DORABench is a benchmark artifact for evaluating dataset-oriented reasoning and analysis workflows across multiple data sources. It includes benchmark data, metadata, indexes, validation reports, scoring outputs, and agent run results for subsets of Australian Open Data and Hugging Face benchmarks. The artifact was created by Lisa-Yao Gan from DORABench Dataset Discoverability Artifacts and was last updated on May 8, 2026.
Use Cases
- Benchmarking dataset discovery agents based on the included Australian Open Data and Hugging Face subsets.
- Reproducing evaluation pipeline results based on the provided execution traces and scoring outputs.
- Inspecting source data and benchmark construction methods for methodological validation.
- Comparing agent performance across different data sources using the provided run results.
Strengths
- Includes multiple components for a full evaluation pipeline: benchmark data, metadata, indexes, validation reports, scoring outputs, and agent run results.
- Supports reproducibility by providing execution traces and final scoring results.
- Benchmarks dataset-oriented reasoning across two distinct data sources: Australian Open Data and Hugging Face.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and file size are unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- DORABench Dataset Discoverability Artifacts
- Collection Method
- Likely constructed as an artifact for a research evaluation pipeline.
- Freshness
- Last updated 2026-05-08 11:53:03; freshness should be verified.