datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
unity-dev-instructions
Unity Developer Instructions
A comprehensive instruction-tuning dataset for Unity game development,
covering C# scripting, XR/VR development, physics, animation, rendering,
UI Toolkit, and performance optimization.
Dataset Summary
Split
Count
Train
46,483
Test
2,446
Total
48,929
Data Sources
| unity_docs | 40,496 |
| stackoverflow | 6,071 |
| github | 2,362 |
Source breakdown:
Source
Count
unity_docs
40,496
stackoverflow
6,071… See the full description on the dataset page: https://huggingface.co/datasets/vishnuOI/unity-dev-instructions.unity_api_2022_3
Unity3d 2022.3 LTS API & Manual
In this dataset, you'll find a series of Q&A for the Unity3d API and Manual.
Dataset Creation
Download the unity offline documentation.
Process documentation, extract title, and description. Clean documentation.
Process each title and description item in llama3-8B-Instruct in order to generate several questions that capture the meaning of the API.
Re-process in llama3-8B-Instruct with question and API to generate the answer.… See the full description on the dataset page: https://huggingface.co/datasets/Hypersniper/unity_api_2022_3.pub-instruct-unity.preview
Pub-Instruct — Unity (Preview)
An open, license-respecting instruction-tuning dataset for Unity 6 (6000.0 LTS) C# development.
Records are generated clean-room by an open-weight model and validated by compiling every code
sample against pinned Unity 6 reference assemblies.
⚠️ Preview — 3,247 records at ~36% of planned coverage. This early cut is foundations-first
(Unity official docs, an OER textbook, and MIT-0 CatLikeCoding tutorials are well covered); applied… See the full description on the dataset page: https://huggingface.co/datasets/wrayy/pub-instruct-unity.preview.Unity_ResolvedQuestions
Unity Resolved Questions Dataset
This dataset contains curated Unity-related technical Q&A pairs, focused on practical problem-solving in Unity, C#, XR, URP/HDRP, and SketchUp.
Data Formats
data/hermes.jsonInstruction-style format (instruction, input, output), suitable for base / completion-style SFT.
data/messages.jsonlChat-style format (system, user, assistant), designed for instruction-following and chat models such as Mistral Instruct v0.3.
Intended Use… See the full description on the dataset page: https://huggingface.co/datasets/Erocal/Unity_ResolvedQuestions.unitybench
UnityBench — by 14Dimension Enterprise
https://www.14dimension.com/
A two-track evaluation for Unity 6 / C# coding models: domain knowledge + real-world bug review.
Generalist code models are strong — but how well do they actually handle Unity 6, and can they
find real bugs in real game code without crying wolf? There was no good public answer. This is one.
Built by 14Dimension Enterprise from a shipping Unity Game — bugs and fixes reconstructed from real
version-control… See the full description on the dataset page: https://huggingface.co/datasets/ldh-sky/unitybench.docs-instruct-unity-20260601-2007
docs-instruct-unity-20260601-2007
Synthetic instruction-tuning dataset generated by the DownFTuner pipeline.
Source: documentation site (unity), crawled and chunked.
Generator: DeepSeek-V4-Pro via Ollama, with EmbeddingGemma dedup + hallucination filter.
Format: chat-format JSONL (messages field), split into train.jsonl and valid.jsonl.
Source URLs are preserved in each row's source metadata where available.