Elamur Research⌕
☰
home / papers / 2026 / Reason-to-Learn (R2L): Multi-Agent Knowledge Distillation for Lightweight LLMs in Sentiment Analysis

Reason-to-Learn (R2L): Multi-Agent Knowledge Distillation for Lightweight LLMs in Sentiment Analysis

Summary

Large language models are costly to run. This work asks how a compact model can absorb the reasoning skills of several larger ones for sentiment analysis, especially the harder aspect-based variant (ABSA). The authors propose Reason-to-Learn (R2L), a distillation pipeline built around a group of heterogeneous teacher agents. Each teacher identifies sentiment-bearing phrases, processes them, and writes explanatory rationales. A GPT-4o meta-agent then merges these outputs and settles disagreements, producing training data for a small student. Because only generated text passes between models, teachers and student can have different architectures. Experiments center on Vietnamese sentiment tasks. Students with 1–1.5B parameters outperform every individual teacher and the GPT-4o aggregator, with the largest margins on ABSA. Ablations show that both rationales and multi-teacher aggregation contribute beyond plain fine-tuning or single-teacher distillation. The Vietnamese-trained student also transfers zero-shot to Swedish ABSA. The approach points to a cheap route toward deployable specialist models for languages with fewer resources.

Key points

  • The pipeline has four phases: phrase/sentiment detection, processing, rationale generation (run separately by each teacher), and aggregation by a meta-agent (GPT-4o).
  • The teacher agents include Qwen 32B and Llama 70B. The R2L-Student models have only 1–1.5B parameters, yet they beat these teachers on Vietnamese SA and ABSA.
  • According to the ablation study, rationale-based training is essential for complex tasks, and collective aggregation outperforms any single-teacher baseline.
  • In a zero-shot cross-lingual test, the student was evaluated on Swedish ABSA (Svensk ABSAbank-Imm v1.1). It reached Krippendorff's Alpha scores comparable to those of the large teacher agents.
  • The paper was published in the LREC 2026 proceedings (pp. 10301–10312) by authors from DopikAI, Umeå University, University of Orleans, and Lund University/DeepTensor.

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

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/3aygmawej3my

@inproceedings{tu2026reason,
  title = {{Reason-to-Learn (R2L): Multi-Agent Knowledge Distillation for Lightweight LLMs in Sentiment Analysis}},
  author = {Tu, Le-Huy and Nguyen, Quan and NGUYEN, Vincent and Bjorklund, Johanna and Vu, Xuan-Son},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/3aygmawej3my},
  url = {https://doi.org/10.63317/3aygmawej3my}
}

References

56 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Berdicevskis, Aleksandrs, Borin, Lars, Rouces, Jacobo, and Tahmasebi, Nina. 2024. Svensk absabank-imm 1.1. find
  2. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… find
  3. Hyungjoo Chae, Yongho Song, Kai Ong, Taeyoon Kwon, Minjin Kim, Youngjae Yu, Dongha Lee, Dongyeop Kang, and Jinyoung Yeo. 2023. Dialogue chain-of-thought distillation for commonsense-aware conversational agents. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing… find
  4. Xiaoshu Chen, Sihang Zhou, Ke Liang, and Xinwang Liu. 2024. Post-semantic-thinking: A robust strategy to distill reasoning capacity from large language models. CoRR, abs/2404.09170. — [arXiv:2404.09170](https://arxiv.org/abs/2404.09170) find
  5. Zeming Chen, Qiyue Gao, Antoine Bosselut, Ashish Sabharwal, and Kyle Richardson. 2023. DISCO: Distilling counterfactuals with large language models. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5514– 5528, Toronto, Canada.… find
  6. Michael Han Daniel Han and Unsloth team. 2023. Unsloth. find
  7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 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,… find
  8. Yao Fu, Hao Peng, Litu Ou, Ashish Sabharwal, and Tushar Khot. 2023. Specializing smaller language models towards multi-step reasoning. In Proceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org. find our card
  9. Kanishk Gandhi, Ayush Chakravarthy, Anikait Singh, Nathan Lile, and Noah D. Goodman. 2025. Cognitive behaviors that enable self-improving reasoners, or, four habits of highly effective stars. find our card
  10. Alec Go, Richa Bhayani, and Lei Huang. 2009. Twitter sentiment classification using distant supervision. CS224N project report, Stanford, 1(12):2009. find
  11. Yuxian Gu, Li Dong, Furu Wei, and Minlie Huang. 2024. Minillm: Knowledge distillation of large language models. In ICLR. find
  12. Shailja Gupta, Rajesh Ranjan, and Surya Narayan Singh. 2024. Comprehensive study on sentiment analysis: From rule-based to modern llm based system. ArXiv, abs/2409.09989. — [arXiv:2409.09989](https://arxiv.org/abs/2409.09989) find
  13. Jan Herrington and Ron Oliver. 2000. An instructional design framework for authentic learning environments. Educational technology research and development, 48(3):23–48. find
  14. Geoffrey Hinton, Oriol Vinyals, and Jeffrey Dean. 2015. Distilling the knowledge in a neural network. In NIPS Deep Learning and Representation Learning Workshop. find our card
  15. Namgyu Ho, Laura Schmid, and Se-Young Yun. 2023. Large language models are reasoning teachers. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 14852– 14882, Toronto, Canada. Association for Computational Linguistics. find our card
  16. Sebastian Hofstätter, Sophia Althammer, Michael Schröder, Mete Sertkan, and Allan Hanbury. 2020. Improving efficient neural ranking models with cross-architecture knowledge distillation. find
  17. Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. 2019. Parameter-efficient transfer learning for NLP. In Proceedings of the 36th International Conference on Machine Learning, volume 97 of Proceedings of Ma… find
  18. Cheng-Yu Hsieh, Chun-Liang Li, Chih-kuan Yeh, Hootan Nakhost, Yasuhisa Fujii, Alex Ratner, Ranjay Krishna, Chen-Yu Lee, and Tomas Pfister. 2023. Distilling step-by-step! outperforming larger language models with less training data and smaller model sizes. In Findings of the Association for Computati… find our card
  19. Alex Jackson. 2025. Countdown numbers game dataset. Released under the MIT License. find
  20. Savitha Viswanadh Kandala, Pramuka Medaranga, and Ambuj Varshney. 2024. Tinyllm: A framework for training and deploying language models at the edge computers. ArXiv, abs/2412.15304. find
  21. Gyeongman Kim, Doohyuk Jang, and Eunho Yang. 2024. PromptKD: Distilling student-friendly knowledge for generative language models via prompt tuning. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 6266–6282, Miami, Florida, USA. Association for Computational Linguisti… find
  22. Jan Ole Krugmann and Jochen Hartmann. 2024. Sentiment analysis in the age of generative ai. Customer Needs and Solutions, 11:1–19. find
  23. Brian Lester, Rami Al-Rfou, and Noah Constant. 2021. The power of scale for parameter-efficient prompt tuning. In Conference on Empirical Methods in Natural Language Processing. find
  24. Chengxi Li, Feiyu Gao, Jiajun Bu, Lu Xu, Xiang Chen, Yu Gu, Zirui Shao, Qi Zheng, Ningyu Zhang, Yongpan Wang, and Zhi Yu. 2021. Sentiprompt: Sentiment knowledge enhanced prompt-tuning for aspect-based sentiment analysis. CoRR, abs/2109.08306. — [arXiv:2109.08306](https://arxiv.org/abs/2109.08306) find
  25. Liunian Harold Li, Jack Hessel, Youngjae Yu, Xiang Ren, Kai-Wei Chang, and Yejin Choi. 2023. Symbolic chain-of-thought distillation: Small models can also “think” step-by-step. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2… find our card
  26. Yuang Liu, Wei Zhang, and Jun Wang. 2020. Adaptive multi-teacher multi-level knowledge distillation. Neurocomputing, 415:106–113. find
  27. Yufan Liu, Jiajiong Cao, Bing Li, Weiming Hu, Jingting Ding, and Liang Li. 2022. Cross-architecture knowledge distillation. In Computer Vision – ACCV 2022: 16th Asian Conference on Computer Vision, Macao, China, December 4–8, 2022, Proceedings, Part V, page 179–195, Berlin, Heidelberg. Springer-Verl… find
  28. Huyen Nguyen, Hung Nguyen, Quyen Ngo, Luong Vu, Vu Mai, Ngo Xuan Bach, and Cuong Le. 2019. Vlsp shared task: Sentiment analysis. Journal of Computer Science and Cybernetics, 34:295– 310. find
  29. Khai-Nguyen Nguyen, Khai Le-Duc, Bach Phan Tat, Le Duy, Jerry Ngo, Long Vo-Dang, Anh Totti Nguyen, and Hy Truong Son. 2024. Sentiment reasoning for healthcare. In Advancements In Medical Foundation Models: Explainability, Robustness, Security, and Beyond. find
  30. Kiet Van Nguyen, Vu Duc Nguyen, Phu X. V. Nguyen, Tham T. H. Truong, and Ngan Luu-Thuy Nguyen. 2018a. Uit-vsfc: Vietnamese students’ feedback corpus for sentiment analysis. In 2018 10th International Conference on Knowledge and Systems Engineering (KSE), pages 19–24. find
  31. Phu X. V. Nguyen, Tham T. T. Hong, Kiet Van Nguyen, and Ngan Luu-Thuy Nguyen. 2018b. Deep learning versus traditional classifiers on vietnamese students’ feedback corpus. In 2018 5th NAFOSTED Conference on Information and Computer Science (NICS), pages 75–80. find
  32. Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. 2002. Thumbs up? sentiment classification using machine learning techniques. In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002), pages 79–86. Association for Computational Linguistics. find our card
  33. Dolores Perin. 2011. Facilitating student learning through contextualization (assessment of evidence series). find
  34. Cuong Pham, Tuan Hoang, and Thanh-Toan Do. 2023. Collaborative multi-teacher knowledge distillation for learning low bit-width deep neural networks. In 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pages 6424–6432. find
  35. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. find
  36. Colin Raffel, Noam M. Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2019. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21:140:1–140:67. find
  37. Meena Rambocas and Barney G Pacheco. 2018. Online sentiment analysis in marketing research: a review. Journal of Research in Interactive Marketing, 12(2):146–163. find
  38. Margarita Rodríguez-Ibánez, Antonio Casánez-Ventura, Félix Castejón-Mateos, and Pedro-Manuel Cuenca-Jiménez. 2023. A review on sentiment analysis from social media platforms. Expert Systems with Applications, 223:119862. find
  39. Adriana Romero, Nicolas Ballas, Samira Ebrahimi Kahou, Antoine Chassang, Carlo Gatta, and Yoshua Bengio. 2015. Fitnets: Hints for thin deep nets. In 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings. find
  40. Jihyeon Seo, Kyusam Oh, Chanho Min, Yongkeun Yun, and Sungwoo Cho. 2023. Deep collective knowledge distillation. ArXiv, abs/2304.08878. — [arXiv:2304.08878](https://arxiv.org/abs/2304.08878) find
  41. Philipp Sonnleitner, Ulrich Keller, Romain Martin, and Martin Brunner. 2013. Students’ complex problem-solving abilities: Their structure and relations to reasoning ability and educational success. Intelligence, 41(5):289–305. find
  42. Dang Van Thin, Vu Duc Nguye, Kiet Van Nguyen, and Ngan Luu-Thuy Nguyen. 2018. Deep learning for aspect detection on vietnamese reviews. In 2018 5th NAFOSTED Conference on Information and Computer Science (NICS), pages 104–109. find
  43. Somin Wadhwa, Silvio Amir, and Byron C Wallace. 2024. Investigating mysteries of CoT-augmented distillation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 6071–6086, Miami, Florida, USA. Association for Computational Linguistics. find
  44. Fanqi Wan, Xinting Huang, Deng Cai, Xiaojun Quan, Wei Bi, and Shuming Shi. 2024. Knowledge fusion of large language models. ICLR. find
  45. Peifeng Wang, Zhengyang Wang, Zheng Li, Yifan Gao, Bing Yin, and Xiang Ren. 2023. SCOTT: Self-consistent chain-of-thought distillation. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5546–5558, Toronto, Canada. Association fo… find our card
  46. Zhiqiang Wang, Yiran Pang, Yanbin Lin, and Xingquan Zhu. 2024. Adaptable and reliable text classification using large language models. find
  47. Xiaohan Xu, Ming Li, Chongyang Tao, Tao Shen, Reynold Cheng, Jinyang Li, Can Xu, Dacheng Tao, and Tianyi Zhou. 2024. A survey on knowledge distillation of large language models. ArXiv, abs/2402.13116. — [arXiv:2402.13116](https://arxiv.org/abs/2402.13116) find
  48. Chuanguang Yang, Xinqiang Yu, Han Yang, Zhulin An, Chengqing Yu, Libo Huang, and Yongjun Xu. 2025. Multi-Teacher Knowledge Distillation with Reinforcement Learning for Visual Recognition. arXiv e-prints, page arXiv:2502.18510. — [arXiv:2502.18510](https://arxiv.org/abs/2502.18510) find
  49. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, and Quoc V. Le. 2019. XLNet: generalized autoregressive pre-training for language understanding. Curran Associates Inc., Red Hook, NY, USA. find
  50. Shan You, Chang Xu, Chao Xu, and Dacheng Tao. 2017. Learning from multiple teacher networks. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’17, page 1285–1294, New York, NY, USA. Association for Computing Machinery. find
  51. Hailin Zhang, Defang Chen, and Can Wang. 2021. Confidence-aware multi-teacher knowledge distillation. find
  52. ICASSP 2022 - 2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 4498–4502. find
  53. Wenxuan Zhang, Yue Deng, Bing Liu, Sinno Pan, and Lidong Bing. 2024a. Sentiment analysis in the era of large language models: A reality check. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 3881– 3906, Mexico City, Mexico. Association for Computational Linguistics. find our card
  54. Wenxuan Zhang, Xin Li, Yang Deng, Lidong Bing, and Wai Lam. 2023. A survey on aspect-based sentiment analysis: Tasks, methods, and challenges. IEEE Transactions on Knowledge and Data Engineering, 35(11):11019–11038. find
  55. Xin Zhang, Yanzhao Zhang, Dingkun Long, Wen Xie, Ziqi Dai, Jialong Tang, Huan Lin, Baosong Yang, Pengjun Xie, Fei Huang, et al. 2024b. mgte: Generalized long-context text representation and reranking models for multilingual text retrieval. In Proceedings of the 2024 Conference on Empirical Methods i… find our card
  56. Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. A robustly optimized BERT pre-training approach with post-training. In Proceedings of the 20th Chinese National Conference on Computational Linguistics, pages 1218–1227, Huhhot, China. Chinese Information Processing Society of China. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tu-2026-reason-learn-r2l-multi-agent/