An official website of the United States government
Over the past few years, Large Language Models of Code (Code LLMs) have started to have a significant impact on programming practice. Code LLMs are also emerging as building blocks for research in programming languages and software engineering. However, the quality of code produced by a Code LLM varies significantly by programming language. Code LLMs produce impressive results on high-resource programming languages that are well represented in their training data (e.g., Java, Python, or JavaScript), but struggle with low-resource languages that have limited training data available (e.g., OCaml, Racket, and several others). This paper presents an effective approach for boosting the performance of Code LLMs on low-resource languages using semi-synthetic data. Our approach, called MultiPL-T, generates high-quality datasets for low-resource languages, which can then be used to fine-tune any pretrained Code LLM. MultiPL-T translates training data from high-resource languages into training data for low-resource languages in the following way. 1) We use a Code LLM to synthesize unit tests for commented code from a high-resource source language, filtering out faulty tests and code with low test coverage. 2) We use a Code LLM to translate the code from the high-resource source language to a target low-resource language. This gives us a corpus of candidate training data in the target language, but many of these translations are wrong. 3) We use a lightweight compiler to compile the test cases generated in (1) from the source language to the target language, which allows us to filter our obviously wrong translations. The result is a training corpus in the target low-resource language where all items have been validated with test cases. We apply this approach to generate tens of thousands of new, validated training items for five low-resource languages: Julia, Lua, OCaml, R, and Racket, using Python as the source high-resource language. Furthermore, we use an open Code LLM (StarCoderBase) with open training data (The Stack), which allows us to decontaminate benchmarks, train models without violating licenses, and run experiments that could not otherwise be done. Using datasets generated with MultiPL-T, we present fine-tuned versions of StarCoderBase and Code Llama for Julia, Lua, OCaml, R, and Racket that outperform other fine-tunes of these base models on the natural language to code task. We also present Racket fine-tunes for two very recent models, DeepSeek Coder and StarCoder2, to show that MultiPL-T continues to outperform other fine-tuning approaches for low-resource languages. The MultiPL-T approach is easy to apply to new languages, and is significantly more efficient and effective than alternatives such as training longer.
more »
« less
BatFix: Repairing language model-based transpilation
To keep up with changes in requirements, frameworks, and coding practices, software organizations might need to migrate code from one language to another. Source-to-source migration, or transpilation, is often a complex, manual process. Transpilation requires expertise both in the source and target language, making it highly laborious and costly. Languages models for code generation and transpilation are becoming increasingly popular. However, despite capturing code-structure well, code generated by language models is often spurious and contains subtle problems. We proposeBatFix, a novel approach that augments language models for transpilation by leveraging program repair and synthesis to fix the code generated by these models.BatFixtakes as input both the original program, the target program generated by the machine translation model, and a set of test cases and outputs a repaired program that passes all test cases. Experimental results show that our approach is agnostic to language models and programming languages.BatFixcan locate bugs spawning multiple lines and synthesize patches for syntax and semantic bugs for programs migrated fromJavatoC++andPythontoC++from multiple language models, including, OpenAI’sCodex.
more »
« less
- Award ID(s):
- 1750116
- PAR ID:
- 10568538
- Publisher / Repository:
- ACM
- Date Published:
- Journal Name:
- ACM Transactions on Software Engineering and Methodology
- Volume:
- 33
- Issue:
- 6
- ISSN:
- 1049-331X
- Page Range / eLocation ID:
- 1 to 29
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We present an enumerative program synthesis framework calledcomponent-based refactoringthat can refactor “direct” style code that does not use library components into equivalent “combinator” style code that does use library components. This framework introduces a sound but incomplete technique to check the equivalence of direct code and combinator code calledequivalence by canonicalizationthat does not rely on input-output examples or logical specifications. Moreover, our approach can repurpose existing compiler optimizations, leveraging decades of research from the programming languages community. We instantiated our new synthesis framework in two contexts: (i) higher-order functional combinators such asmapandfilterin the staticallytyped functional programming language Elm and (ii) high-performance numerical computing combinators provided by the NumPy library for Python. We implemented both instantiations in a tool calledCobblerand evaluated it on thousands of real programs to test the performance of the component-based refactoring framework in terms of execution time and output quality. Our work offers evidence that synthesis-backed refactoring can apply across a range of domains without specification beyond the input program.more » « less
-
Compiler fuzzing tools such as Csmith have uncovered many bugs in compilers by randomly sampling programs from a generative model. The success of these tools is often attributed to their ability to generate unexpected corner case inputs that developers tend to overlook during manual testing. At the same time, their chaotic nature makes fuzzer-generated test cases notoriously hard to interpret, which has lead to the creation of input simplification tools such as C-Reduce (for C compiler bugs). In until now unrelated work, researchers have also shown that human-written software tends to be rather repetitive and predictable to language models. Studies show that developers deliberately write more predictable code, whereas code with bugs is relatively unpredictable. In this study, we ask the natural questions of whether this high predictability property of code also, and perhaps counter-intuitively, applies to fuzzer-generated code. That is, we investigate whether fuzzer-generated compiler inputs are deemed unpredictable by a language model built on human-written code and surprisingly conclude that it is not. To the contrary, Csmith fuzzer-generated programs are more predictable on a per-token basis than human-written C programs. Furthermore, bug-triggering tended to be more predictable still than random inputs, and the C-Reduce minimization tool did not substantially increase this predictability. Rather, we find that bug-triggering inputs are unpredictable relative to Csmith's own generative model. This is encouraging; our results suggest promising research directions on incorporating predictability metrics in the fuzzing and reduction tools themselves.more » « less
-
Static analysis tools have demonstrated effectiveness at finding bugs in real world code. Such tools are increasingly widely adopted to improve software quality in practice. Automated Program Repair (APR) has the potential to further cut down on the cost of improving software quality. However, there is a disconnect between these effective bug-finding tools and APR. Recent advances in APR rely on test cases, making them inapplicable to newly discovered bugs or bugs difficult to test for deterministically (like memory leaks). Additionally, the quality of patches generated to satisfy a test suite is a key challenge. We address these challenges by adapting advances in practical static analysis and verification techniques to enable a new technique that finds and then accurately fixes real bugs without test cases. We present a new automated program repair technique using Separation Logic. At a high-level, our technique reasons over semantic effects of existing program fragments to fix faults related to general pointer safety properties: resource leaks, memory leaks, and null dereferences. The procedure automatically translates identified fragments into source-level patches, and verifies patch correctness with respect to reported faults. In this work we conduct the largest study of automatically fixing undiscovered bugs in real-world code to date. We demonstrate our approach by correctly fixing 55 bugs, including 11 previously undiscovered bugs, in 11 real-world projects.more » « less
-
Michael Pradel (Ed.)Large language models have demonstrated the ability to generate both natural language and programming language text. Although contemporary code generation models are trained on corpora with several programming languages, they are tested using benchmarks that are typically monolingual. The most widely used code generation benchmarks only target Python, so there is little quantitative evidence of how code generation models perform on other programming languages. We propose MultiPL-E, a system for translating unit test-driven code generation benchmarks to new languages. We create the first massively multilingual code generation benchmark by using MultiPL-E to translate two popular Python code generation benchmarks to 18 additional programming languages. We use MultiPL-E to extend the HumanEval benchmark and MBPP benchmark to 18 languages that encompass a range of programming paradigms and popularity. Using these new parallel benchmarks, we evaluate the multi-language performance of three state-of-the-art code generation models: Codex, CodeGen and InCoder. We find that Codex matches or even exceeds its performance on Python for several other languages. The range of programming languages represented in MultiPL-E allow us to explore the impact of language frequency and language features on model performance. Finally, the MultiPL-E approach of compiling code generation benchmarks to new programming languages is both scalable and extensible, making it straightforward to evaluate new models, benchmarks, and languages.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1145/3658668
