An agentic system to answer questions on tabular data

LLM-Powered Data Discovery System for Tabular Data

This repository provides data, code, and experiments for evaluating how well LLMs can predict statistical correlations. We propose and assess the Logit-Based Calibrated Prior (LCP), a novel method for using LLM logits to estimate correlation priors in a calibrated way.

Identify useful correlations efficiently in a large repository of spatio-temporal tabular datasets

Temporary repository to host demo scripts combining Pneuma and Palimpzest

Data Discovery Tools and Systems