Search

While there has been plenty of work on generating tests from existing code, there has been limited work on generating tests from issues. A correct test must validate the code patch that resolves the issue. In this work, we focus on the scenario where the code patch does not exist yet. This approach supports two major use-cases. First, it supports TDD (test-driven development), the discipline of "test first, write code later" that has well-documented benefits for human software engineers. Second, it also validates SWE (software engineering) agents, which generate code patches for resolving issues. This paper introduces Otter, an LLM-based solution for generating tests from issues. Otter augments LLMs with rule-based analysis to check and repair their outputs, and introduces a novel self-reflective action planning stage. Experiments show Otter outperforming state-of-the-art systems for generating tests from issues, in addition to enhancing systems that generate patches from issues. We hope that Otter helps make developers more productive at resolving issues and leads to more robust, well-tested code.

Test-driven development (TDD) is the practice of writing tests first and coding later, and the proponents of TDD expound its numerous benefits. For instance, given an issue on a source code repository, tests can clarify the desired behavior among stake-holders before anyone writes code for the agreed-upon fix. Although there has been a lot of work on automated test generation for the practice "write code first, test later", there has been little such automation for TDD. Ideally, tests for TDD should be fail-to-pass (i.e., fail before the issue is resolved and pass after) and have good adequacy with respect to covering the code changed during issue resolution. This paper introduces TDD-Bench Verified, a high-quality benchmark suite of 449 issues mined from real-world GitHub code repositories. The benchmark's evaluation harness runs only relevant tests in isolation for simple yet accurate coverage measurements, and the benchmark's dataset is filtered both by human judges and by execution in the harness. This paper also presents Auto-TDD, an LLM-based solution that takes as input an issue description and a codebase (prior to issue resolution) and returns as output a test that can be used to validate the changes made for resolving the issue. Our evaluation shows that Auto-TDD yields a better fail-to-pass rate than the strongest prior work while also yielding high coverage adequacy. Overall, we hope that this work helps make developers more productive at resolving issues while simultaneously leading to more robust fixes.

This white paper, developed through close collaboration between IBM Research and UIUC researchers within the IIDAI Institute, envisions transforming hybrid cloud systems to meet the growing complexity of AI workloads through innovative, full-stack co-design approaches, emphasizing usability, manageability, affordability, adaptability, efficiency, and scalability. By integrating cutting-edge technologies such as generative and agentic AI, cross-layer automation and optimization, unified control plane, and composable and adaptive system architecture, the proposed framework addresses critical challenges in energy efficiency, performance, and cost-effectiveness. Incorporating quantum computing as it matures will enable quantum-accelerated simulations for materials science, climate modeling, and other high-impact domains. Collaborative efforts between academia and industry are central to this vision, driving advancements in foundation models for material design and climate solutions, scalable multimodal data processing, and enhanced physics-based AI emulators for applications like weather forecasting and carbon sequestration. Research priorities include advancing AI agentic systems, LLM as an Abstraction (LLMaaA), AI model optimization and unified abstractions across heterogeneous infrastructure, end-to-end edge-cloud transformation, efficient programming model, middleware and platform, secure infrastructure, application-adaptive cloud systems, and new quantum-classical collaborative workflows. These ideas and solutions encompass both theoretical and practical research questions, requiring coordinated input and support from the research community. This joint initiative aims to establish hybrid clouds as secure, efficient, and sustainable platforms, fostering breakthroughs in AI-driven applications and scientific discovery across academia, industry, and society., Comment: 70 pages, 27 figures

Large language models (LLMs) have taken the world by storm by making many previously difficult uses of AI feasible. LLMs are controlled via highly expressive textual prompts and return textual answers. Unfortunately, this unstructured text as input and output makes LLM-based applications brittle. This motivates the rise of prompting frameworks, which mediate between LLMs and the external world. However, existing prompting frameworks either have a high learning curve or take away control over the exact prompts from the developer. To overcome this dilemma, this paper introduces the Prompt Declaration Language (PDL). PDL is a simple declarative data-oriented language that puts prompts at the forefront, based on YAML. PDL works well with many LLM platforms and LLMs. It supports writing interactive applications that call LLMs and tools, and makes it easy to implement common use-cases such as chatbots, RAG, or agents. We hope PDL will make prompt programming simpler, less brittle, and more enjoyable.

As Python is increasingly being adopted for large and complex programs, the importance of static analysis for Python (such as type inference) grows. Unfortunately, static analysis for Python remains a challenging task due to its dynamic language features and its abundant external libraries. To help fill this gap, this paper presents PoTo, an Andersen-style context-insensitive and flow-insensitive points-to analysis for Python. PoTo addresses Python-specific challenges and works for large programs via a novel hybrid evaluation, integrating traditional static points-to analysis with concrete evaluation in the Python interpreter for external library calls. Next, this paper presents PoTo+, a static type inference for Python built on the points-to analysis. We evaluate PoTo+ and compare it to two state-of-the-art Python type inference techniques: (1) the static rule-based Pytype and (2) the deep-learning based DLInfer. Our results show that PoTo+ outperforms both Pytype and DLInfer on existing Python packages.

Large language models (LLMs) have become remarkably good at improving developer productivity for high-resource programming languages. These models use two kinds of data: large amounts of unlabeled code samples for pre-training and relatively smaller amounts of labeled code samples for fine-tuning or in-context learning. Unfortunately, many programming languages are low-resource, lacking labeled samples for most tasks and often even lacking unlabeled samples. Therefore, users of low-resource languages (e.g., legacy or new languages) miss out on the benefits of LLMs. Cross-lingual transfer uses data from a source language to improve model performance on a target language. It has been well-studied for natural languages, but has received little attention for programming languages. This paper reports extensive experiments on four tasks using a transformer-based LLM and 11 to 41 programming languages to explore the following questions. First, how well does cross-lingual transfer work for a given task across different language pairs. Second, given a task and target language, how should one choose a source language. Third, which characteristics of a language pair are predictive of transfer performance, and how does that depend on the given task. Our empirical study with 1,808 experiments reveals practical and scientific insights, such as Kotlin and JavaScript being the most transferable source languages and different tasks relying on substantially different features. Overall, we find that learning transfers well across several programming languages., Comment: 15 pages, 9 figures, 8 tables

There have been many papers with algorithms for improving fairness of machine-learning classifiers for tabular data. Unfortunately, most use only very few datasets for their experimental evaluation. We introduce a suite of functions for fetching 20 fairness datasets and providing associated fairness metadata. Hopefully, these will lead to more rigorous experimental evaluations in future fairness-aware machine learning research.

Sliding-window aggregation is a foundational stream processing primitive that efficiently summarizes recent data. The state-of-the-art algorithms for sliding-window aggregation are highly efficient when stream data items are evicted or inserted one at a time, even when some of the insertions occur out-of-order. However, real-world streams are often not only out-of-order but also burtsy, causing data items to be evicted or inserted in larger bulks. This paper introduces a new algorithm for sliding-window aggregation with bulk eviction and bulk insertion. For the special case of single insert and evict, our algorithm matches the theoretical complexity of the best previous out-of-order algorithms. For the case of bulk evict, our algorithm improves upon the theoretical complexity of the best previous algorithm for that case and also outperforms it in practice. For the case of bulk insert, there are no prior algorithms, and our algorithm improves upon the naive approach of emulating bulk insert with a loop over single inserts, both in theory and in practice. Overall, this paper makes high-performance algorithms for sliding window aggregation more broadly applicable by efficiently handling the ubiquitous cases of out-of-order data and bursts., Comment: Extended version for VLDB 2023 paper

Low-code programming allows citizen developers to create programs with minimal coding effort, typically via visual (e.g. drag-and-drop) interfaces. In parallel, recent AI-powered tools such as Copilot and ChatGPT generate programs from natural language instructions. We argue that these modalities are complementary: tools like ChatGPT greatly reduce the need to memorize large APIs but still require their users to read (and modify) programs, whereas visual tools abstract away most or all programming but struggle to provide easy access to large APIs. At their intersection, we propose LowCoder, the first low-code tool for developing AI pipelines that supports both a visual programming interface (LowCoder_VP) and an AI-powered natural language interface (LowCoder_NL). We leverage this tool to provide some of the first insights into whether and how these two modalities help programmers by conducting a user study. We task 20 developers with varying levels of AI expertise with implementing four ML pipelines using LowCoder, replacing the LowCoder_NL component with a simple keyword search in half the tasks. Overall, we find that LowCoder is especially useful for (i) Discoverability: using LowCoder_NL, participants discovered new operators in 75% of the tasks, compared to just 32.5% and 27.5% using web search or scrolling through options respectively in the keyword-search condition, and (ii) Iterative Composition: 82.5% of tasks were successfully completed and many initial pipelines were further successfully improved. Qualitative analysis shows that AI helps users discover how to implement constructs when they know what to do, but still fails to support novices when they lack clarity on what they want to accomplish. Overall, our work highlights the benefits of combining the power of AI with low-code programming.

Bias mitigators can improve algorithmic fairness in machine learning models, but their effect on fairness is often not stable across data splits. A popular approach to train more stable models is ensemble learning, but unfortunately, it is unclear how to combine ensembles with mitigators to best navigate trade-offs between fairness and predictive performance. To that end, we built an open-source library enabling the modular composition of 8 mitigators, 4 ensembles, and their corresponding hyperparameters, and we empirically explored the space of configurations on 13 datasets. We distilled our insights from this exploration in the form of a guidance diagram for practitioners that we demonstrate is robust and reproducible., Comment: arXiv admin note: text overlap with arXiv:2202.00751

Traditionally, computer programming has been the prerogative of professional developers using textual programming languages such as C, Java, or Python. Low-code programming promises an alternative: letting citizen developers create programs using visual abstractions, demonstrations, or natural language. While low-code programming is currently getting a lot of attention in industry, the relevant research literature is scattered, and in fact, rarely uses the term "low-code". This article brings together low-code literature from various research fields, explaining how techniques work while providing a unified point of view. Low-code has the potential to empower more people to automate tasks by creating computer programs, making them more productive and less dependent on scarce professional software developers.

Machine learning in practice often involves complex pipelines for data cleansing, feature engineering, preprocessing, and prediction. These pipelines are composed of operators, which have to be correctly connected and whose hyperparameters must be correctly configured. Unfortunately, it is quite common for certain combinations of datasets, operators, or hyperparameters to cause failures. Diagnosing and fixing those failures is tedious and error-prone and can seriously derail a data scientist's workflow. This paper describes an approach for automatically debugging an ML pipeline, explaining the failures, and producing a remediation. We implemented our approach, which builds on a combination of AutoML and SMT, in a tool called Maro. Maro works seamlessly with the familiar data science ecosystem including Python, Jupyter notebooks, scikit-learn, and AutoML tools such as Hyperopt. We empirically evaluate our tool and find that for most cases, a single remediation automatically fixes errors, produces no additional faults, and does not significantly impact optimal accuracy nor time to convergence., Comment: Extended version of MAPS 2022 paper

A central function of code review is to increase understanding; helping reviewers understand a code change aids in knowledge transfer and finding bugs. Comments in code largely serve a similar purpose, helping future readers understand the program. It is thus natural to study what happens when these two forms of understanding collide. We ask: what documentation-related comments do reviewers make and how do they affect understanding of the contribution? We analyze ca.700K review comments on 2,000 (Java and Python) GitHub projects, and propose several filters to identify which comments are likely to be either in response to a change in documentation and/or call for such a change. We identify 65K such cases. We next develop a taxonomy of the reviewer intents behind such "comments on comments". We find that achieving a shared understanding of the code is key: reviewer comments most often focused on clarification, followed by pointing out issues to fix, such as typos and outdated comments. Curiously, clarifying comments were frequently suggested (often verbatim) by the reviewer, indicating a desire to persist their understanding acquired during code review. We conclude with a discussion of implications of our comments-on-comments dataset for research on improving code review, including the potential benefits for automating code review.

There are several bias mitigators that can reduce algorithmic bias in machine learning models but, unfortunately, the effect of mitigators on fairness is often not stable when measured across different data splits. A popular approach to train more stable models is ensemble learning. Ensembles, such as bagging, boosting, voting, or stacking, have been successful at making predictive performance more stable. One might therefore ask whether we can combine the advantages of bias mitigators and ensembles? To explore this question, we first need bias mitigators and ensembles to work together. We built an open-source library enabling the modular composition of 10 mitigators, 4 ensembles, and their corresponding hyperparameters. Based on this library, we empirically explored the space of combinations on 13 datasets, including datasets commonly used in fairness literature plus datasets newly curated by our library. Furthermore, we distilled the results into a guidance diagram for practitioners. We hope this paper will contribute towards improving stability in bias mitigation.

GraphQL is a query language for APIs and a runtime for executing those queries, fetching the requested data from existing microservices, REST APIs, databases, or other sources. Its expressiveness and its flexibility have made it an attractive candidate for API providers in many industries, especially through the web. A major drawback to blindly servicing a client's query in GraphQL is that the cost of a query can be unexpectedly large, creating computation and resource overload for the provider, and API rate-limit overages and infrastructure overload for the client. To mitigate these drawbacks, it is necessary to efficiently estimate the cost of a query before executing it. Estimating query cost is challenging, because GraphQL queries have a nested structure, GraphQL APIs follow different design conventions, and the underlying data sources are hidden. Estimates based on worst-case static query analysis have had limited success because they tend to grossly overestimate cost. We propose a machine-learning approach to efficiently and accurately estimate the query cost. We also demonstrate the power of this approach by testing it on query-response data from publicly available commercial APIs. Our framework is efficient and predicts query costs with high accuracy, consistently outperforming the static analysis by a large margin.

Sliding-window aggregation is a widely-used approach for extracting insights from the most recent portion of a data stream. The aggregations of interest can usually be expressed as binary operators that are associative but not necessarily commutative nor invertible. Non-invertible operators, however, are difficult to support efficiently. In a 2017 conference paper, we introduced DABA, the first algorithm for sliding-window aggregation with worst-case constant time. Before DABA, if a window had size $n$, the best published algorithms would require $O(\log n)$ aggregation steps per window operation---and while for strictly in-order streams, this bound could be improved to $O(1)$ aggregation steps on average, it was not known how to achieve an $O(1)$ bound for the worst-case, which is critical for latency-sensitive applications. This article is an extended version of our 2017 paper. Besides describing DABA in more detail, this article introduces a new variant, DABA Lite, which achieves the same time bounds in less memory. Whereas DABA requires space for storing $2n$ partial aggregates, DABA Lite only requires space for $n+2$ partial aggregates. Our experiments on synthetic and real data support the theoretical findings.

Automated machine learning makes it easier for data scientists to develop pipelines by searching over possible choices for hyperparameters, algorithms, and even pipeline topologies. Unfortunately, the syntax for automated machine learning tools is inconsistent with manual machine learning, with each other, and with error checks. Furthermore, few tools support advanced features such as topology search or higher-order operators. This paper introduces Lale, a library of high-level Python interfaces that simplifies and unifies automated machine learning in a consistent way., Comment: KDD Workshop on Automation in Machine Learning (AutoML@KDD), August 2020

AI automation tools need machine-readable hyperparameter schemas to define their search spaces. At the same time, AI libraries often come with good human-readable documentation. While such documentation contains most of the necessary information, it is unfortunately not ready to consume by tools. This paper describes how to automatically mine Python docstrings in AI libraries to extract JSON Schemas for their hyperparameters. We evaluate our approach on 119 transformers and estimators from three different libraries and find that it is effective at extracting machine-readable schemas. Our vision is to reduce the burden to manually create and maintain such schemas for AI automation tools and broaden the reach of automation to larger libraries and richer schemas.

JSON is a popular data format used pervasively in web APIs, cloud computing, NoSQL databases, and increasingly also machine learning. JSON Schema is a language for declaring the structure of valid JSON data. There are validators that can decide whether a JSON document is valid with respect to a schema. Unfortunately, like all instance-based testing, these validators can only show the presence and never the absence of a bug. This paper presents a complementary technique: JSON subschema checking, which can be used for static type checking with JSON Schema. Deciding whether one schema is a subschema of another is non-trivial because of the richness of the JSON Schema specification language. Given a pair of schemas, our approach first canonicalizes and simplifies both schemas, then decides the subschema question on the canonical forms, dispatching simpler subschema queries to type-specific checkers. We apply an implementation of our subschema checking algorithm to 8,548 pairs of real-world JSON schemas from different domains, demonstrating that it can decide the subschema question for most schema pairs and is always correct for schema pairs that it can decide. We hope that our work will bring more static guarantees to hard-to-debug domains, such as cloud computing and artificial intelligence.

Machine-learning automation tools, ranging from humble grid-search to hyperopt, auto-sklearn, and TPOT, help explore large search spaces of possible pipelines. Unfortunately, each of these tools has a different syntax for specifying its search space, leading to lack of portability, missed relevant points, and spurious points that are inconsistent with error checks and documentation of the searchable base components. This paper proposes using types (such as enum, float, or dictionary) both for checking the correctness of, and for automatically searching over, hyperparameters and pipeline configurations. Using types for both of these purposes guarantees consistency. We present Lale, an embedded language that resembles scikit learn but provides better automation, correctness checks, and portability. Lale extends the reach of existing automation tools across data modalities (tables, text, images, time-series) and programming languages (Python, Java, R). Thus, data scientists can leverage automation while remaining in control of their work.

Systems that can automatically analyze EEG signals can aid neurologists by reducing heavy workload and delays. However, such systems need to be first trained using a labeled dataset. While large corpuses of EEG data exist, a fraction of them are labeled. Hand-labeling data increases workload for the very neurologists we try to aid. This paper proposes a semi-supervised learning workflow that can not only extract meaningful information from large unlabeled EEG datasets but also make predictions with minimal supervision, using labeled datasets as small as 5 examples., Comment: Machine Learning for Health (ML4H) at NeurIPS 2019 - Extended Abstract

Stan is a popular probabilistic programming language with a self-contained syntax and semantics that is close to graphical models. Unfortunately, existing embeddings of Stan in Python use multi-line strings. That approach forces users to switch between two different language styles, with no support for syntax highlighting or simple error reporting within the Stan code. This paper tackles the question of whether Stan could use Python syntax while retaining its self-contained semantics. The answer is yes, that can be accomplished by reinterpreting the Python syntax. This paper introduces Yaps, a new frontend to Stan based on reinterpreted Python. We tested Yaps on over a thousand Stan models and made it available open-source.

Sliding-window aggregation summarizes the most recent information in a data stream. Users specify how that summary is computed, usually as an associative binary operator because this is the most general known form for which it is possible to avoid naively scanning every window. For strictly in-order arrivals, there are algorithms with $O(1)$ time per window change assuming associative operators. Meanwhile, it is common in practice for streams to have data arriving slightly out of order, for instance, due to clock drifts or communication delays. Unfortunately, for out-of-order streams, one has to resort to latency-prone buffering or pay $O(\log n)$ time per insert or evict, where $n$ is the window size. This paper presents the design, analysis, and implementation of FiBA, a novel sliding-window aggregation algorithm with an amortized upper bound of $O(\log d)$ time per insert or evict, where $d$ is the distance of the inserted or evicted value to the closer end of the window. This means $O(1)$ time for in-order arrivals and nearly $O(1)$ time for slightly out-of-order arrivals, with a smooth transition towards $O(\log n)$ as $d$ approaches $n$. We also prove a matching lower bound on running time, showing optimality. Our algorithm is as general as the prior state-of-the-art: it requires associativity, but not invertibility nor commutativity. At the heart of the algorithm is a careful combination of finger-searching techniques, lazy rebalancing, and position-aware partial aggregates. We further show how to answer range queries that aggregate subwindows for window sharing. Finally, our experimental evaluation shows that FiBA performs well in practice and supports the theoretical findings.

Stan is a probabilistic programming language that is popular in the statistics community, with a high-level syntax for expressing probabilistic models. Stan differs by nature from generative probabilistic programming languages like Church, Anglican, or Pyro. This paper presents a comprehensive compilation scheme to compile any Stan model to a generative language and proves its correctness. We use our compilation scheme to build two new backends for the Stanc3 compiler targeting Pyro and NumPyro. Experimental results show that the NumPyro backend yields a 2.3x speedup compared to Stan in geometric mean over 26 benchmarks. Building on Pyro we extend Stan with support for explicit variational inference guides and deep probabilistic models. That way, users familiar with Stan get access to new features without having to learn a fundamentally new language.

Deep probabilistic programming languages try to combine the advantages of deep learning with those of probabilistic programming languages. If successful, this would be a big step forward in machine learning and programming languages. Unfortunately, as of now, this new crop of languages is hard to use and understand. This paper addresses this problem directly by explaining deep probabilistic programming languages and indirectly by characterizing their current strengths and weaknesses.

We discuss the suitability of spreadsheet processors as tools for programming streaming systems. We argue that, while spreadsheets can function as powerful models for stream operators, their fundamental boundedness limits their scope of application. We propose two extensions to the spreadsheet model and argue their utility in the context of programming streaming systems., Comment: In Proceedings of the 2nd Workshop on Software Engineering Methods in Spreadsheets (http://spreadsheetlab.org/sems15/)