PaDaS-Lab/ecir2027-curse-dense-retrieval-run-files
Dataset · https://huggingface.co/datasets/PaDaS-Lab/ecir2027-curse-dense-retrieval-run-files
Full-corpus run files: Revisiting Dense Information Retrieval for Large Index Sizes This dataset contains the retrieval run files (top 1,000 documents per query) of the ECIR 2027 reproducibility paper "Revisiting Dense Information Retrieval for Large Index Sizes". They were produced by BM25, SPLADE and three uncompressed dense retrievers on the full (not subsampled) corpora used in the paper. The code that produced and consumes them is the code repository linked in the paper.… See the full description on the dataset page: https://huggingface.co/datasets/PaDaS-Lab/ecir2027-curse-dense-retrieval-run-files.
First seen by Oriel on .
AI summary
This dataset provides full-corpus retrieval runs for an ECIR 2027 reproducibility paper comparing BM25, SPLADE, and uncompressed dense retrievers at large index sizes.
- It includes gzipped TREC runs with up to 1,000 retrieved documents per query across nine corpora, including 100-million-document CoRE and synthetic-noise corpora.
- Six systems are represented: two BM25 implementations, SPLADE v3, and three dense embedding models; dense runs use exact brute-force cosine search over 768-dimensional embeddings.
- The files support rebuilding the MS MARCO v1 pooled corpus sample and recomputing the significance tests in Table 4.
- The download is about 3.2 GB; subsampled-corpus runs and truncated or binarized dense configurations are not included.
- One Elasticsearch run omits a query on random-noise corpora, and dense-gemma is absent for one random-words corpus.
#information retrieval #dense retrieval #bm25 #splade
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Datasets: 12, latest observation ; +2 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Datasets: 94, latest observation ; +30 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Datasets | PaDaS-Lab/ecir2027-curse-dense-retrieval-run-files | https://huggingface.co/datasets/PaDaS-Lab/ecir2027-curse-dense-retrieval-run-files Read discussion |