This folder keeps paper-specific documentation for the Data Efficacy codebase. The root README introduces the shared motivation and pipeline, while each page here records the details needed for a specific paper.
| Work | Status | Page |
|---|---|---|
| Data Efficacy for Language Model Training (DELT) | arXiv 2025 | Data Efficacy for Language Model Training |
| Demystifying Data Organization for Enhanced LLM Training | ACL 2026 | Demystifying Data Organization for Enhanced LLM Training |
When adding a new data efficacy paper:
- Add a new page under
docs/. - Put paper-specific figures under
docs/assets/<paper_slug>/. - Link the page from both this index and the root README.
- Keep reusable code in the shared modules, such as
data_scoring,data_selection,data_ordering,model_train, ormodel_eval.