ColHNQue (ColPaliHardNegativeQueries) is a dataset introduced in the paper 'DocReRank: Single‑Page Hard Negative Query Generation for Training Multi‑Modal RAG Rerankers'. It addresses limitations of document-level mining by generating hard negative queries at the page/image level. The dataset was created by DocReRank and was last updated on July 22, 2025.
Use Cases
- Training reranker models based on page-level hard negative queries.
- Evaluating retrieval performance based on semantically similar but incorrect queries.
- Benchmarking multimodal RAG systems based on the described query generation method.
Strengths
- The dataset is designed to address a specific limitation in document-level hard negative mining.
- It was introduced in a published research paper, suggesting a research-driven creation method.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- DocReRank
- Collection Method
- Likely generated for the research paper 'DocReRank: Single‑Page Hard Negative Query Generation for Training Multi‑Modal RAG Rerankers'.
- Freshness
- Last updated 2025-07-22 11:18:06; freshness should be verified.