Knowledge-Guided Reinforcement Learning for Multi-Step Question Answering with Small Language Models
Keywords:
small language models; knowledge-guided reinforcement learning; multi-step question answering; retrieval augmentation; policy optimization; system governance; robustness; fairness; sustainability; deploymentAbstract
Multi-step question answering requires systems to retrieve evidence, decompose problems, coordinate intermediate reasoning, and produce final answers under uncertainty. While large language models have demonstrated strong performance on such tasks, their computational, financial, and environmental costs limit deployment in many public-sector, educational, and edge-oriented settings. Small language models offer a compelling alternative, but they often struggle with long-horizon reasoning, fragile retrieval use, and sparse reward signals. This paper examines knowledge-guided reinforcement learning as a system-level strategy for enabling small language models to perform multi-step question answering. Rather than treating reinforcement learning as a purely optimization problem, the analysis emphasizes architecture, knowledge infrastructure, governance, robustness, fairness, and sustainability. The paper discusses how retrieval systems, symbolic verifiers, tool interfaces, and process-level reward models can guide policy learning while reducing sample complexity. It also considers structural trade-offs involving latency, energy use, auditability, and deployment autonomy. The central argument is that knowledge-guided reinforcement learning can make small language models more capable and governable in multi-step question answering, but only if training objectives are coupled with retrieval quality controls, transparent evaluation, and institutional safeguards. The paper concludes by outlining research directions for hybrid neuro-symbolic systems, federated knowledge infrastructure, and policy frameworks that support accountable deployment.
References
1. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
2. Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., ... & Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
3. Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (pp. 4171–4186). Association for Computational Linguistics.
4. Raffel, C., Shazeer, N., Roberts, A., Lee, K., Narang, S., Matena, M., ... & Liu, P. J. (2020). Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140), 1–67.
5. Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., ... & Zhou, D. (2022). Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35, 24824–24837.
6. Kojima, T., Gu, S. S., Reid, M., Matsuo, Y., & Iwasawa, Y. (2022). Large language models are zero-shot reasoners. Advances in Neural Information Processing Systems, 35, 22199–22213.
7. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing reasoning and acting in language models. In International Conference on Learning Representations.
8. Schick, T., Dwivedi-Yu, J., Dessì, R., Raileanu, R., Lomeli, M., Zettlemoyer, L., ... & Scialom, T. (2023). Toolformer: Language models can teach themselves to use tools. Advances in Neural Information Processing Systems, 36.
9. Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., ... & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459–9474.
10. Borgeaud, S., Mensch, A., Hoffmann, J., Cai, T., Rutherford, E., Millican, K., ... & Sifre, L. (2022). Improving language models by retrieving from trillions of tokens. In International Conference on Machine Learning (pp. 2206–2240). PMLR.
11. Christiano, P. F., Leike, J., Brown, T. B., Martic, M., Legg, S., & Amodei, D. (2017). Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems, 30.
12. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., ... & Lowe, R. (2022). Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35, 27730–27744.
13. Bai, Y., Jones, A., Ndousse, K., Askell, A., Chen, A., DasSarma, N., ... & Kaplan, J. (2022). Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862.
14. Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017). Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347.
15. Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., ... & Hassabis, D. (2015). Human-level control through deep reinforcement learning. Nature, 518(7540), 529–533.
16. Li, Q. (2026, May). TrajLite PPO: Scalable Reasoning Path Filtering and Alignment for Small Parameter Large Language Models. In 2026 2nd International Conference on Artificial Intelligence and Digital Ethics (ICAIDE) (pp. 29-32). IEEE.
17. Zelikman, E., Wu, Y., Mu, J., & Goodman, N. D. (2022). Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35, 15476–15488.
18. Trinh, T. H., Wu, Y., Le, Q. V., He, H., & Luong, T. (2024). Solving olympiad geometry without human demonstrations. Nature, 625(7995), 476–482.
19. Wang, X., Wei, J., Schuurmans, D., Le, Q., Chi, E., Narang, S., ... & Zhou, D. (2023). Self-consistency improves chain of thought reasoning in language models. In International Conference on Learning Representations.
20. Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., ... & Liang, P. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Data Intelligence and AI Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
