Beyond Retrieval: Bidirectional Neuro-Symbolic Validation for Reliable Graph-RAG

Leonardo Arazo de Oliveira Araújo; Didier Vega-Oliveros

doi:10.5753/kdmile.2026.32086

Leonardo Arazo de Oliveira Araújo UNIFESP
Didier Vega-Oliveros UNIFESP

DOI: https://doi.org/10.5753/kdmile.2026.32086

Resumo

Retrieval-Augmented Generation (RAG) and its graph-based extensions improve factual grounding in LLMs but do not guarantee logical consistency when responses must satisfy explicit domain constraints. We propose a bidirectional neuro-symbolic validation framework for Graph-RAG that combines First-Order Logic reasoning over a domain knowledge graph with an Annotated Datalog engine built on PyReason, propagating confidence bounds across inferred facts. A bidirectional mechanism supports symbolic enrichment, post-hoc validation with corrective rewriting, and direct symbolic execution. In a Brazilian public university advisory benchmark, the framework achieves strict accuracy of 84.2% and weighted accuracy of 87.7% (with partial-credit weighting), an 87% reduction in factual errors relative to LLM-only; a backend-sensitivity study with gemma4:12b preserves the baseline ordering with no answer labeled incorrect (0/57).

Palavras-chave: graph-rag, knowledge graphs, neuro-symbolic reasoning, retrieval-augmented generation

Referências

Aditya, D., Mukherji, K., Balasubramanian, S., Chaudhary, A., and Shakarian, P. Pyreason: Software for open world temporal logic. In AAAI Spring Symposium on Challenges Requiring the Combination of Machine Learning and Knowledge Engineering, 2023.

Brown, A., Roman, M., and Devereux, B. A systematic literature review of retrieval-augmented generation: Techniques, metrics, and challenges. arXiv preprint arXiv:2508.06401, 2025.

Cheng, M., Luo, Y., Ouyang, J., Liu, Q., Liu, H., Li, L., Yu, S., Zhang, B., Cao, J., Ma, J., Wang, D., and Chen, E. A survey on knowledge-oriented retrieval-augmented generation. arXiv preprint arXiv:2503.10677, 2025.

Cohen, J. A coefficient of agreement for nominal scales. Educational and Psychological Measurement 20 (1): 37–46, 1960.

Garcez, A. S. d. and Lamb, L. C. Neurosymbolic ai: The third wave. Artificial Intelligence Review vol. 56, pp. 12387–12406, 2023.

Han, H., Ma, L., Wang, Y., Shomer, H., Lei, Y., Qi, Z., Guo, K., Hua, Z., Long, B., Liu, H., Aggarwal, C. C., and Tang, J. Rag vs. graphrag: A systematic evaluation and key insights. arXiv preprint arXiv:2502.11371, 2025.

Linders, J. and Tomczak, J. M. Knowledge graph-extended retrieval augmented generation for question answering. Applied Intelligence vol. 55, pp. 1102, 2025.

Saha, A., Agarwal, P., et al. A zero-shot neuro-symbolic approach for complex knowledge graph question answering. arXiv preprint arXiv:2601.04568, 2026.

Saxena, Y. and Gaur, M. Neurosymbolic retrievers for retrieval-augmented generation. IEEE Intelligent Systems, 2026.

Wu, J., Zhu, J., Qi, Y., Chen, J., Xu, M., Menolascina, F., and Grau, V. Medical graph rag: Towards safe medical large language model via graph retrieval-augmented generation. arXiv preprint arXiv:2408.04187, 2024.

Yu, D., Yang, B., Liu, D., Wang, H., and Pan, S. A survey on neural-symbolic learning systems. Neural Networks vol. 166, pp. 105–126, 2023.

Zhang, Q., Chen, S., Bei, Y., Yuan, Z., Zhou, H., Hong, Z., Chen, H., Xiao, Y., Zhou, C., Dong, J., Chang, Y., and Huang, X. A survey of graph retrieval-augmented generation for customized large language models. arXiv preprint arXiv:2501.13958, 2025.

Zhu, X., Xie, Y., Liu, Y., Li, Y., and Hu, W. Knowledge graph-guided retrieval augmented generation. In Proceedings of the 2025 Annual Conference of the Nations of the Americas Chapter of the ACL (NAACL 2025), 2025.