Skip to content

Dataset · nebius

126k real GitHub pull requests in 18+ languages

nebius/SWE-rebench-V2-PRs

Large-scale corpus of real GitHub pull requests by Nebius for training and evaluating coding agents.

What's inside

The dataset contains 126,300 examples of real English-language GitHub pull requests. It covers numerous programming languages, including Go, Python, JavaScript, TypeScript, Rust, Java, C, C++, Julia, Elixir, Kotlin, PHP, and others. The collection is designed for developing and testing code generation and software engineering systems.

What you can do with it

  • Fine-tuning LLMs and autonomous agents for automated bug fixing and patch writing
  • Evaluating code generation quality across multiple popular programming languages simultaneously
Tasks
Text generation
Languages
English
Size
100K–1M rows
Downloads this month
64.8K
Likes
19

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nebius/SWE-rebench-V2-PRs")