Themis-Git-Commits-Merged

Overview

Themis-Git-Commits-Merged is a large-scale dataset of ~3.98M single-file code commits from permissively licensed GitHub repositories that have been cross-referenced with GHTorrent pull request data to retain only commits that are part of successfully merged, non-reverted pull requests. This provides implicit human validation of each code change — a merge decision by project maintainers confirms the intent and quality of the commit.

This dataset is derived from Themis-Git-Commits (the raw, unfiltered commit pool) by joining against GHTorrent pull request records through end of 2021 (the temporal extent of the GHTorrent PR data used). It serves as the direct input to the aspect classification and preference construction stages of the Themis pipeline, which produces the commit-based preference pairs in Themis-CodePreference and Themis-CodeRewardBench.

The SQL query used for initial commit mining restricts to repositories under permissive open-source licenses only (MIT, Apache-2.0, BSD-2/3-Clause, ISC, CC0-1.0, EPL-1.0, MPL-2.0, Unlicense, AGPL-3.0, LGPL-2.1, Artistic-2.0). The BigQuery snapshot used contains commits up to early 2022 — predating the widespread availability of LLM code generation tools — ensuring that all code changes in the dataset represent genuine human-authored preferences.

Relationship to Themis-Git-Commits

	Themis-Git-Commits	Themis-Git-Commits-Merged (this dataset)
Scope	All mined single-file commits	Only commits in merged, non-reverted PRs
Extension filtering	No	Yes (file extension matches target language)
PR validation	No	Yes (via GHTorrent, through end of 2021)
Human validation	Implicit (authored by humans)	Explicit (merge decision by maintainers)
Downstream use	Raw pool for further filtering	Direct input to aspect classification

Collection Pipeline

The commit mining pipeline is described in detail in the Themis paper and the Dataset folder in the GitHub repository. The BigQuery SQL query and scraping infrastructure are modified from the OctoPack pipeline (CommitPack); the subsequent filtering, classification, and preference construction stages are original to Themis.

1. BigQuery Mining — A GoogleSQL query (modified from OctoPack) extracts single-file commits from permissively licensed repositories in bigquery-public-data.github_repos, filtering for target programming languages and non-trivial commit messages.

2. Repository Reputation Filtering — Commits are subset to those originating from curated high-reputation repositories (15+ GitHub stars, 5+ contributors, 10+ issues).

3. Extension Filtering — Commits are further filtered so the changed file's extension matches a target programming language.

4. Content Retrieval — The pre-commit (old_contents) and post-commit (new_contents) file contents are fetched from GitHub via shallow git fetches using retrieve_commit_contents.py.

5. MinHash Deduplication — Near-duplicate content is removed using MinHash LSH deduplication (shingle size 5, 256 permutations, Jaccard threshold 0.7).

6. Pull Request Cross-Referencing (this step produces this dataset) — Commits are joined with GHTorrent pull request data (through end of 2021) to retain only non-reverted commits that are part of successfully merged pull requests from reputable repositories, ensuring implicit human validation of each code change.

Downstream Processing (Not in This Dataset)

The steps below are applied downstream to produce the final preference pairs in Themis-CodePreference and Themis-CodeRewardBench, and are not reflected in this dataset:

• Language Subsetting — The 24 languages in this dataset are narrowed to the 8 target languages (C, C#, C++, Go, Java, JavaScript, Python, Ruby) used in Themis-CodePreference and Themis-CodeRewardBench.
• Temporal Subsetting — For training data (Themis-CodePreference), only commits pushed before March 2019 are retained. For benchmark data (Themis-CodeRewardBench), commits are scoped to June 2019 – January 2021 from disjoint repositories.
• Aspect Classification — Commits are assigned to quality dimensions (Functional Correctness, Runtime Efficiency, Memory Efficiency, Security Hardness, Readability & Maintainability) using criteria-specialized ModernBERT commit classifiers, trained on seed positives retrieved via curated term lists.
• LLM Scoring & Instruction Synthesis — Frontier LMs validate preference strength and generate realistic inverse instructions.
• LLM-as-a-Judge Preference Labelling — Multi-sample voting with frontier LMs produces consensus preference labels.

Dataset Schema

Column	Type	Description
commit	string	Git commit SHA
subject	string	First line of the commit message
message	string	Full commit message body
repos	string	Comma-separated list of repository names containing this commit
file_path	string	Path of the changed file
license	string	SPDX license identifier of the source repository
unix_time	int64	Committer timestamp (seconds since epoch)
new_contents	string	File contents after the commit (post-commit)
old_contents	string	File contents before the commit (pre-commit)

Language Distribution

The dataset is partitioned by programming language, with one config per language. Each config has a single train split.

Language	Commits		Language	Commits
Python	914,849		Rust	57,383
JavaScript	694,051		Scala	46,557
Java	402,369		Swift	40,271
Ruby	345,124		Groovy	16,723
PHP	281,390		PowerShell	16,729
C++	194,218		Kotlin	16,467
Go	189,131		Erlang	15,623
C	172,736		Haskell	13,851
C#	138,000		Dart	8,890
TypeScript	114,769		Perl	4,930
Shell	95,305		Julia	2,662
			R	656
			Assembly	507
Total				~3.98M

Filters Applied

All filters from the upstream Themis-Git-Commits pipeline apply, plus the PR merge filter:

Filter	Purpose
License allowlist	MIT, Apache-2.0, BSD-2-Clause, BSD-3-Clause, ISC, CC0-1.0, EPL-1.0, MPL-2.0, Unlicense, AGPL-3.0, LGPL-2.1, Artistic-2.0
Language allowlist (SQL)	Python, Java, JavaScript, C, C#, C++, TypeScript, Go, Ruby (repo-level filter in BigQuery query)
Message length	10 < length < 15,000 characters
Message blocklist	~50 low-signal messages excluded (e.g., "initial commit", "wip", "yolo")
Pattern exclusion	Merge commits and CI push messages filtered out
Same-path constraint	old_path = new_path — file was modified in place, not renamed or moved
Single-file constraint	Commit touches exactly one file
Content retrieval	Both pre-commit and post-commit file contents successfully fetched
Near-deduplication	MinHash LSH with Jaccard threshold 0.7
Extension filtering	Changed file's extension mapped to one of 24 target programming languages (see distribution table)
PR merge filter	Commit is part of a successfully merged, non-reverted pull request (GHTorrent data through end of 2021)

Usage

from datasets import load_dataset

# Load a single language
python_commits = load_dataset("project-themis/git-commits-merged", "Python")
sample = python_commits["train"][0]
print(f"Commit: {sample['commit']}")
print(f"Subject: {sample['subject']}")
print(f"License: {sample['license']}")
print(f"File: {sample['file_path']}")

# Load all languages
for lang in ["C", "CSharp", "Cpp", "Go", "Java", "JavaScript", "Python", "Ruby", "TypeScript"]:
    ds = load_dataset("project-themis/git-commits-merged", lang)
    print(f"{lang}: {len(ds['train'])} commits")

License

This dataset is released under the Apache 2.0 License. The source commits are drawn exclusively from repositories with permissive open-source licenses (see filter table above).

Citation

@article{themis2025,
  title={Themis: Training Robust Multilingual Code Reward Models for Flexible Multi-Criteria Scoring},
  author={Paul, Indraneil and Gurevych, Iryna and Glava\v{s}, Goran},
  journal={arXiv preprint arXiv:2605.00754},
  year={2025}
}