Towards Probabilistic Question Answering Over Tabular Data

Name: Towards Probabilistic Question Answering Over Tabular Data
Uploaded: 2026-10-06T10:05:01.589383+00:00
Duration: 1196 s
Description: Current methods for answering questions over tables struggle with reasoning under uncertainty. This work introduces LUCARIO, a new benchmark for probabilistic QA over tabular data, and Auto-BN, a framework that uses Bayesian Networks and Large Language Models to compute answers based on statistical rigor.

Current methods for answering questions over tables struggle with reasoning under uncertainty. This work introduces LUCARIO, a new benchmark for probabilistic QA over tabular data, and Auto-BN, a framework that uses Bayesian Networks and Large Language Models to compute answers based on statistical rigor.

Open the full episode