TabJoinBench
TabJoinBench: A Benchmark for Joinable Table Discovery
A benchmark evaluating join discovery methods across semantic, relational, and hybrid data lake scenarios using WikiTables, MMQA, and NYC Open Data sources.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info tabjoinbench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
5 sourced entriesEvery number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
Content & Construction: TabJoinBench constructs query–candidate pairs from three distinct sources (WikiTables for semantic joins, MMQA for equi-joins, and NYC Open Data for realistic lakes). It introduces structural, representation, and semantic changes via composable perturbations while preserving ground truth.
Scoring & Findings: The benchmark evaluates set-based, feature-based, learned, and embedding approaches. It aims to facilitate reproducible evaluation by releasing processed datasets, annotations, and generation pipelines.