Cooki-e/search-swe-development
Search-SWE development inputs — bingyu Public input assets for the implementation task task-1-x-1, authored by bingyu. This is a contributor development dataset, not the official Search-SWE release. Files live under development/bingyu/1-x-1/: corpus/: YFCC-10M vectors, CSR tags and format metadata. validation/: ten public development queries and exact filtered Top-10 labels. example/: an independent synthetic 64-document example with 17 queries. The nine files total 2,865,730… See the full description on the dataset page: https://huggingface.co/datasets/Cooki-e/search-swe-development.
Search-SWE development inputs — bingyu
Public input assets for the implementation task task-1-x-1, authored by bingyu. This is a contributor development dataset, not the official Search-SWE release.
Files live under development/bingyu/1-x-1/:
corpus/: YFCC-10M vectors, CSR tags and format metadata.validation/: ten public development queries and exact filtered Top-10 labels.example/: an independent synthetic 64-document example with 17 queries.
The nine files total 2,865,730,488 bytes. manifest.json records their exact sizes and SHA-256 hashes. In the GitHub task package they are restored under data/, using the immutable HF revision pinned in that package's assets.json.
The upstream YFCC-10M release is attributed and linked under CC BY 4.0 in SOURCES.md. That file distinguishes upstream data and derived inputs from task-authored metadata and the synthetic fixture, whose standalone license has not been supplied. The repository-level other label refers to those per-file terms; it does not replace or relicense the upstream CC BY 4.0 grant.
Only public input assets are included. Held-out task queries/labels, reference solutions, job results, credentials and task code are not part of this dataset. Keep this pinned development snapshot available through task review and until any official replacement has been published and verified.
