Measuring The Impact Of Programming Language Distribution

TL;DR

This paper introduces BabelCode, a framework for execution-based evaluation of neural code models across many languages, and demonstrates that balancing language distributions in training data improves performance on low-resource languages.

Contribution

It presents BabelCode for language-agnostic evaluation and a new translation dataset, TP3, to study the impact of language distribution balancing on model performance.

Findings

Balanced training data improves low-resource language performance by 12.34% on average.

Balancing increases low-resource language pass@k by 30.77%.

Slight decrease in high-resource language performance by 12.94%..

Abstract

Current benchmarks for evaluating neural code models focus on only a small subset of programming languages, excluding many popular languages such as Go or Rust. To ameliorate this issue, we present the BabelCode framework for execution-based evaluation of any benchmark in any language. BabelCode enables new investigations into the qualitative performance of models' memory, runtime, and individual test case results. Additionally, we present a new code translation dataset called Translating Python Programming Puzzles (TP3) from the Python Programming Puzzles (Schuster et al. 2021) benchmark that involves translating expert-level python functions to any language. With both BabelCode and the TP3 benchmark, we investigate if balancing the distributions of 14 languages in a training dataset improves a large language model's performance on low-resource languages. Training a model on a balanced corpus results in, on average, 12.34% higher $pass@k$ across all tasks and languages compared to the baseline. We find that this strategy achieves 66.48% better $pass@k$ on low-resource languages at the cost of only a 12.94% decrease to high-resource languages. In our three translation tasks, this strategy yields, on average, 30.77% better low-resource $pass@k$ while having 19.58% worse high-resource $pass@k$.