Resource type
Corpus
Description
Dataset arranged in benchmark mode for the evaluation of the ability of language models (LLMs) to reason and answer questions about data stored in tables. Databench is composed of 65 tabular datasets from real problems.
Databench is the result of collaboration with Graphext and CardiffNLP.
How to cite
Jorge Jorge Osés Grijalba and Luis Alfonso Ureña-López and
Eugenio Martínez Cámara and Jose Camacho-Collados. (2024). Question Answering over Tabular Data with DataBench: A Large-Scale Empirical Evaluation of LLMs. En Proceedings of LREC-COLING 2024, Turiín, Italia.
Enlace