SWE-rebench-V2

# SWE-rebench-V2 ## Dataset Summary SWE-rebench-V2 is a curated dataset of software-engineering tasks derived from real GitHub issues and pull requests. The dataset contains 32,079 samples covering Python, Go, TypeScript, JavaScript, Rust, Java, PHP, Kotlin, Julia, Elixir, Scala, Swift, Dart, C, C++, C#, R, Clojure, OCaml, and Lua. For log parser functions, base Dockerfiles, and the prompts used, please see https://github.com/SWE-rebench/SWE-rebench-V2 The detailed technical report is available at [“SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale”](https://arxiv.org/abs/2602.23866). ## Quick Start ```python from datasets import load_dataset ds = load_dataset("nebius/SWE-rebench-V2", split="train") print(len(ds)) # 32079 ``` ## Dataset Structure | Field | Type | Description | |---|---|---| | `instance_id` | `string` | Unique identifier for the instance | | `repo` | `string` | GitHub repository in `owner/repo` format | | `base_commit` | `string` | Git commit SHA of the base before the fix | | `patch` | `string` | The gold patch that resolves the issue | | `test_patch` | `string` | Diff adding or modifying tests that verify the fix | | `problem_statement` | `string` | Issue description the patch addresses | | `pr_description` | `string` | Full pull request description | | `created_at` | `int64` | Unix timestamp (milliseconds) of the issue/PR creation | | `image_name` | `string` | Docker image name used for the evaluation environment | | `language` | `string` | Primary programming language of the repository | | `interface` | `string` | Description of the code interface changed by the patch | | `license` | `string` | SPDX license identifier of the repository | | `FAIL_TO_PASS` | `list[string]` | Test IDs that fail before the patch and pass after | | `PASS_TO_PASS` | `list[string]` | Test IDs that pass both before and after the patch | | `install_config` | `struct` | Configuration needed to reproduce the test environment | | `meta` | `struct` | Metadata and LLM-generated quality annotations | # License The dataset is licensed under the Creative Commons Attribution 4.0 license. However, please respect the license of each specific repository on which a particular instance is based. To facilitate this, the license of each repository at the time of the commit is provided for every instance. # Citation ```bibtex @misc{badertdinov2026swerebenchv2languageagnosticswe, title={SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale}, author={Ibragim Badertdinov and Maksim Nekrashevich and Anton Shevtsov and Alexander Golubev}, year={2026}, eprint={2602.23866}, archivePrefix={arXiv}, primaryClass={cs.SE}, url={https://arxiv.org/abs/2602.23866}, }