The field of Large Language Models (LLMs) is moving towards more robust and accurate methods for tabular reasoning and query optimization. Recent developments have focused on improving the ability of LLMs to generate and execute SQL queries, as well as to evaluate the semantic equivalence of generated SQL. Additionally, there is a growing interest in using LLMs for domain-specific applications, such as nuclear power plants, where data privacy and security are of utmost importance. Noteworthy papers in this area include: LLM-Symbolic Integration for Robust Temporal Tabular Reasoning, which introduces a synthetic dataset and a symbolic intermediate representation to enhance generalization and mitigate biases. TableRAG: A Retrieval Augmented Generation Framework for Heterogeneous Document Reasoning, which proposes a hybrid framework that unifies textual understanding and complex manipulations over tabular data, establishing a new state-of-the-art for heterogeneous document question answering.
Advances in Large Language Models for Tabular Reasoning and Query Optimization
Sources
LLM-Symbolic Integration for Robust Temporal Tabular Reasoning
Training-Free Query Optimization via LLM-Based Plan Similarity
Small Models, Big Support: A Local LLM Framework for Teacher-Centric Content Creation and Assessment using RAG and CAG
Reinforcing Code Generation: Improving Text-to-SQL with Execution-Based Learning
Towards Secure and Private Language Models for Nuclear Power Plants
Unlocking the Potential of Large Language Models in the Nuclear Industry with Synthetic Data
Enhancing Accuracy and Maintainability in Nuclear Plant Data Retrieval: A Function-Calling LLM Approach Over NL-to-SQL
Taming SQL Complexity: LLM-Based Equivalence Evaluation for Text-to-SQL
Bridging the Gap Between Open-Source and Proprietary LLMs in Table QA
TableRAG: A Retrieval Augmented Generation Framework for Heterogeneous Document Reasoning