Summary of Long Code Arena: a Set Of Benchmarks For Long-context Code Models, by Egor Bogomolov et al.
Long Code Arena: a Set of Benchmarks for Long-Context Code Models
by Egor Bogomolov, Aleksandra Eliseeva, Timur Galimzyanov, Evgeniy Glukhov, Anton Shapkin, Maria Tigina, Yaroslav Golubev, Alexander Kovrigin, Arie van Deursen, Maliheh Izadi, Timofey Bryksin
First submitted to arxiv on: 17 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Software Engineering (cs.SE)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper bridges the gap in code processing benchmarks by introducing Long Code Arena, a suite of six tasks that require project-wide context. The Long Code Arena covers various aspects of code processing, including library-based code generation, CI builds repair, and commit message generation. The authors provide manually verified datasets for testing, an evaluation suite, and open-source baseline solutions based on popular language models (LLMs). These solutions showcase the usage of the dataset and simplify adoption by other researchers. The paper’s introduction is a significant step forward in advancing code processing capabilities. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Long Code Arena is a new benchmark that helps computers understand large projects better. It includes six tasks, such as writing code suggestions and fixing broken builds. The authors provide data for testing and solutions to get started quickly. This means researchers can focus on developing new ideas rather than creating their own data and starting from scratch. |
