Datasets:
Contributing to ifc-bench
We welcome contributions to the ifc-bench dataset! This document outlines the guidelines for contributing new models, QA pairs, documentation, or code improvements.
How to Contribute
1. Reporting Issues
If you find a bug, an error in the dataset, or have a suggestion, please open an issue on our GitHub issue tracker.
2. Contributing New IFC Models
We are always looking for new IFC models to expand the dataset. When contributing a new model, please ensure:
- Licensing: The model must be available under a permissive open-source license (e.g., CC BY 4.0, MIT). Models with copyleft licenses (e.g., GPLv3) may be accepted but will be clearly flagged in the dataset documentation, as they impose additional redistribution requirements on downstream users.
- Format: The model must be in the IFC format.
- Documentation: Provide a brief description of the model, including its purpose, size, and complexity.
- Organization: Place the model files in a new folder under the
projects/directory. Include alicense.txtfile and amodel_card.md(Markdown format) with the model's metadata. - QA Pairs (Optional): Include question-answer pairs for the new model.
3. Contributing New QA Pairs
If you want to add more question-answer pairs to existing models:
- Format: Add new rows to the
questions/ifc-bench-v2.csvfile. - Category: Each question must include a
categoryvalue (1-4): 1=Direct Information Retrieval, 2=Computational Aggregation, 3=Geometric/Spatial Computation, 4=Incomplete Information. - Accuracy: Ensure the answers are accurate and verifiable.
- Clarity: Questions should be clear and unambiguous.
- Diversity: Try to cover different aspects of the model (spatial, properties, systems, etc.).
- Consistency: Follow the existing format for columns:
question,ground_truth,ifc_model,project,category.
4. Correcting Existing Answers
If you find inaccuracies in existing QA pairs:
- Verification: Provide evidence for the correction (screenshots, model measurements)
- Format: Modify the answer cell in
questions/ifc-bench-v2.csvwhile keeping the original question - Traceability: Include a brief explanation in the pull request description
5. Contributing Documentation
If you want to improve the documentation:
- Clarity: Ensure the documentation is clear, concise, and easy to understand.
- Accuracy: Ensure the documentation is accurate and up-to-date.
- Format: Follow the Markdown format.
- Organization: Place new documentation files in the
docs/directory.
Contribution Workflow
- Fork the repository on GitHub.
- Create a new branch for your changes.
- Make your changes and commit them with clear messages.
- Push your branch to your forked repository.
- Submit a pull request to the main repository.
Questions
If you have any questions, please open an issue on our GitHub issue tracker.
Before starting major work:
- Check open issues for existing discussions
- For large contributions, consider opening an issue first to discuss the approach
Thank you for your contributions!