Abstract
Text classification, where each input text is assigned to a single category, is a fundamental task in natural language processing. In this paper, we propose a novel framework that combines BERT-based contextual embeddings with a reinforcement learning technique known as group relative policy optimization (GRPO). To enrich input representations, we employ prompt-based text generation using GPT models. It helps create diverse paraphrased variants that are fused with the original embeddings. It is then used to train a classifier using the GRPO, which improves stability and reduces reliance on conventional supervised objectives. We conduct experiments on the baseline BERT model and perform an ablation study of our proposed pipeline using a financial fraud classification dataset. It shows that the integration of generative augmentation and policy-based learning leads to improved classification performance and training efficacy.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Airani, P., Pipada, N., Shah, P.: Classification of complaints text data by ensembling large language models. In: Proceedings of the 17th International Conference on Agents and Artificial Intelligence - Volume 3: ICAART, pp. 679–686. INSTICC, SciTePress (2025). https://doi.org/10.5220/0013173900003890
Balkus, S.V., Yan, D.: Improving short text classification with augmented data using gpt-3 (2022). https://doi.org/10.48550/arXiv.2205.10981, https://arxiv.org/abs/2205.10981
Chai, D., Wu, W., Han, Q., Wu, F., Li, J.: Description based text classification with reinforcement learning. In: Proceedings of the 37th International Conference on Machine Learning, pp. 1360–1369. PMLR (2020). http://proceedings.mlr.press/v119/chai20a/chai20a.pdf
Dai, H., et al.: Auggpt: Leveraging chatgpt for text data augmentation (2023). https://arxiv.org/abs/2302.13007
Devlin, J., Chang, M.W., Lee, K., Toutanova, K.: Bert: Pre-training of deep bidirectional transformers for language understanding (2019). https://doi.org/10.48550/arXiv.1810.04805, https://arxiv.org/abs/1810.04805
Don‘aciolo, H., Barri’e, A., South, L.E., Forbes, F.: Dynamic learning rate for deep reinforcement learning: a bandit approach (2024). https://arxiv.org/abs/2410.12598
Guo, D., et al.: Deepseek-r1: Incentivizing reasoning capability in LLMs via reinforcement learning (2025). https://arxiv.org/abs/2501.12948
Kumar, V., et al.: Data augmentation using pre-trained transformer models (2020). https://doi.org/10.48550/arXiv.2003.02245, https://arxiv.org/abs/2003.02245
Kuznetsov, S., colleagues: Using large language models for data augmentation in text classification models. Int. J. Comput. 24(1) (2025). https://doi.org/10.47839/ijc.24.1.3886
Lin, Z., Lin, M., Xie, Y., Ji, R.: Cppo: accelerating the training of group relative policy optimization-based reasoning models (2025). https://doi.org/10.48550/arXiv.2503.22342, https://arxiv.org/abs/2503.22342, submitted Mar 28 (2025)
Piduguralla, M., Bhatt, J.S.: An analytical CNN: Use of wavelets for learning image structures in cross-domain generalization. In: 2024 National Conference on Communications (NCC), pp. 1–6. Chennai, India (2024). https://doi.org/10.1109/NCC60321.2024.10485918
Refai, D., Abo-Soud, S., Abdel-Rahman, M.: Data augmentation using transformers and similarity measures for improving Arabic text classification (2022). https://doi.org/10.48550/arXiv.2212.13939, https://arxiv.org/abs/2212.13939
Sane, S.: Hybrid group relative policy optimization: a multi-sample approach to enhancing policy optimization (2025). https://arxiv.org/abs/2502.01652, posted Jan 30, 2025
Shao, Z., et al.: Deepseekmath: Pushing the limits of mathematical reasoning in open language models (2024). https://doi.org/10.48550/arXiv.2402.03300, https://arxiv.org/abs/2402.03300
Xue, Z., et al.: Dancegrpo: Unleashing grpo on visual generation (2025). https://arxiv.org/abs/2505.07818, released on arXiv May 12, 2025
Ye, Z., et al.: Zero-shot text classification via reinforced self-training. In: Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (2020). https://aclanthology.org/2020.acl-main.272.pdf
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2026 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Vaghasiya, V., Shimpi, D., Shah, P., Bhatt, J.S. (2026). Text Classification Using LLM and Group Relative Policy Optimization (GRPO). In: Mitra, S., Saha, S., Panigrahi, B.K., Sarkar, S., Chaudhury, S. (eds) Pattern Recognition and Machine Intelligence. PReMI 2025. Lecture Notes in Computer Science, vol 16357. Springer, Cham. https://doi.org/10.1007/978-3-032-18477-1_65
Download citation
DOI: https://doi.org/10.1007/978-3-032-18477-1_65
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-032-18476-4
Online ISBN: 978-3-032-18477-1
eBook Packages: Computer ScienceComputer Science (R0)Springer Nature Proceedings Computer Science