201 source documents and 2,024 benchmark questions with topology and evidence annotations, created by diandianone123 and last updated on May 7, 2026. The benchmark annotations and metadata are provided as JSON files, while larger document artifacts are stored as compressed archives.
Use Cases
- Benchmarking retrieval-augmented generation (RAG) systems based on topology and evidence annotations.
- Evaluating document-level information retrieval (MMDocIR) performance based on the provided questions and source documents.
- Training models for complex question answering that require evidence extraction from multiple documents.
Strengths
- Contains 2,024 benchmark questions with structured topology and evidence annotations.
- Includes 201 source documents (MMDocIR) used as the basis for the benchmark.
- Large document artifacts are compressed to keep repository size manageable.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count for the benchmark questions is unknown, which may limit suitability assessment.
- Last updated 2026-05-07 07:38:19; freshness should be verified.
Provenance
- Source
- huggingface
- Freshness
- 2026-05-07 07:38:19