A new dataset for multi-table question answering has been developed, enhancing models' ability to derive answers through improved reasoning supervision. The use of synthetic contrastive reasoning has led to significant performance gains in various large language models, indicating a promising advancement in AI capabilities for complex data interactions.