-
Have your Queries Already Seen the Data? Data-Privilege in Tabular Benchmarks
We take a look at how many popular benchmarks for evaluating natural language interfaces to tabular data contain "data-privileged" queries—questions that reference specific components of the data that real users wouldn't know about in open-domain settings.
-
TRL reading group discussion: "Large Language Models(LLMs) on Tabular Data: Prediction, Generation, and Understanding - A Survey"
Discussion protocol and results of the TRL reading group on the paper "Large Language Models(LLMs) on Tabular Data: Prediction, Generation, and Understanding - A Survey" by X. Fang et al. (2024).