensim) for code, demos, and videos.