table-reasoning

📚 A curated & auto-maintained reading list on LLM-based Data Agents — data preparation, NL2SQL, table reasoning, data analysis, data science & AutoML agents, database operations, agent memory, and benchmarks.

[NAACL 2024] TabSQLify: Enhancing Reasoning Capabilities of LLMs Through Table Decomposition

[NAACL 2025] This is the official code for the paper "H-STAR: LLM-driven Hybrid SQL-Text Adaptive Reasoning on Tables"

(EMNLP 2025) Weaver: A modular agentic pipeline that dynamically combines SQL and LLMs for advanced table-based question answering

[EMNLP 2024] NormTab: Improving Symbolic Reasoning in LLMs Through Tabular Data Normalization

This repository contain the implementation of DenTab: A Dataset for Table Recognition and Visual QA on Real-World Dental Estimates

MMTabReal is a benchmark suite for multimodal table reasoning on real-world tables with mixed text, images, charts, maps, and visual encodings. It includes curated QA pairs, baseline implementations, and evaluation scripts for multimodal table understanding research.

A small, locally-trained LLM answers constrained table questions and cites the exact cells it used: the model comprehends, a deterministic engine does the arithmetic. 59% to 96% answer accuracy on a locked test.

(IJCNLP-AACL 2025) MAMMQA: Rethinking Information Synthesis in Multimodal Question Answering from a Multi-Agent Perspective