The CogniSQL Positive Sample Corpus is a large-scale dataset of 36,356 weakly supervised queries, where each example is annotated with six semantically diverse reasoning paths leading to correct, executable SQL. This corpus is designed to support multi-path reasoning research and reinforcement learning approaches for Text-to-SQL generation.
Unlike traditional single-annotation datasets, this corpus captures the inherent… See the full description on the dataset page: https://huggingface.co/datasets/CogniSQL/Positive_Sample_Corpus.
