L-TQA: A Modular Multi-Agent Architecture for Context-Constrained Tabular Question Answering

• Helen B. Alves UECE
• Diego D. Fernandes UECE
• Gustavo A. G. S. Dias UECE
• João A. M. B. Cavalcanti UECE
• Antônio Í. M. Pinto UECE
• Lara S. Silva UECE
• Rafael S. Santos UECE
• Thalyson G. N. Silva UECE / IFCE
• Cleilton L. Rocha Atlantic Institute
• Leonardo F. da Costa UECE
• Gustavo A. L. Campos UECE
• Ana Luiza B. P. Barros UECE

DOI: https://doi.org/10.5753/semish.2026.23560

Resumo

Question Answering over Tabular Data requires accurate numerical reasoning and robust schema interpretation, both of which remain challenging for restricted-parameter language models. This paper presents Laura Tabular Question Answer (L-TQA), a multi-agent architecture designed to mitigate context limitations by decoupling semantic interpretation, schema filtering, and executable code generation. The proposed system improves query synthesis through targeted contextual constraints, combining a response-type classifier to enforce output formats with a dynamic column selector to reduce schema noise. Evaluated on the SemEval-2025 Task 8 DataBench benchmark, L-TQA establishes a new state-of-the-art for models with up to 9 billion parameters, achieving 77.01% accuracy on the test split. In addition, the proposed modular design generalizes effectively to larger proprietary models, highlighting its structural robustness and practical value for privacy-preserving, data-driven applications.

Referências

Bangar, A., Kumar, A., Mishra, S., and Modi, A. (2025). Exploration lab IITK at SemEval-2025 task 8: Multi-LLM agent QA over tabular data. In Rosenthal, S., Rosá, A., Ghosh, D., and Zampieri, M., editors, Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025), pages 2165–2169, Vienna, Austria. Association for Computational Linguistics.

Chen, W. (2023). Large language models are few(1)-shot table reasoners. In Vlachos, A. and Augenstein, I., editors, Findings of the Association for Computational Linguistics: EACL 2023, pages 1120–1130, Dubrovnik, Croatia. Association for Computational Linguistics.

Grijalba, J. O., Ureñ-López, L. A., Martínez-Cámara, E., and Camacho-Collados, J. (2025). Semeval-2025 task 8: Question answering over tabular data. In Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025), pages 2512–2522.

Grijalba, J. O., Ureña-López, L. A., Cámara, E. M., and Camacho-Collados, J. (2024). Question answering over tabular data with databench: A large-scale empirical evaluation of llms. In Proceedings of LREC-COLING 2024, Turin, Italy.

Jiang, Y., Wei, F., Bao, E., Li, Y., Ding, B., Yang, Y., and Xiao, X. (2026). Accurate table question answering with accessible llms.

Jin, R., Wang, X., Wang, D., Zheng, H., Qi, Y., Yang, S., and Zhang, M. (2025). TALON: A multi-agent framework for long-table exploration and question answering. In Christodoulopoulos, C., Chakraborty, T., Rose, C., and Peng, V., editors, Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 27397–27413, Suzhou, China. Association for Computational Linguistics.

Luo, J., Zhang, W., Yuan, Y., Zhao, Y., Yang, J., Gu, Y., Wu, B., Chen, B., Qiao, Z., Long, Q., Tu, R., Luo, X., Ju, W., Xiao, Z., Wang, Y., Xiao, M., Liu, C., Yuan, J., Zhang, S., Jin, Y., Zhang, F., Wu, X., Zhao, H., Tao, D., Yu, P. S., and Zhang, M. (2025). Large language model agent: A survey on methodology, applications and challenges.

Nan, L., Hsieh, C., Mao, Z., Lin, X. V., Verma, N., Zhang, R., Kryściński, W., Schoelkopf, H., Kong, R., Tang, X., Mutuma, M., Rosand, B., Trindade, I., Bandaru, R., Cunningham, J., Xiong, C., and Radev, D. (2022). FeTaQA: Free-form table question answering. Transactions of the Association for Computational Linguistics, 10:35–49.

Raghav, R., Vemali, A. P., Aswal, D., Ramesh, R., and Bhupal, A. (2025). Scottyposeidon at semeval-2025 task 8: Llm-driven code generation for zero-shot question answering on tabular data. In Proceedings of the 19th International Workshop on Semantic Evaluation (SemEval-2025), pages 2197–2204.

Reback, J., McKinney, W., Van Den Bossche, J., Augspurger, T., Cloud, P., Klein, A., Hawkins, S., Roeschke, M., Tratner, J., She, C., et al. (2020). pandas-dev/pandas: Pandas 1.0. 5. Zenodo.

Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., Rodriguez, A., Joulin, A., Grave, E., and Lample, G. (2023). Llama: Open and efficient foundation language models.

Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., Zhao, W. X., Wei, Z., and Wen, J. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6).

Wang, Z., Moriyama, S., Wang, W.-Y., Gangopadhyay, B., and Takamatsu, S. (2025). Talk structurally, act hierarchically: A collaborative framework for llm multi-agent systems.

Zhang, J., Fan, Y., Cai, K., Sun, X., and Wang, K. (2025). Osc: Cognitive orchestration through dynamic knowledge alignment in multi-agent llm collaboration.

Zhang, W., Zeng, L., Xiao, Y., Li, Y., Cui, C., Zhao, Y., Hu, R., Liu, Y., Zhou, Y., and An, B. (2026). Agentorchestra: Orchestrating multi-agent intelligence with the tool-environment-agent(tea) protocol.

Zhu, J.-P., Cai, P., Xu, K., Li, L., Sun, Y., Zhou, S., Su, H., Tang, L., and Liu, Q. (2024). Autotqa: Towards autonomous tabular question answering through multi-agent large language models. Proc. VLDB Endow., 17:3920–3933.