Elamur Research⌕
☰
home / papers / 2026 / LCR-RAG: Enhancing Logical Consistency in Retrieval-Augmented Generation via Neuro-symbolic Reinforcement Learning

LCR-RAG: Enhancing Logical Consistency in Retrieval-Augmented Generation via Neuro-symbolic Reinforcement Learning

Paper · 2026 · Wenxiang Zheng, Guo Tang, Shixin Jiang et al.

Summary

Retrieval-augmented generation (RAG) helps large language models stay factually grounded. In multi-step reasoning, however, answers can still contradict themselves or skip inference steps. Iterative and self-reflective systems such as IRCoT and Self-RAG try to fix this, but they depend on the model judging its own output. The authors call the resulting gap a "logic deficit" and describe it with a three-level "RAG Reliability Pyramid": factual grounding, semantic coherence, and logical soundness.

LCR-RAG addresses the top level of that pyramid. An external neuro-symbolic verifier combines an NLI-based fact check with symbolic reasoning over RDF triples extracted from the answer. Detected defects are turned into a Logic-Consistency-driven Reward (LCR). A PPO agent uses this reward to choose high-level query rewrites over several iterations. A lightweight router decides beforehand which queries need this costly loop.

Across QA benchmarks and five 8–9B backbones, the method beats strong RAG baselines. Ablations attribute most of the gains to the LCR. This matters because logical consistency becomes an explicit, checkable training objective rather than a hoped-for side effect.

Key points

  • Data and models: evaluated on HotpotQA, ASQA, MuSiQue and TriviaQA. The abstract lists only three of these and omits MuSiQue. Backbones were Qwen3-8B, Llama-3.3-8B, DeepSeek-R1-8B, GLM-4-9B and Ministral-3-8B, with an additional Qwen3 8B vs. 32B scale comparison.
  • Results: compared with standard RAG, the method reports up to 30% relative improvement in answer accuracy and more than 10 points absolute gain on faithfulness-type metrics. It also outperforms baselines including RAG-RL and SeaKR, and stays robust under noisy or conflicting retrieval.
  • Verifier pipeline:
    • an NLI checker for factual support;
    • a T5 semantic parser that converts claims into RDF triples;
    • a Datalog engine that applies a rule base to find contradictions or incomplete inference chains, producing a defect report.
  • Routing module: a DistilBERT-based classifier sends each query to one of three lanes (FAST, STANDARD, DEEP). Its features include entity density from spaCy NER, logical keywords with nested clauses, and IDF-based term sparsity. Predictions below 0.5 confidence fall back to STANDARD.
  • Limitations: the symbolic rules are written by hand and fixed, which may hinder transfer to new domains. Verification and correction also do not go deeper than the query-optimization loop.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Harbin Institute of TechnologyHarbin · China185 works · 29,140 citations🔬Peng Cheng LaboratoryShenzhen · China47 works · 2,699 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Information retrieval & search · 552 works

Cite

Cite as Wenxiang Zheng, Guo Tang, Shixin Jiang, Liangyu Huo, Xiyuan Zhang, Jian Xie and Ming Liu. 2026. LCR-RAG: Enhancing Logical Consistency in Retrieval-Augmented Generation via Neuro-symbolic Reinforcement Learning. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.814

@inproceedings{zheng2026lcr,
  title = {{LCR-RAG: Enhancing Logical Consistency in Retrieval-Augmented Generation via Neuro-symbolic Reinforcement Learning}},
  author = {Wenxiang Zheng, Guo Tang, Shixin Jiang, Liangyu Huo, Xiyuan Zhang, Jian Xie and Ming Liu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.814},
  url = {https://doi.org/10.18653/v1/2026.acl-long.814}
}

References

38 references, 22 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mistral AI. 2025. Introducing mistral 3. https://mistral.ai/news/mistral-3. ↗
  2. AI@Meta. 2024. Llama 3 model card. find
  3. Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2024. Self-RAG: Learning to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations.
  4. Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George van den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, Diego de Las Casas, Aurelia Guy, Jacob Menick, Roman Ring, T. W. Hennigan, Saffron Huang, Lorenzo Maggiore, Chris Jones, Albin… find
  5. Fengxiang Cheng, Haoxuan Li, Fenrong Liu, Robert van Rooij, Kun Zhang, and Zhouchen Lin. 2025. Empowering llms with logical reasoning: A comprehensive survey. Preprint, arXiv:2502.15652. find
  6. DeepSeek-AI. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. Preprint, arXiv:2501.12948.
  7. Shahul Es, Jithin James, Luis Espinosa Anke, and Steven Schockaert. 2024. RAGAs: Automated evaluation of retrieval augmented generation. In Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations, pages 150–158, St. Julians,…
  8. Luyu Gao, Xueguang Ma, Jimmy Lin, and Jamie Callan. 2023a. Precise zero-shot dense retrieval without relevance labels. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1762–1777, Toronto, Canada. Association for Computational L…
  9. Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Qianyu Guo, Meng Wang, and Haofen Wang. 2023b. Retrieval-augmented generation for large language models: A survey. ArXiv, abs/2312.10997.
  10. Bishwamittra Ghosh, Sarah Hasan, Naheed Anjum Arafat, and Arijit Khan. 2025. Logical consistency of large language models in fact-checking. In The Thirteenth International Conference on Learning Representations. find
  11. Claire Glanois, Zhaohui Jiang, Xuening Feng, Paul Weng, Matthieu Zimmer, Dong Li, Wulong Liu, and Jianye Hao. 2022. Neuro-symbolic hierarchical rule induction. In Proceedings of the 39th International Conference on Machine Learning, volume 162 of Proceedings of Machine Learning Research, pages 7583–… find
  12. Team GLM, Aohan Zeng, Bin Xu, Bowen Wang, Chenhui Zhang, Da Yin, Diego Rojas, Guanyu Feng, Hanlin Zhao, Hanyu Lai, Hao Yu, Hongning Wang, Jiadai Sun, Jiajie Zhang, Jiale Cheng, Jiayi Gui, Jie Tang, Jing Zhang, Juanzi Li, and 37 others. 2024. Chatglm: A family of large language models from glm-130b t…
  13. Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. Deberta: Decoding-enhanced bert with disentangled attention. In International Conference on Learning Representations.
  14. Jerry Huang, Siddarth Madala, Risham Sidhu, Cheng Niu, Hao Peng, Julia Hockenmaier, and Tong Zhang. 2025. Rag-rl: Advancing retrieval-augmented generation via rl and curriculum learning. Preprint, arXiv:2503.12759. — [arXiv:2503.12759](https://arxiv.org/abs/2503.12759) find
  15. Pengcheng Jiang, Jiacheng Lin, Lang Cao, Runchu Tian, SeongKu Kang, Zifeng Wang, Jimeng Sun, and Jiawei Han. 2025. Deepretrieval: Hacking real search engines and retrievers with large language models via reinforcement learning. Preprint, arXiv:2503.00223. — [arXiv:2503.00223](https://arxiv.org/abs/2503.00223) find
  16. Mandar Joshi, Eunsol Choi, Daniel Weld, and Luke Zettlemoyer. 2017. TriviaQA: A large scale distantly supervised challenge dataset for reading comprehension. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1601–1611, Vancouver…
  17. Nikhil Kandpal, Haikang Deng, Adam Roberts, Eric Wallace, and Colin Raffel. 2023. Large language models struggle to learn long-tail knowledge. In Proceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org.
  18. Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. 2020. Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6769–6781, Onli…
  19. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. In Proceedings of the 34th Internation…
  20. Yuan Li, Qi Luo, Xiaonan Li, Bufan Li, Qinyuan Cheng, Bo Wang, Yining Zheng, Yuxin Wang, Zhangyue Yin, and Xipeng Qiu. 2025. R3-rag: Learning step-by-step reasoning and retrieval for llms via reinforcement learning. Preprint, arXiv:2505.23794. find
  21. Sheng-Chieh Lin, Luyu Gao, Barlas Oguz, Wenhan Xiong, Jimmy Lin, Wen-tau Yih, and Xilun Chen. 2024. Flame: Factuality-aware alignment for large language models. Advances in Neural Information Processing Systems, 37:115588–115614. find
  22. Yinhong Liu, Zhijiang Guo, Tianya Liang, Ehsan Shareghi, Ivan Vulić, and Nigel Collier. 2025. Aligning with logic: Measuring, evaluating and improving logical preference consistency in large language models. Preprint, arXiv:2410.02205. — [arXiv:2410.02205](https://arxiv.org/abs/2410.02205) find
  23. Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, and Luc De Raedt. 2018. Deepproblog: neural probabilistic logic programming. In Proceedings of the 32nd International Conference on Neural Information Processing Systems, NIPS’18, page 3753–3763, Red Hook, NY, USA. Curran Associ…
  24. Jiayuan Mao, Chuang Gan, Pushmeet Kohli, Joshua B Tenenbaum, and Jiajun Wu. 2019. The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision. In International Conference on Learning Representations.
  25. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. 2022. Train…
  26. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21(1).
  27. Albert Sawczyn, Jakub Binkowski, Denis Janiak, Bogdan Gabrys, and Tomasz Kajdanowicz. 2025. Fact-selfcheck: Fact-level black-box hallucination detection for llms. Preprint, arXiv:2503.17229. — [arXiv:2503.17229](https://arxiv.org/abs/2503.17229) find
  28. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. Preprint, arXiv:1707.06347. find
  29. Luciano Serafini and Artur d’Avila Garcez. 2016. Logic tensor networks: Deep learning and logical reasoning from data and knowledge. Preprint, arXiv:1606.04422. — [arXiv:1606.04422](https://arxiv.org/abs/1606.04422) find
  30. Ivan Stelmakh, Yi Luan, Bhuwan Dhingra, and Ming-Wei Chang. 2022. ASQA: Factoid questions meet long-form answers. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 8273–8288, Abu Dhabi, United Arab Emirates. Association for Computational Linguistics.
  31. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. MuSiQue: Multi-hop questions via single-hop question composition. Transactions of the Association for Computational Linguistics, 10:539–554. find
  32. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2023. Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p…
  33. Xueru Wen, Jie Lou, Xinyu Lu, Ji Yuqiu, Xinyan Guan, Yaojie Lu, Hongyu Lin, Ben He, Xianpei Han, Debing Zhang, and Le Sun. 2025. On-policy self-alignment with fine-grained knowledge feedback for hallucination mitigation. Preprint, arXiv:2406.12221. find
  34. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, • Neuro-Symbolic Verifier (NSV): We utilize DeBERTa-v3-large for Natural Language Inference (NLI) to verify claim entailment. The semantic parser is a T5-large model fine-tuned to translate natural language… — Qwen3 Technical Report
  35. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2369…
  36. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. ReAct: Synergizing reasoning and acting in language models. In International Conference on Learning Representations (ICLR).
  37. Zijun Yao, Weijian Qi, Liangming Pan, Shulin Cao, Linmei Hu, Weichuan Liu, Lei Hou, and Juanzi Li. 2024. Seakr: Self-aware knowledge retrieval for adaptive retrieval augmented generation. Preprint, arXiv:2406.19215.
  38. Wenhao Yu, Dan Iter, Shuohang Wang, Yichong Xu, Mingxuan Ju, Soumya Sanyal, Chenguang Zhu, Michael Zeng, and Meng Jiang. 2023. Generate rather than retrieve: Large language models are strong context generators. In International Conference for Learning Representation (ICLR).

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zheng-2026-lcr-rag-enhancing-logical-consistency/