Latent Bridges for Multi-Table Question Answering
Not on the map. Why?
Abstract (arXiv)
We introduce GRAB, a constructor-encoder-bridge pipeline for table question answering. Our method lifts relational data into an heterogeneous graph, encodes it via message passing, and transfers the signals to an LLM through a small set of query-conditioned latent tokens. This provides the LLM with a compact, task-relevant structural representation together with the flattened text. Crucially, the LLM remains strictly frozen to preserve its general reasoning capabilities; we train only the lightweight graph encoder and latent bridge (91M parameters), allowing the entire pipeline to be trained efficiently. Our pipeline significantly improves performance on relational Question Answering, with the largest gains in demanding multi-table settings, offering an efficient, principled way to connect relational deep learning with LLMs.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 35 references that are in the map. First posted on arXiv 27 Jun 2026.
Builds on
22 of 35 references.
References · 22
- 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,166
- 2018How Powerful are Graph Neural Networks?cited by 4,393
- 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,915
- 2017Neural Message Passing for Quantum Chemistrycited by 3,487
- 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,876
- 2018Weisfeiler and Leman Go Neural: Higher-order Graph Neural Networkscited by 881
- 2017Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learningcited by 687
- 2015Compositional Semantic Parsing on Semi-Structured Tablescited by 515
- 2020TAPAS: Weakly Supervised Table Parsing via Pre-trainingcited by 434
- 2023LoraHub: Efficient Cross-Task Generalization via Dynamic LoRA Compositioncited by 364
- 2022Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoningcited by 350
- 2021TAPEX: Table Pre-training via Learning a Neural SQL Executorcited by 235
10 more
- 2023TableLlama: Towards Open Large Generalist Models for Tablescited by 137
- 2022MultiHiertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Datacited by 121
- 2021HiTab: A Hierarchical Table Dataset for Question Answering and Natural Language Generationcited by 118
- 2024RelBench: A Benchmark for Deep Learning on Relational Databasescited by 96
- 2022TableFormer: Robust Transformer Modeling for Table-Text Encodingcited by 91
- 2023MultiTabQA: Generating Tabular Answers for Multi-Table Question Answeringcited by 32
- 2024Is Table Retrieval a Solved Problem? Exploring Join-Aware Multi-Table Retrievalcited by 27
- 2024TQA-Bench: Evaluating LLMs for Multi-Table Question Answeringcited by 17
- 2024Improving Retrieval-augmented Text-to-SQL with AST-based Ranking and Schema Pruningcited by 16
- 2024SCITAT: A Question Answering Benchmark for Scientific Tables and Text Covering Diverse Reasoning Typescited by 10
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2022QAGCN: Answering Multi-Relation Questions via Single-Step Implicit Reasoning over Knowledge Graphs
- 2024GraphOTTER: Evolving LLM-based Graph Reasoning for Complex Table Question Answering
- 2026Evolving from Lessons: Skill-Augmented Table Graph Reasoning for Operation-wise Table Question Answering
- 2022Relational Graph Convolutional Neural Networks for Multihop Reasoning: A Comparative Study
- 2025UniRel: Relation-Centric Knowledge Graph Question Answering with RL-Tuned LLM Reasoning
- 2026H2Table: Hierarchical Hypergraph-Enhanced Large Language Models for Complex Table Reasoning
- 2025Efficient Multi-Hop Question Answering over Knowledge Graphs via LLM Planning and Embedding-Guided Search
- 2024Integrating Large Language Models with Graph-based Reasoning for Conversational Question Answering