arXiv subjects

Michele Tufano

Publications and source records attributed to Michele Tufano.

At least 19 recordsLinked to original sources

arXiv2026

Dynamic Cogeneration of Bug Reproduction Test in Agentic Program Repair

Bug Reproduction Tests (BRTs) have been used in many Automated Program Repair (APR) systems, primarily for validating fixes and aiding fix generation. In practice, when developers submit a patch, they often implement the BRT alongside the fix. Our experience deploying agentic APR reveals that developers desire a BRT within AI-generated patches to increase their confidence. However, canonical APR systems tend to generate BRTs and fixes separately, and focus on producing only the fix in the final patch. In this paper, we study agentic APR in the context of cogeneration, where the APR agent is instructed to generate both a fix and a BRT in the same patch. We evaluate the effectiveness of different cogeneration strategies on 120 human-reported bugs at Google and characterize different cogeneration strategies by their influence on APR agent behavior. We develop and evaluate patch selectors that account for test change to select patches with plausible fixes (and plausible BRTs). Finally, we analyze the root causes of failed cogeneration trajectories. We show that cogeneration allows the APR agent to generate BRTs for at least as many bugs as a dedicated BRT agent, without compromising the generation rate of plausible fixes, thereby reducing engineering effort in maintaining and coordinating separate generation pipelines for fix and BRT at scale.

cs.SE↗

arXiv2026

Grounding AI Agents in Contracts: An Empirical Evaluation of Spec-Driven Test Generation

LLM-based agents are increasingly used for coding tasks, where they have outperformed many classical approaches and scaled to repository-level tasks, such as test generation. However, when directly prompted to generate tests, these agents can fail to reason about the code and its underlying contracts, thereby missing edge cases and behavioral boundaries that affect test quality. To address this limitation, we propose Spec-Driven Test Generation, where we instruct an agent to first reason about -- and explicitly document -- code pre-conditions, post-conditions, and undefined behaviors. This intermediate semi-formal specification acts as a cognitive scaffold to guide subsequent test generation. Our evaluation on production bugs from Google shows that the spec-driven agent can deliver a 9.8 percentage points ($p = 0.0352$) improvement in bug detection rate and a 2.5 percentage point ($p = 0.0034$) improvement in branch coverage, compared to a traditional test generation agent baseline. Using LLM-as-a-Judge, we further show that test suites generated by the spec-driven agent are superior to the baseline and human-authored tests in 77.8% and 56.7% of the cases, respectively, and demonstrated improvements on following best practices, readability, and edge-case coverage.

cs.SE↗

arXiv2026

Enabling Global, Human-Centered Explanations for LLMs:From Tokens to Interpretable Code and Test Generation

As Large Language Models for Code (LM4Code) become integral to software engineering, establishing trust in their output becomes critical. However, standard accuracy metrics obscure the underlying reasoning of generative models, offering little insight into how decisions are made. Although post-hoc interpretability methods attempt to fill this gap, they often restrict explanations to local, token-level insights, which fail to provide a developer-understandable global analysis. Our work highlights the urgent need for \textbf{global, code-based} explanations that reveal how models reason across code. To support this vision, we introduce \textit{code rationales} (CodeQ), a framework that enables global interpretability by mapping token-level rationales to high-level programming categories. Aggregating thousands of these token-level explanations allows us to perform statistical analyses that expose systemic reasoning behaviors. We validate this aggregation by showing it distills a clear signal from noisy token data, reducing explanation uncertainty (Shannon entropy) by over 50%. Additionally, we find that a code generation model (\textit{codeparrot-small}) consistently favors shallow syntactic cues (e.g., \textbf{indentation}) over deeper semantic logic. Furthermore, in a user study with 37 participants, we find its reasoning is significantly misaligned with that of human developers. These findings, hidden from traditional metrics, demonstrate the importance of global interpretability techniques to foster trust in LM4Code.

cs.SE↗

arXiv2026

Abstain and Validate: A Dual-LLM Policy for Reducing Noise in Agentic Program Repair

Agentic Automated Program Repair (APR) is increasingly tackling complex, repository-level bugs in industry, but ultimately these patches still need to be reviewed by a human before committing them to ensure they address the bug. Showing patches unlikely to be accepted can lead to substantial noise, wasting valuable developer time and eroding trust in automated code changes. We introduce two complementary LLM-based policies to reduce such noise: bug abstention and patch validation policies. Bug abstention excludes bugs that the agentic APR system is unlikely to fix. Patch validation rejects patches that are unlikely to be a good fix for the given bug. We evaluate both policies on three sets of bugs from Google's codebase, and their candidate patches generated by an internal agentic APR system. On a set of 174 human-reported bugs, removing bugs and patches rejected by our policies can raise success rates by up to 13 percentage points and 15 percentage points, respectively, and by up to 39 percentage points in combination. On null pointer exceptions and sanitizer-reported bugs with machine-generated bug reports, patch validation also improves average single-sample success rates. This two-policy approach provides a practical path to the reliable, industrial-scale deployment of agentic APR systems.

cs.SE↗

arXiv2025

Towards a Human-in-the-Loop Framework for Reliable Patch Evaluation Using an LLM-as-a-Judge

Reliable evaluation is crucial for advancing Automated Program Repair (APR), but prevailing benchmarks rely on execution-based evaluation methods (unit test pass@k), which fail to capture true patch validity. Determining validity can require costly manual annotation. To reduce this cost, we introduce a human-in-the-loop approach to LLM-based patch validity judgment. Inspired by the observation that human judgment is better aligned when using a shared rubric, we first employ an LLM to generate a per-bug rubric, followed by a one-time human review and optional refinement to this rubric, and then employ an LLM to judge patches using the refined rubric. We apply this approach to assign binary validity labels to patches for issues found by Google sanitizer tools. Our results show that this approach yields substantial agreement with human consensus (Cohen's kappa 0.75), high recall (0.94) and high precision (0.80), when considering patches that have unanimous agreement from 3 human raters on the validity labels. On the full dataset including patches where human raters disagree, we find this approach can still be further improved (Cohen's kappa 0.57, recall 0.93, precision 0.65) and identify possible future directions.

cs.SE↗

arXiv2025

Towards Verified Code Reasoning by LLMs

While LLM-based agents are able to tackle a wide variety of code reasoning questions, the answers are not always correct. This prevents the agent from being useful in situations where high precision is desired: (1) helping a software engineer understand a new code base, (2) helping a software engineer during code review sessions, and (3) ensuring that the code generated by an automated code generation system meets certain requirements (e.g. fixes a bug, improves readability, implements a feature). As a result of this lack of trustworthiness, the agent's answers need to be manually verified before they can be trusted. Manually confirming responses from a code reasoning agent requires human effort and can result in slower developer productivity, which weakens the assistance benefits of the agent. In this paper, we describe a method to automatically validate the answers provided by a code reasoning agent by verifying its reasoning steps. At a very high level, the method consists of extracting a formal representation of the agent's response and, subsequently, using formal verification and program analysis tools to verify the agent's reasoning steps. We applied this approach to a benchmark set of 20 uninitialized variable errors detected by sanitizers and 20 program equivalence queries. For the uninitialized variable errors, the formal verification step was able to validate the agent's reasoning on 13/20 examples, and for the program equivalence queries, the formal verification step successfully caught 6/8 incorrect judgments made by the agent.

cs.SE↗

arXiv2025

Agentic Bug Reproduction for Effective Automated Program Repair at Google

Bug reports often lack sufficient detail for developers to reproduce and fix the underlying defects. Bug Reproduction Tests (BRTs), tests that fail when the bug is present and pass when it has been resolved, are crucial for debugging, but they are rarely included in bug reports, both in open-source and in industrial settings. Thus, automatically generating BRTs from bug reports has the potential to accelerate the debugging process and lower time to repair. This paper investigates automated BRT generation within an industry setting, specifically at Google, focusing on the challenges of a large-scale, proprietary codebase and considering real-world industry bugs extracted from Google's internal issue tracker. We adapt and evaluate a state-of-the-art BRT generation technique, LIBRO, and present our agent-based approach, BRT Agent, which makes use of a fine-tuned Large Language Model (LLM) for code editing. Our BRT Agent significantly outperforms LIBRO, achieving a 28% plausible BRT generation rate, compared to 10% by LIBRO, on 80 human-reported bugs from Google's internal issue tracker. We further investigate the practical value of generated BRTs by integrating them with an Automated Program Repair (APR) system at Google. Our results show that providing BRTs to the APR system results in 30% more bugs with plausible fixes. Additionally, we introduce Ensemble Pass Rate (EPR), a metric which leverages the generated BRTs to select the most promising fixes from all fixes generated by APR system. Our evaluation on EPR for Top-K and threshold-based fix selections demonstrates promising results and trade-offs. For example, EPR correctly selects a plausible fix from a pool of 20 candidates in 70% of cases, based on its top-1 ranking.

cs.SE↗

arXiv2025

Evaluating Agent-based Program Repair at Google

Agent-based program repair offers to automatically resolve complex bugs end-to-end by combining the planning, tool use, and code generation abilities of modern LLMs. Recent work has explored the use of agent-based repair approaches on the popular open-source SWE-Bench, a collection of bugs from highly-rated GitHub Python projects. In addition, various agentic approaches such as SWE-Agent have been proposed to solve bugs in this benchmark. This paper explores the viability of using an agentic approach to address bugs in an enterprise context. To investigate this, we curate an evaluation set of 178 bugs drawn from Google's issue tracking system. This dataset spans both human-reported (78) and machine-reported bugs (100). To establish a repair performance baseline on this benchmark, we implement Passerine, an agent similar in spirit to SWE-Agent that can work within Google's development environment. We show that with 20 trajectory samples and Gemini 1.5 Pro, Passerine can produce a patch that passes bug tests (i.e., plausible) for 73% of machine-reported and 25.6% of human-reported bugs in our evaluation set. After manual examination, we found that 43% of machine-reported bugs and 17.9% of human-reported bugs have at least one patch that is semantically equivalent to the ground-truth patch. These results establish a baseline on an industrially relevant benchmark, which as we show, contains bugs drawn from a different distribution -- in terms of language diversity, size, and spread of changes, etc. -- compared to those in the popular SWE-Bench dataset.

cs.SE↗

arXiv2025

Reinforcement Learning from Automatic Feedback for High-Quality Unit Test Generation

Software testing is a crucial aspect of software development, and the creation of high-quality tests that adhere to best practices is essential for effective maintenance. Recently, Large Language Models (LLMs) have gained popularity for code generation, including the automated creation of test cases. However, these LLMs are often trained on vast amounts of publicly available code, which may include test cases that do not adhere to best practices and may even contain test smells (anti-patterns). To address this issue, we propose a novel technique called Reinforcement Learning from Static Quality Metrics (RLSQM). To begin, we analyze the anti-patterns generated by the LLM and show that LLMs can generate undesirable test smells. Thus, we train specific reward models for each static quality metric, then utilize Proximal Policy Optimization (PPO) to train models for optimizing a single quality metric at a time. Furthermore, we amalgamate these rewards into a unified reward model aimed at capturing different best practices and quality aspects of tests. By comparing RL-trained models with those trained using supervised learning, we provide insights into how reliably utilize RL to improve test generation quality and into the effects of various training strategies. Our experimental results demonstrate that the RL-optimized model consistently generated high-quality test cases compared to the base LLM, improving the model by up to 21%, and successfully generates nearly 100% syntactically correct code. RLSQM also outperformed GPT-4 on four out of seven metrics. This represents a significant step towards enhancing the overall efficiency and reliability of software testing through Reinforcement Learning and static quality metrics. Our data are available at https://figshare.com/s/ded476c8d4c221222849.

cs.SE↗

arXiv2025

Reinforcement Learning from Automatic Feedback for High-Quality Unit Test Generation

Software testing is a crucial but time-consuming aspect of software development, and recently, Large Language Models (LLMs) have gained popularity for automated test case generation. However, because LLMs are trained on vast amounts of open-source code, they often generate test cases that do not adhere to best practices and may even contain test smells (anti-patterns). To address this issue, we propose Reinforcement Learning from Static Quality Metrics (RLSQM), wherein we utilize Reinforcement Learning to generate high-quality unit tests based on static analysis-based quality metrics. First, we analyzed LLM-generated tests and show that LLMs frequently do generate undesirable test smells -- up to 37% of the time. Then, we implemented lightweight static analysis-based reward model and trained LLMs using this reward model to optimize for five code quality metrics. Our experimental results demonstrate that the RL-optimized Codex model consistently generated higher-quality test cases than the base LLM, improving quality metrics by up to 23%, and generated nearly 100% syntactically-correct code. RLSQM also outperformed GPT-4 on all code quality metrics, in spite of training a substantially cheaper Codex model. We provide insights into how reliably utilize RL to improve test generation quality and show that RLSQM is a significant step towards enhancing the overall efficiency and reliability of automated software testing. Our data are available at https://doi.org/10.6084/m9.figshare.25983166.

cs.SE↗

arXiv2024

AutoDev: Automated AI-Driven Development

The landscape of software development has witnessed a paradigm shift with the advent of AI-powered assistants, exemplified by GitHub Copilot. However, existing solutions are not leveraging all the potential capabilities available in an IDE such as building, testing, executing code, git operations, etc. Therefore, they are constrained by their limited capabilities, primarily focusing on suggesting code snippets and file manipulation within a chat-based interface. To fill this gap, we present AutoDev, a fully automated AI-driven software development framework, designed for autonomous planning and execution of intricate software engineering tasks. AutoDev enables users to define complex software engineering objectives, which are assigned to AutoDev's autonomous AI Agents to achieve. These AI agents can perform diverse operations on a codebase, including file editing, retrieval, build processes, execution, testing, and git operations. They also have access to files, compiler output, build and testing logs, static analysis tools, and more. This enables the AI Agents to execute tasks in a fully automated manner with a comprehensive understanding of the contextual information required. Furthermore, AutoDev establishes a secure development environment by confining all operations within Docker containers. This framework incorporates guardrails to ensure user privacy and file security, allowing users to define specific permitted or restricted commands and operations within AutoDev. In our evaluation, we tested AutoDev on the HumanEval dataset, obtaining promising results with 91.5% and 87.8% of Pass@1 for code generation and test generation respectively, demonstrating its effectiveness in automating software engineering tasks while maintaining a secure and user-controlled development environment.

cs.SE↗

arXiv2024

Copilot Evaluation Harness: Evaluating LLM-Guided Software Programming

The integration of Large Language Models (LLMs) into Development Environments (IDEs) has become a focal point in modern software development. LLMs such as OpenAI GPT-3.5/4 and Code Llama offer the potential to significantly augment developer productivity by serving as intelligent, chat-driven programming assistants. However, utilizing LLMs out of the box is unlikely to be optimal for any given scenario. Rather, each system requires the LLM to be honed to its set of heuristics to ensure the best performance. In this paper, we introduce the Copilot evaluation harness: a set of data and tools for evaluating LLM-guided IDE interactions, covering various programming scenarios and languages. We propose our metrics as a more robust and information-dense evaluation than previous state of the art evaluation systems. We design and compute both static and execution based success metrics for scenarios encompassing a wide range of developer tasks, including code generation from natural language (generate), documentation generation from code (doc), test case generation (test), bug-fixing (fix), and workspace understanding and query resolution (workspace). These success metrics are designed to evaluate the performance of LLMs within a given IDE and its respective parameter space. Our learnings from evaluating three common LLMs using these metrics can inform the development and validation of future scenarios in LLM guided IDEs.

cs.SE↗

arXiv2023

Predicting Code Coverage without Execution

Code coverage is a widely used metric for quantifying the extent to which program elements, such as statements or branches, are executed during testing. Calculating code coverage is resource-intensive, requiring code building and execution with additional overhead for the instrumentation. Furthermore, computing coverage of any snippet of code requires the whole program context. Using Machine Learning to amortize this expensive process could lower the cost of code coverage by requiring only the source code context, and the task of code coverage prediction can be a novel benchmark for judging the ability of models to understand code. We propose a novel benchmark task called Code Coverage Prediction for Large Language Models (LLMs). We formalize this task to evaluate the capability of LLMs in understanding code execution by determining which lines of a method are executed by a given test case and inputs. We curate and release a dataset we call COVERAGEEVAL by executing tests and code from the HumanEval dataset and collecting code coverage information. We report the performance of four state-of-the-art LLMs used for code-related tasks, including OpenAI's GPT-4 and GPT-3.5-Turbo, Google's BARD, and Anthropic's Claude, on the Code Coverage Prediction task. Finally, we argue that code coverage as a metric and pre-training data source are valuable for overall LLM performance on software engineering tasks.

cs.SE↗

arXiv2023

InferFix: End-to-End Program Repair with LLMs

Software development life cycle is profoundly influenced by bugs: their introduction, identification, and eventual resolution account for a significant portion of software cost. This has motivated software engineering researchers and practitioners to propose different approaches for automating the identification and repair of software defects. Large language models have been adapted to the program repair task through few-shot demonstration learning and instruction prompting, treating this as an infilling task. However, these models have only focused on learning general bug-fixing patterns for uncategorized bugs mined from public repositories. In this paper, we propose InferFix: a transformer-based program repair framework paired with a state-of-the-art static analyzer to fix critical security and performance bugs. InferFix combines a Retriever -- transformer encoder model pretrained via contrastive learning objective, which aims at searching for semantically equivalent bugs and corresponding fixes; and a Generator -- a large language model (Codex Cushman) finetuned on supervised bug-fix data with prompts augmented via bug type annotations and semantically similar fixes retrieved from an external non-parametric memory. To train and evaluate our approach, we curated InferredBugs, a novel, metadata-rich dataset of bugs extracted by executing the Infer static analyzer on the change histories of thousands of Java and C# repositories. Our evaluation demonstrates that InferFix outperforms strong LLM baselines, with a top-1 accuracy of 65.6% for generating fixes in C# and 76.8% in Java. We discuss the deployment of InferFix alongside Infer at Microsoft which offers an end-to-end solution for detection, classification, and localization of bugs, as well as fixing and validation of candidate patches, integrated in the continuous integration pipeline to automate the software development workflow.

cs.SE↗

arXiv2023

An Empirical Investigation into the Use of Image Captioning for Automated Software Documentation

Existing automated techniques for software documentation typically attempt to reason between two main sources of information: code and natural language. However, this reasoning process is often complicated by the lexical gap between more abstract natural language and more structured programming languages. One potential bridge for this gap is the Graphical User Interface (GUI), as GUIs inherently encode salient information about underlying program functionality into rich, pixel-based data representations. This paper offers one of the first comprehensive empirical investigations into the connection between GUIs and functional, natural language descriptions of software. First, we collect, analyze, and open source a large dataset of functional GUI descriptions consisting of 45,998 descriptions for 10,204 screenshots from popular Android applications. The descriptions were obtained from human labelers and underwent several quality control mechanisms. To gain insight into the representational potential of GUIs, we investigate the ability of four Neural Image Captioning models to predict natural language descriptions of varying granularity when provided a screenshot as input. We evaluate these models quantitatively, using common machine translation metrics, and qualitatively through a large-scale user study. Finally, we offer learned lessons and a discussion of the potential shown by multimodal models to enhance future techniques for automated software documentation.

cs.SE↗

arXiv2022

Exploring and Evaluating Personalized Models for Code Generation

Large Transformer models achieved the state-of-the-art status for Natural Language Understanding tasks and are increasingly becoming the baseline model architecture for modeling source code. Transformers are usually pre-trained on large unsupervised corpora, learning token representations and transformations relevant to modeling generally available text, and are then fine-tuned on a particular downstream task of interest. While fine-tuning is a tried-and-true method for adapting a model to a new domain -- for example, question-answering on a given topic -- generalization remains an on-going challenge. In this paper, we explore and evaluate transformer model fine-tuning for personalization. In the context of generating unit tests for Java methods, we evaluate learning to personalize to a specific software project using several personalization techniques. We consider three key approaches: (i) custom fine-tuning, which allows all the model parameters to be tuned; (ii) lightweight fine-tuning, which freezes most of the model's parameters, allowing tuning of the token embeddings and softmax layer only or the final layer alone; (iii) prefix tuning, which keeps model parameters frozen, but optimizes a small project-specific prefix vector. Each of these techniques offers a trade-off in total compute cost and predictive performance, which we evaluate by code and task-specific metrics, training time, and total computational operations. We compare these fine-tuning strategies for code generation and discuss the potential generalization and cost benefits of each in various deployment scenarios.

cs.SE↗

arXiv2022

Methods2Test: A dataset of focal methods mapped to test cases

Unit testing is an essential part of the software development process, which helps to identify issues with source code in early stages of development and prevent regressions. Machine learning has emerged as viable approach to help software developers generate automated unit tests. However, generating reliable unit test cases that are semantically correct and capable of catching software bugs or unintended behavior via machine learning requires large, metadata-rich, datasets. In this paper we present Methods2Test: A dataset of focal methods mapped to test cases: a large, supervised dataset of test cases mapped to corresponding methods under test (i.e., focal methods). This dataset contains 780,944 pairs of JUnit tests and focal methods, extracted from a total of 91,385 Java open source projects hosted on GitHub with licenses permitting re-distribution. The main challenge behind the creation of the Methods2Test was to establish a reliable mapping between a test case and the relevant focal method. To this aim, we designed a set of heuristics, based on developers' best practices in software testing, which identify the likely focal method for a given test case. To facilitate further analysis, we store a rich set of metadata for each method-test pair in JSON-formatted files. Additionally, we extract textual corpus from the dataset at different context levels, which we provide both in raw and tokenized forms, in order to enable researchers to train and evaluate machine learning models for Automated Test Generation. Methods2Test is publicly available at: https://github.com/microsoft/methods2test

cs.SE↗

arXiv2021

Long-Range Modeling of Source Code Files with eWASH: Extended Window Access by Syntax Hierarchy

Statistical language modeling and translation with transformers have found many successful applications in program understanding and generation tasks, setting high benchmarks for tools in modern software development environments. The finite context window of these neural models means, however, that they will be unable to leverage the entire relevant context of large files and packages for any given task. While there are many efforts to extend the context window, we introduce an architecture-independent approach for leveraging the syntactic hierarchies of source code for incorporating entire file-level context into a fixed-length window. Using concrete syntax trees of each source file we extract syntactic hierarchies and integrate them into context window by selectively removing from view more specific, less relevant scopes for a given task. We evaluate this approach on code generation tasks and joint translation of natural language and source code in Python programming language, achieving a new state-of-the-art in code completion and summarization for Python in the CodeXGLUE benchmark. We also introduce new CodeXGLUE benchmarks for user-experience-motivated tasks: code completion with normalized literals, method body completion/code summarization conditioned on file-level context.

cs.LG↗