TQAgent: Enhancing Table-Based Question Answering with Knowledge Graphs and Tree-Structured Reasoning
Source abstract
Abstract
Table-based question answering (TableQA) has emerged as an important task in natural language processing, yet existing models face challenges in handling complex reasoning and mitigating hallucinations, especially when dealing with diverse table structures. We introduce TQAgent, a framework designed to enhance table-based reasoning by incorporating knowledge graphs and tree-structured reasoning paths. TQAgent reduces hallucinations and improves model reliability by grounding reasoning in external knowledge and dynamically sampling high-confidence paths. Additionally, it employs knowledge distillation techniques for lightweight deployment. Experimental results on the TabFact, WikiTQ, and FeTaQA datasets show significant performance improvements, with accuracy increases of up to 4% over baseline models. TQAgent’s dynamic operation planning and knowledge graph integration enable effective multi-step reasoning and better handling of diverse table data. Furthermore, the framework achieves state-of-the-art results, surpassing traditional large-scale models in both reasoning accuracy and computational efficiency. These findings open new avenues for future research in table-based question answering and model deployment optimization.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Topic Modeling · Data Quality and Management · Semantic Web and Ontologies
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
State Grid Hebei Information and Telecommunication Branch, Shijiazhuang 050051, China
School of Computer Science (National Pilot Software Engineering School), Beijing University of Posts and Telecommunications, Beijing 100876, China
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing 100876, China
Funding
State Grid Hebei Electric Power Company
References (40)
Show source-provided references
1. Fang, X., Xu, W., Tan, F.A., Zhang, J., Hu, Z., Qi, Y., Nickleach, S., Socolinsky, D., Sengamedu, S., and Faloutsos, C. (2024). Large Language Models (LLMs) on Tabular Data: Prediction, Generation, and Understanding—A Survey. arXiv.
2. Ritze, D., Lehmberg, O., and Bizer, C. (2015, January 13–15). Matching html tables to dbpedia. Proceedings of the 5th International Conference on Web Intelligence, Mining and Semantics, Larnaca, Cyprus.10.1145/2797115.2797118 ↗
3. Zhang, S., and Balog, K. (2017, January 7–11). Entitables: Smart assistance for entity-focused tables. Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval, Tokyo, Japan.10.1145/3077136.3080796 ↗
4. Cheng, Z., Dong, H., Wang, Z., Jia, R., Guo, J., Gao, Y., Han, S., Lou, J.G., and Zhang, D. (2021). Hitab: A hierarchical table dataset for question answering and natural language generation. arXiv.10.18653/v1/2022.acl-long.78 ↗
5. FeTaQA: Free-form table question answering10.1162/tacl_a_00446 ↗
6. Liu, Q., Chen, B., Guo, J., Ziyadi, M., Lin, Z., Chen, W., and Lou, J.G. (2021). TAPEX: Table pre-training via learning a neural SQL executor. arXiv.
7. OpticalBERT and OpticalTable-SQA: Text-and table-based language models for the optical-materials domain10.1021/acs.jcim.2c01259 ↗
8. Turl: Table understanding through representation learning10.1145/3542700.3542709 ↗
9. Survey of hallucination in natural language generation10.1145/3571730 ↗
10. Lenat, D., and Marcus, G. (2023). Getting from generative ai to trustworthy ai: What llms might learn from cyc. arXiv.
11. Mallen, A., Asai, A., Zhong, V., Das, R., Khashabi, D., and Hajishirzi, H. (2022). When not to trust language models: Investigating effectiveness of parametric and non-parametric memories. arXiv.10.18653/v1/2023.acl-long.546 ↗
12. Zhang, M., Press, O., Merrill, W., Liu, A., and Smith, N.A. (2023). How language model hallucinations can snowball. arXiv.
13. Mialon, G., Dessì, R., Lomeli, M., Nalmpantis, C., Pasunuru, R., Raileanu, R., Rozière, B., Schick, T., Dwivedi-Yu, J., and Celikyilmaz, A. (2023). Augmented language models: A survey. arXiv.
14. Mruthyunjaya, V., Pezeshkpour, P., Hruschka, E., and Bhutani, N. (2023). Rethinking language models as symbolic knowledge graphs. arXiv.
15. Strich, J. (2025, January 19–20). Adapt LLM for Multi-turn Reasoning QA using Tidy Data. Proceedings of the Joint Workshop of the 9th Financial Technology and Natural Language Processing (FinNLP), the 6th Financial Narrative Processing (FNP), and the 1st Workshop on Large Language Models for Finance and Legal (LLMFinLegal), Abu Dhabi, United Arab Emirates.
16. Martynova, A., Tishin, V., and Semenova, N. (2025, January 19). Learn Together: Joint Multitask Finetuning of Pretrained KG-enhanced LLM for Downstream Tasks. Proceedings of the Workshop on Generative AI and Knowledge Graphs (GenAIK), Abu Dhabi, United Arab Emirates.
17. Natural Language Interfaces for Tabular Data Querying and Visualization: A Survey10.1109/TKDE.2024.3400824 ↗
18. Zhao, Y., Chen, L., Cohan, A., and Zhao, C. (2024, January 11–16). TaPERA: Enhancing faithfulness and interpretability in long-form table QA by content planning and execution-based reasoning. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Bangkok, Thailand.10.18653/v1/2024.acl-long.692 ↗
19. Ye, Y., Hui, B., Yang, M., Li, B., Huang, F., and Li, Y. (2023, January 23–27). Large language models are versatile decomposers: Decomposing evidence and questions for table-based reasoning. Proceedings of the 46th International ACM SIGIR Conference on Research and development In Information Retrieval, Taipei, China.10.1145/3539618.3591708 ↗
20. He, X., Zhou, M., Xu, X., Ma, X., Ding, R., Du, L., Gao, Y., Jia, R., Chen, X., and Han, S. (2024, January 20–27). Text2analysis: A benchmark of table question answering with advanced data analysis and unclear queries. Proceedings of the AAAI Conference on Artificial Intelligence, Vancouver, BC, Canada.10.1609/aaai.v38i16.29779 ↗
21. Table-gpt: Table fine-tuned gpt for diverse table tasks
22. Sui, Y., Zhou, M., Zhou, M., Han, S., and Zhang, D. (2024, January 4–8). Table meets llm: Can large language models understand structured table data? A benchmark and empirical study. Proceedings of the 17th ACM International Conference on Web Search and Data Mining, Virtual.10.1145/3616855.3635752 ↗
23. Rajkumar, N., Li, R., and Bahdanau, D. (2022). Evaluating the Text-to-SQL Capabilities of Large Language Models. arXiv.
24. Combining Small Language Models and Large Language Models for Zero-Shot NL2SQL10.14778/3681954.3681960 ↗
25. Zhang, T., Yue, X., Li, Y., and Sun, H. (2024). TableLlama: Towards Open Large Generalist Models for Tables. arXiv.10.18653/v1/2024.naacl-long.335 ↗
26. Chen, Y., Yuan, Y., Zhang, Z., Zheng, Y., Liu, J., Ni, F., Hao, J., Mao, H., and Zhang, F. (2025). SheetAgent: Towards A Generalist Agent for Spreadsheet Reasoning and Manipulation via Large Language Models. arXiv.
27. Knowledge Graphs, Large Language Models, and Hallucinations: An NLP Perspective10.1016/j.websem.2024.100844 ↗
28. Guan, X., Liu, Y., Lin, H., Lu, Y., He, B., Han, X., and Sun, L. (2024, January 20–27). Mitigating large language model hallucinations via autonomous knowledge graph-based retrofitting. Proceedings of the AAAI Conference on Artificial Intelligence, Vancouver, BC, Canada.10.1609/aaai.v38i16.29770 ↗
29. Ledger, G., and Mancinni, R. (Authorea Prepr., 2024). Detecting llm hallucinations using monte carlo simulations on token probabilities, Authorea Prepr., preprint.10.36227/techrxiv.171822396.61518693/v1 ↗
30. Haloscope: Harnessing unlabeled llm generations for hallucination detection
31. Fairburn, S., and Ainsworth, J. (Authorea Prepr., 2024). Mitigate large language model hallucinations with probabilistic inference in graph neural networks, Authorea Prepr., preprint.10.22541/au.171987466.61187752/v1 ↗
32. Taveekitworachai, P., Abdullah, F., and Thawonmas, R. (2024, January 12–16). Null-shot prompting: Rethinking prompting large language models with hallucination. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, Miami, FL, USA.10.18653/v1/2024.emnlp-main.740 ↗
33. The Impact of LLM Hallucinations on Motor Skill Learning: A Case Study in Badminton10.1109/ACCESS.2024.3444783 ↗
34. GPTs and Hallucination: Why do large language models hallucinate?10.1145/3688007 ↗
35. Li, S., Ji, H., and Han, J. (2021, January 6–11). Document-Level Event Argument Extraction by Conditional Generation. Proceedings of the North American Chapter of the Association for Computational Linguistics, Online.10.18653/v1/2021.naacl-main.69 ↗
36. Chen, W., Wang, H., Chen, J., Zhang, Y., Wang, H., Li, S., Zhou, X., and Wang, W.Y. (2019). Tabfact: A large-scale dataset for table-based fact verification. arXiv.
37. Srivastava, P., Ganu, T., and Guha, S. (2022). Towards Zero-Shot and Few-Shot Table Question Answering using GPT-3. arXiv.
38. TableGPT: A novel table understanding method based on table recognition and large language model collaborative enhancement10.1007/s10489-024-05937-6 ↗
39. Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., and Zhou, D. (2023). Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. arXiv.
40. Hernandez Farias, D.I., Hope, T., and Li, M. (2024). OpenT2T: An Open-Source Toolkit for Table-to-Text Generation. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, Miami, FL, USA, 12–16 November 2024, Association for Computational Linguistics.