SWE-QA: A Dataset and Benchmark for Complex Code Understanding

TL;DR

SWE-QA introduces a comprehensive dataset and benchmark for evaluating complex, multi-hop code understanding in Python, highlighting the challenges faced by current language models in real-world software reasoning.

Contribution

The paper presents a novel dataset and benchmark that focus on multi-hop reasoning in code comprehension, bridging the gap between simple tasks and real-world software development complexity.

Findings

Current language models struggle with multi-hop reasoning in code understanding.

Dense architectures outperform mixture-of-experts models by 10-14 percentage points.

Best model achieves 74.41% accuracy on the SWE-QA benchmark.

Abstract

In this paper, we introduce SWE-QA, a text and code corpus aimed at benchmarking multi-hop code comprehension, addressing the gap between simplified evaluation tasks and the complex reasoning required in real-world software development. While existing code understanding benchmarks focus on isolated snippets, developers must routinely connect information across multiple dispersed code segments. The dataset comprises 9,072 multiple-choice questions systematically generated from 12 Python repositories of SWE-bench, evaluating several recurrent reasoning patterns like Declaration-and-Call questions that link entity definitions to their usage, and Interacting-Entity questions that examine the dynamic relationships among multiple collaborating components. Generated through parsing-based entity extraction and Large Language Model assisted question construction with carefully validated distractors, the benchmark distinguishes genuine comprehension from superficial pattern matching. Evaluation of 15 language models (360M to 671B parameters) reveals significant challenges in multi-hop reasoning, with best performance reaching 74.41% accuracy. Dense architectures consistently outperform mixture-of-experts models by 10-14 percentage points, while reasoning-enhanced variants show inconsistent benefits.