SALT-KG: A Benchmark for Semantics-Aware Learning on Enterprise Tables
Isaiah Onando Mulang’, Felix Sasaki, Tassilo Klein, Jonas Kolk, Nikolay Grechanov, Johannes Hoffart · 2026
Read SALT-KG: A Benchmark for Semantics-Aware Learning on Enterprise Tables by Isaiah Onando Mulang’, Felix Sasaki, Tassilo Klein, Jonas Kolk, Nikolay Grechanov, Johannes Hoffart (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-07638.txt.
Description
Isaiah Onando Mulang' and colleagues build a benchmark for learning from business tables with meaning attached. SALT-KG extends the earlier SALT benchmark for relational prediction. The new version links multi-table transaction data to a structured knowledge graph of operational business knowledge. That graph records field-level descriptions, relations between fields, and business object types. Their 2026 preprint defines the resource.
Most table benchmarks give models numbers without the business context behind them. Semantics-aware learning tests whether models can use that context to predict better. Enterprise data is exactly where such context exists but is rarely shared. A public benchmark lets teams compare methods fairly. The paper targets machine learning scholars who work with relational and enterprise data.
Catalog record
| CREATOR | Isaiah Onando Mulang’, Felix Sasaki, Tassilo Klein, Jonas Kolk, Nikolay Grechanov, Johannes Hoffart |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-07638