Update README.md

README.md

@@ -22,10 +22,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 25664270.827623513
     num_examples: 507
-  download_size: 6383845
-  dataset_size: 25664270.827623513
 - config_name: C
   features:
   - name: commit
@@ -48,11 +45,8 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 13034110929.154518
     num_examples: 172736
-  download_size: 4953140102
-  dataset_size: 13034110929.154518
-- config_name: C#
+- config_name: CSharp
   features:
   - name: commit
     dtype: string
@@ -74,11 +68,8 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 5397646765.3076105
     num_examples: 138000
-  download_size: 1656363259
-  dataset_size: 5397646765.3076105
-- config_name: C++
+- config_name: Cpp
   features:
   - name: commit
     dtype: string
@@ -100,10 +91,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 23547757257
     num_examples: 194218
-  download_size: 7569744858
-  dataset_size: 23547757257
 - config_name: Dart
   features:
   - name: commit
@@ -126,10 +114,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 397294625.4988167
     num_examples: 8890
-  download_size: 154161531
-  dataset_size: 397294625.4988167
 - config_name: Erlang
   features:
   - name: commit
@@ -152,10 +137,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 889166050
     num_examples: 15623
-  download_size: 262956120
-  dataset_size: 889166050
 - config_name: Go
   features:
   - name: commit
@@ -178,10 +160,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 13724186600.079618
     num_examples: 189131
-  download_size: 7560506015
-  dataset_size: 13724186600.079618
 - config_name: Groovy
   features:
   - name: commit
@@ -204,10 +183,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 569309862
     num_examples: 16723
-  download_size: 156155250
-  dataset_size: 569309862
 - config_name: Haskell
   features:
   - name: commit
@@ -230,10 +206,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 584392586
     num_examples: 13851
-  download_size: 208126412
-  dataset_size: 584392586
 - config_name: Java
   features:
   - name: commit
@@ -256,10 +229,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 15135951136.42295
     num_examples: 402369
-  download_size: 5020550095
-  dataset_size: 15135951136.42295
 - config_name: JavaScript
   features:
   - name: commit
@@ -282,10 +252,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 36390392763.94772
     num_examples: 694051
-  download_size: 12357643118
-  dataset_size: 36390392763.94772
 - config_name: Julia
   features:
   - name: commit
@@ -308,10 +275,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 99752915.85875197
     num_examples: 2662
-  download_size: 35164706
-  dataset_size: 99752915.85875197
 - config_name: Kotlin
   features:
   - name: commit
@@ -334,10 +298,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 339000058
     num_examples: 16467
-  download_size: 101204442
-  dataset_size: 339000058
 - config_name: PHP
   features:
   - name: commit
@@ -360,10 +321,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 8837629617.411818
     num_examples: 281390
-  download_size: 3345666898
-  dataset_size: 8837629617.411818
 - config_name: Perl
   features:
   - name: commit
@@ -386,10 +344,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 260477378
     num_examples: 4930
-  download_size: 92901210
-  dataset_size: 260477378
 - config_name: PowerShell
   features:
   - name: commit
@@ -412,10 +367,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 553399997.1505141
     num_examples: 16729
-  download_size: 301061644
-  dataset_size: 553399997.1505141
 - config_name: Python
   features:
   - name: commit
@@ -438,10 +390,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 43401708067
     num_examples: 914849
-  download_size: 14393853016
-  dataset_size: 43401708067
 - config_name: R
   features:
   - name: commit
@@ -464,10 +413,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 22395918.040545523
     num_examples: 656
-  download_size: 4801427
-  dataset_size: 22395918.040545523
 - config_name: Ruby
   features:
   - name: commit
@@ -490,10 +436,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 4806098257
     num_examples: 345124
-  download_size: 1555780240
-  dataset_size: 4806098257
 - config_name: Rust
   features:
   - name: commit
@@ -516,10 +459,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 2750630964.173699
     num_examples: 57383
-  download_size: 1072232323
-  dataset_size: 2750630964.173699
 - config_name: Scala
   features:
   - name: commit
@@ -542,10 +482,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 1502090669
     num_examples: 46557
-  download_size: 507218235
-  dataset_size: 1502090669
 - config_name: Shell
   features:
   - name: commit
@@ -568,10 +505,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 2727002433.72094
     num_examples: 95305
-  download_size: 623236489
-  dataset_size: 2727002433.72094
 - config_name: Swift
   features:
   - name: commit
@@ -594,10 +528,7 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 831556938.5793096
     num_examples: 40271
-  download_size: 345171247
-  dataset_size: 831556938.5793096
 - config_name: TypeScript
   features:
   - name: commit
@@ -620,124 +551,24 @@ dataset_info:
     dtype: string
   splits:
   - name: train
-    num_bytes: 11236698679
     num_examples: 114769
-  download_size: 3124417123
-  dataset_size: 11236698679
-configs:
-- config_name: Assembly
-  data_files:
-  - split: train
-    path: Assembly/train-*
-- config_name: C
-  data_files:
-  - split: train
-    path: C/train-*
-- config_name: C#
-  data_files:
-  - split: train
-    path: C#/train-*
-- config_name: C++
-  data_files:
-  - split: train
-    path: C++/train-*
-- config_name: Dart
-  data_files:
-  - split: train
-    path: Dart/train-*
-- config_name: Erlang
-  data_files:
-  - split: train
-    path: Erlang/train-*
-- config_name: Go
-  data_files:
-  - split: train
-    path: Go/train-*
-- config_name: Groovy
-  data_files:
-  - split: train
-    path: Groovy/train-*
-- config_name: Haskell
-  data_files:
-  - split: train
-    path: Haskell/train-*
-- config_name: Java
-  data_files:
-  - split: train
-    path: Java/train-*
-- config_name: JavaScript
-  data_files:
-  - split: train
-    path: JavaScript/train-*
-- config_name: Julia
-  data_files:
-  - split: train
-    path: Julia/train-*
-- config_name: Kotlin
-  data_files:
-  - split: train
-    path: Kotlin/train-*
-- config_name: PHP
-  data_files:
-  - split: train
-    path: PHP/train-*
-- config_name: Perl
-  data_files:
-  - split: train
-    path: Perl/train-*
-- config_name: PowerShell
-  data_files:
-  - split: train
-    path: PowerShell/train-*
-- config_name: Python
-  data_files:
-  - split: train
-    path: Python/train-*
-- config_name: R
-  data_files:
-  - split: train
-    path: R/train-*
-- config_name: Ruby
-  data_files:
-  - split: train
-    path: Ruby/train-*
-- config_name: Rust
-  data_files:
-  - split: train
-    path: Rust/train-*
-- config_name: Scala
-  data_files:
-  - split: train
-    path: Scala/train-*
-- config_name: Shell
-  data_files:
-  - split: train
-    path: Shell/train-*
-- config_name: Swift
-  data_files:
-  - split: train
-    path: Swift/train-*
-- config_name: TypeScript
-  data_files:
-  - split: train
-    path: TypeScript/train-*
+tags:
+- code
+- github
+- commits
+- multilingual
 license: apache-2.0
 task_categories:
 - text-generation
 language:
-- en
-tags:
 - code
-- multilingual
-- github
-- commits
 size_categories:
-- 10M<n<100M
+- 1M<n<10M
 ---
 
 <div align="center">
 
-# Themis-Git-Commits
+# Themis-Git-Commits-Merged
 
 [![arXiv](https://img.shields.io/badge/arXiv-xxxx.xxxxx-b31b1b.svg)](https://arxiv.org/abs/xxxx.xxxxx)
 [![Models](https://img.shields.io/badge/%F0%9F%A4%97%20Models-Themis--RM-yellow)](https://huggingface.co/collections/project-themis/themis-reward-model-collection)
@@ -748,15 +579,31 @@ size_categories:
 
 ## Overview
 
-**Themis-Git-Commits** is a large-scale dataset of single-file code commits mined from openly licensed GitHub repositories via the [BigQuery GitHub public dataset](https://console.cloud.google.com/marketplace/product/github/github-repos). This is the **raw commit dataset** — prior to merging with pull request data to subset only for merged commits. It serves as the foundational data source for the commit-based preference pairs in [Themis-CodePreference](https://huggingface.co/collections/project-themis/themis-preference-datasets-and-benchmarks), which is used to train the [Themis-RM](https://huggingface.co/collections/project-themis/themis-reward-model-collection) suite of multilingual code reward models.
+**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](https://ghtorrent.org/) 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](https://huggingface.co/datasets/project-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](https://github.com/iNeil77/Themis) pipeline, which produces the commit-based preference pairs in [Themis-CodePreference](https://huggingface.co/datasets/project-themis/Themis-CodePreference) and [Themis-CodeRewardBench](https://huggingface.co/datasets/project-themis/Themis-CodeRewardBench).
 
-Each row represents a single commit that changes exactly one file in a repository with a permissive open-source license. The dataset includes the commit metadata (SHA, message, timestamp, license) along with the pre-commit and post-commit file contents, enabling downstream construction of code-change preference pairs across multiple quality dimensions.
+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
+
+<div align="center">
+
+| | [Themis-Git-Commits](https://huggingface.co/datasets/project-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 |
+
+</div>
 
 ## Collection Pipeline
 
-The commit mining pipeline is described in detail in the [Themis paper](https://arxiv.org/abs/xxxx.xxxxx) and the [Dataset](https://github.com/iNeil77/Themis/tree/main/Dataset) folder in the GitHub repository. At a high level:
+The commit mining pipeline is described in detail in the [Themis paper](https://arxiv.org/abs/xxxx.xxxxx) and the [Dataset](https://github.com/iNeil77/Themis/tree/main/Dataset) folder in the GitHub repository. The BigQuery SQL query and scraping infrastructure are modified from the [OctoPack](https://arxiv.org/abs/2308.07124) pipeline ([CommitPack](https://huggingface.co/datasets/bigcode/commitpack)); the subsequent filtering, classification, and preference construction stages are original to Themis.
 
-1. **BigQuery Mining** — A [GoogleSQL query](https://github.com/iNeil77/Themis/blob/main/Dataset/Commit_Mining_SQL/consolidated_query.sql) extracts single-file commits from `bigquery-public-data.github_repos`, filtering for permissive licenses, target programming languages, and non-trivial commit messages.
+1. **BigQuery Mining** — A [GoogleSQL query](https://github.com/iNeil77/Themis/blob/main/Dataset/Commit_Mining_SQL/consolidated_query.sql) (modified from [OctoPack](https://arxiv.org/abs/2308.07124)) 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](https://github.com/iNeil77/Themis/tree/main/Dataset/Repos) (15+ GitHub stars, 5+ contributors, 10+ issues).
 
@@ -766,11 +613,14 @@ The commit mining pipeline is described in detail in the [Themis paper](https://
 
 5. **MinHash Deduplication** — Near-duplicate content is removed using [MinHash LSH deduplication](https://github.com/iNeil77/Themis/blob/main/Dataset/Utils/minHash_dedupe_local.py) (shingle size 5, 256 permutations, Jaccard threshold 0.7).
 
+6. **Pull Request Cross-Referencing** *(this step produces this dataset)* — Commits are joined with [GHTorrent](https://ghtorrent.org/) 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 are **not** reflected in this raw 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:
 
-- **Pull Request Merging** — Commits are cross-referenced with GHTorrent pull request data to retain only non-reverted commits that are part of successfully merged pull requests, ensuring implicit human validation.
+- **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](https://huggingface.co/datasets/project-themis/Themis-CodePreference) and [Themis-CodeRewardBench](https://huggingface.co/datasets/project-themis/Themis-CodeRewardBench).
+- **Temporal Subsetting** — For training data ([Themis-CodePreference](https://huggingface.co/datasets/project-themis/Themis-CodePreference)), only commits pushed before **March 2019** are retained. For benchmark data ([Themis-CodeRewardBench](https://huggingface.co/datasets/project-themis/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](https://huggingface.co/answerdotai/ModernBERT-base) commit classifiers, trained on seed positives retrieved via [curated term lists](https://github.com/iNeil77/Themis/tree/main/Dataset/Commit_Mining_Terms).
 - **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.
@@ -793,14 +643,41 @@ The steps below are applied downstream to produce the final preference pairs in
 
 </div>
 
-## Filters Applied During Mining
+## Language Distribution
+
+The dataset is partitioned by programming language, with one config per language. Each config has a single `train` split.
+
+<div align="center">
+
+| 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** |
+
+</div>
+
+## Filters Applied
+
+All filters from the upstream [Themis-Git-Commits](https://huggingface.co/datasets/project-themis/git-commits) pipeline apply, plus the PR merge filter:
 
 <div align="center">
 
 | 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** | Python, Java, JavaScript, C, C#, C++, TypeScript, Go, Ruby |
+| **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 |
@@ -808,6 +685,8 @@ The steps below are applied downstream to produce the final preference pairs in
 | **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) |
 
 </div>
 
@@ -816,16 +695,18 @@ The steps below are applied downstream to produce the final preference pairs in
 ```python
 from datasets import load_dataset
 
-dataset = load_dataset("project-themis/Themis-Git-Commits")
-
-# Inspect a sample
-sample = dataset["train"][0]
+# 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']}")
-print(f"Old contents length: {len(sample['old_contents'])}")
-print(f"New contents length: {len(sample['new_contents'])}")
+
+# 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
@@ -841,4 +722,4 @@ This dataset is released under the [Apache 2.0 License](https://www.apache.org/l
   journal={arXiv preprint arXiv:xxxx.xxxxx},
   year={2025}
 }
-```
+```