Elamur Research⌕
☰
home / papers / 2026 / ActiveLLM: Large Language Model-Based Active Learning for Textual Few-Shot Scenarios

ActiveLLM: Large Language Model-Based Active Learning for Textual Few-Shot Scenarios

Paper · 2026 · Markus Bayer, Justin Lutz and Christian Reuter

Summary

Active learning (AL) tries to reduce annotation cost by choosing the examples that are most useful to label. Common strategies, such as uncertainty sampling, need a classifier already trained on some labeled data. This is the "cold-start" problem. These strategies also retrain models like BERT repeatedly during annotation, which is slow, and they gain little when the query model differs from the final (successor) model.

The authors propose ActiveLLM. It gives the selection step to a large language model (GPT-4, o1, Llama 3, Mistral Large), which receives unlabeled instances in a prompt and decides which ones annotators should label. No seed labels or training during annotation are needed, so querying is independent of the successor model.

In few-shot experiments, BERT classifiers trained on ActiveLLM-selected data beat those built with conventional AL. The selections also improved the few-shot methods ADAPET, PERFECT and SetFit. An iterative mode extends the approach to larger budgets and can warm-start other AL strategies.

This matters because the LLM is used only during labeling. The resulting small classifier can run locally, cheaply and without external APIs.

Key points

  • The baseline AL strategies discussed include uncertainty-based Least Confidence (LC) and Prediction Entropy (PE), plus diversity-based Embedding KMeans (EKM). ActiveLLM targets their cold-start and model-mismatch weaknesses.
  • The prompt was tuned for GPT-4 on a lesser-known dataset to limit benchmark leakage, then reused unchanged across other LLMs and datasets. Experiments mostly used the LLMs' chat interfaces.
  • The authors report that long prompts degrade selection: LLMs lose track of the task or ignore instructions, such as the provided labels in the iterated mode. They link this to findings by Hsieh et al. (2024) on long-context reasoning.
  • Exact scores appear in the part of the text that was not provided, so they are not included here.
  • Code is available at github.com/PEASEC/ActiveLLM. The paper appears in TACL vol. 14 (2026), pp. 1–22, under a CC-BY 4.0 license.

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

🎓Technische Universität DarmstadtDarmstadt · Germany109 works · 35,920 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Markus Bayer, Justin Lutz and Christian Reuter. 2026. ActiveLLM: Large Language Model-Based Active Learning for Textual Few-Shot Scenarios. Transactions of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1162/tacl.a.63

@article{bayer2026activellm,
  title = {{ActiveLLM: Large Language Model-Based Active Learning for Textual Few-Shot Scenarios}},
  author = {Markus Bayer, Justin Lutz and Christian Reuter},
  journal = {Transactions of the Association for Computational Linguistics},
  year = {2026},
  doi = {10.1162/tacl.a.63},
  url = {https://doi.org/10.1162/tacl.a.63}
}

References

54 references, 21 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rohan Anil, Sebastian Borgeaud, Yonghui Wu, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M. Dai, Anja Hauth, Katie Millican, David Silver, Slav Petrov, Melvin Johnson, Ioannis Antonoglou, Julian Schrittwieser, Amelia Glaese, Jilin Chen, Emily Pitler, Timothy P. Lillicrap,… — Gemini: A Family of Highly Capable Multimodal Models
  2. Jason Baldridge and Miles Osborne. 2004. Active learning and the total cost of annotation. In Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing, EMNLP 2004, A meeting of SIGDAT, a Special Interest Group of the ACL, held in conjunction with ACL 2004, 25–26 July 20… find
  3. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. BERT: Pre-training of deep bidirectional transformers for language understanding. CoRR, abs/1810.04805.
  4. Bosheng Ding, Chengwei Qin, Ruochen Zhao, Tianze Luo, Xinze Li, Guizhen Chen, Wenhan Xia, Junjie Hu, Anh Tuan Luu, and Shafiq Joty. 2024. Data augmentation using LLMs: Data perspectives, learning paradigms and challenges. CoRR, abs/2403.02990. https://doi.org/10 .18653/v1/2024.findings-acl.97 find
  5. Markus Bayer, Tobias Frey, and Christian Reuter. 2022. Multi-level fine-tuning, data augmentation, and few-shot learning for specialized cyber threat intelligence. CoRR, abs/2207.11076. https://doi.org/10.1016/j.cose.2023.103430 — [arXiv:2207.11076](https://arxiv.org/abs/2207.11076) ↗
  6. Liat Ein-Dor, Alon Halfon, Ariel Gera, Eyal Shnarch, Lena Dankin, Leshem Choshen, Marina Danilevsky, Ranit Aharonov, Yoav Katz, and Noam Slonim. 2020. Active learning for BERT: An empirical study. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020,… find
  7. Mitchell Bosley, Musashi Jacobs-Harukawa, Hauke Licht, and Alexander Hoyle. 2023. Do we still need BERT in the age of GPT? Comparing the benefits of domain-adaptation and in-context-learning approaches to using LLMs for political science research. In 2023 Annual Meeting of the Midwest Political Scie… find
  8. 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,… — Language Models are Few-Shot Learners
  9. Tianyu Gao, Adam Fisch, and Danqi Chen. 2020. Making pre-trained language models better few-shot learners. CoRR, abs/2012.15723.
  10. Daniel Grießhaber, Johannes Maucher, and Ngoc Thang Vu. 2020. Fine-tuning BERT for low-resource natural language understanding via active learning. CoRR, abs/2012.02462. https://doi.org/10.18653/v1/2020 .coling-main.100 ↗
  11. Neil Houlsby, Ferenc Huszar, Zoubin Ghahramani, and Máté Lengyel. 2011. Bayesian active learning for classification and preference learning. CoRR, abs/1112.5745. find
  12. Liangyu Chen, Yutong Bai, Siyu Huang, Yongyi Lu, Bihan Wen, Alan L. Yuille, and Zongwei Zhou. 2023. Making your first choice: To address cold start problem in medical active learning. In Medical Imaging with Deep Learning, MIDL 2023, 10–12 July 2023, Nashville, TN, USA, volume 227 of Proceedings of… find
  13. Cheng-Ping Hsieh, Simeng Sun, Samuel Kriman, Shantanu Acharya, Dima Rekesh, Fei Jia, Yang Zhang, and Boris Ginsburg. 2024. RULER: What’s the real context size of your long-context language models? CoRR, abs/2404.06654.
  14. Pieter Floris Jacobs, Gideon Maillette de Buy Wenniger, Marco A. Wiering, and Lambert Schomaker. 2021. Active learning for reducing labeling effort in text classification tasks. CoRR, abs/2109.04847. — [arXiv:2109.04847](https://arxiv.org/abs/2109.04847) find
  15. David A. Cohn, Les E. Atlas, and Richard E. Ladner. 1994. Improving generalization with active learning. Machine Learning, 15(2):201–221. https://doi.org/10.1023/A:1022673506211 ↗
  16. Albert Q. Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de Las Casas, Emma Bou Hanna, Florian Bressand, Gianna Lengyel, Guillaume Bour, Guillaume Lample, Lélio Renard Lavaud, Lucile Saulnier, Marie-Anne Lachaux, Pierre Stock… find
  17. OpenAI. 2023. GPT-4 technical report. CoRR, abs/2303.08774.
  18. OpenAI. 2024. Learning to reason with LLMS. find
  19. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence embeddings using Siamese BERT-networks. CoRR, abs/1908 .10084. https://doi.org/10.18653/v1/D19-1410 ↗
  20. Sebastian Ruder, Matthew E. Peters, Swabha Swayamdipta, and Thomas Wolf. 2019. Transfer learning in natural language processing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapo…
  21. Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts. 2023. DSPy: Compiling declarative language model calls into self-improving pipelines. C…
  22. David Samuel. 2024. BERTs are generative in-context learners. CoRR, abs/2406.04823. https://doi.org/10.52202/079017 -0084 — [doi:10.52202/079017](https://doi.org/10.52202/079017) ↗
  23. Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023. Lost in the middle: How language models use long contexts. CoRR, abs/2307.03172.
  24. Timo Schick and Hinrich Schütze. 2020a. Exploiting cloze questions for few-shot text classification and natural language inference. CoRR, abs/2001.07676. https://doi.org/10.18653/v1/2021.eacl-main.20 ↗
  25. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. RoBERTa: A robustly optimized BERT pretraining approach. CoRR, abs/1907.11692.
  26. Timo Schick and Hinrich Schütze. 2020b. It’s not just size that matters: Small language models are also few-shot learners. CoRR, abs/ 2009.07118. https://doi.org/10.18653/v1/2021.naacl-main.185 ↗
  27. Rabeeh Karimi Mahabadi, Luke Zettlemoyer, James Henderson, Marzieh Saeidi, Lambert Mathias, Veselin Stoyanov, and Majid Yazdani. 2022. PERFECT: Prompt-free and efficient few-shot learning with language models. CoRR, abs/2204.01172. https://doi.org/10.18653/v1/2022.acl-long.254 ↗
  28. Christopher Schröder, Lydia Müller, Andreas Niekler, and Martin Potthast. 2021. Small-text: Active learning for text classification in Python. CoRR, abs/2107.10314. find
  29. Marius Mosbach, Maksym Andriushchenko, and Dietrich Klakow. 2020. On the stability of fine-tuning BERT: Misconceptions, explanations, and strong baselines. CoRR, abs/2006.04884. find
  30. Christopher Schröder, Andreas Niekler, and Martin Potthast. 2022. Revisiting uncertainty-based query strategies for active learning with transformers. In Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22–27, 2022, pages 2194–2203. Association for Computati… find
  31. Minh Van Nguyen, Nghia Trung Ngo, Bonan Min, and Thien Huu Nguyen. 2022. FAMIE: A fast active learning framework for multilingual information extraction. CoRR, abs/2202.08316. https://doi.org/10.18653/v1/2022.naacl-demo.14 ↗
  32. Seungmin Seo, Donghyun Kim, Youbin Ahn, and Kyong-Ho Lee. 2022. Active learning on pre-trained language model with task-independent triplet loss. In Thirty-Sixth AAAI Conference on Artificial Intelligence, AAAI 2022, Thirty-Fourth Conference on Innovative Applications of Artificial Intelligence, IAA…
  33. Derek Tam, Rakesh R. Menon, Mohit Bansal, Shashank Srivastava, and Colin Raffel. 2021. Improving and simplifying pattern exploiting training. CoRR, abs/2103.11955. https://doi.org/10.18653/v1/2021.emnlp-main .407 ↗
  34. Simon Tong and Daphne Koller. 2001. Support vector machine active learning with applications to text classification. Journal of Machine Learning Research, 2:45–66. find
  35. Burr Settles. 2012. Active Learning. Synthesis Lectures on Artificial Intelligence and Machine Learning. Morgan & Claypool Publishers. https://doi.org/10.1007/978-3-031 -01560-1 ↗
  36. Marcos V. Treviso, Tianchu Ji, Ji-Ung Lee, Betty van Aken, Qingqing Cao, Manuel R. Ciosici, Michael Hassid, Kenneth Heafield, Sara Hooker, Pedro Henrique Martins, André F. T. Martins, Peter A. Milder, Colin Raffel, Edwin Simpson, Noam Slonim, Niranjan Balasubramanian, Leon Derczynski, and Roy Schwa… — Efficient Methods for Natural Language Processing: A Survey
  37. Burr Settles and Mark Craven. 2008. An analysis of active learning strategies for sequence labeling tasks. In 2008 Conference on Empirical Methods in Natural Language Processing, EMNLP 2008, Proceedings of the Conference, 25–27 October 2008, Honolulu, Hawaii, USA, A meeting of SIGDAT, a Special Inte… find
  38. Akim Tsvigun, Artem Shelmanov, Gleb Kuzmin, Leonid Sanochkin, Daniil Larionov, Gleb Gusev, Manvel Avetisian, and Leonid Zhukov. 2022. Towards Computationally Feasible Deep Active Learning. CoRR, abs/2205.03598. https://doi.org/10.18653/v1/2022 .findings-naacl.90 ↗
  39. Artem Shelmanov, Dmitry Puzyrev, Lyubov Kupriyanova, Denis Belyakov, Daniil Larionov, Nikita Khromov, Olga Kozlova, Ekaterina Artemova, Dmitry V. Dylov, and Alexander Panchenko. 2021. Active learning for sequence tagging with deep pre-trained models and Bayesian uncertainty estimates. CoRR, abs/2101… find
  40. Lewis Tunstall, Nils Reimers, Unso Eun Seo Jo, Luke Bates, Daniel Korat, Moshe Wasserblat, and Oren Pereg. 2022. Efficient few-shot learning without prompts. CoRR, abs/2209.11055. — [arXiv:2209.11055](https://arxiv.org/abs/2209.11055) find
  41. Oleh Shliazhko, Alena Fenogenova, Maria Tikhonova, Vladislav Mikhailov, Anastasia Kozlova, and Tatiana Shavrina. 2022. mGPT: Few-shot learners go multilingual. CoRR, abs/2204.07580.
  42. Thomas Walshe, Sae Young Moon, Chunyang Xiao, Yawwani Gunawardana, and Fran Silavong. 2025. Automatic labelling with open-source LLMs using dynamic label schema integration. CoRR, abs/2501.12332. — [arXiv:2501.12332](https://arxiv.org/abs/2501.12332) find
  43. Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Y. Ng, and Christopher Potts. 2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, EMNLP 20…
  44. Fali Wang, Zhiwei Zhang, Xianren Zhang, Zongyu Wu, Tzuhao Mo, Qiuhao Lu, Wanjing Wang, Rui Li, Junjie Xu, Xianfeng Tang, Qi He, Yao Ma, Ming Huang, and Suhang Wang. 2024a. A comprehensive survey of small language models in the era of large language models: Techniques, enhancements, applications, col… find
  45. Michelle Yuan, Hsuan-Tien Lin, and Jordan L. Boyd-Graber. 2020. Cold-start active learning through self-supervised language modeling. CoRR, abs/2010.09535. https://doi.org/10.18653/v1/2020.emnlp-main.637 ↗
  46. Ningyu Zhang, Luoqiu Li, Xiang Chen, Shumin Deng, Zhen Bi, Chuanqi Tan, Fei Huang, and Huajun Chen. 2021. Differentiable prompt makes pre-trained language models better few-shot learners. CoRR, abs/2108.13161. — [arXiv:2108.13161](https://arxiv.org/abs/2108.13161) find
  47. Xinru Wang, Hannah Kim, Sajjadur Rahman, Kushan Mitra, and Zhengjie Miao. 2024b. Human-LLM collaborative annotation through effective verification of LLM labels. In Proceedings of the CHI Conference on Human Factors in Computing Systems, CHI 2024, Honolulu, HI, USA, May 11–16, 2024, pages 303:1–303:… — [doi:10.1145/3613904.3641960](https://doi.org/10.1145/3613904.3641960) find
  48. Xiang Zhang, Junbo Jake Zhao, and Yann LeCun. 2015. Character-level convolutional networks for text classification. CoRR, abs/1509.01626.
  49. Yiming Zhang, Shi Feng, and Chenhao Tan. 2022a. Active example selection for in-context learning. CoRR, abs/2211.04486. https://doi.org/10.18653/v1/2022 .emnlp-main.622 ↗
  50. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed H. Chi, Quoc Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. CoRR, abs/2201.11903.
  51. Zhisong Zhang, Emma Strubell, and Eduard H. Hovy. 2022b. A survey of active learning for natural language processing. CoRR, abs/2210.10109. https://doi.org/10 .18653/v1/2022.emnlp-main.414 find
  52. Mengyang Xu, Qihuang Zhong, and Juhua Liu. 2024. LLM-as-an-Augmentor: Improving the data augmentation for aspect-based sentiment analysis with large language models. ICIC2024. https://doi.org/10.65286/icic.v20i2.10911 ↗
  53. Haoqi Zheng, Qihuang Zhong, Liang Ding, Zhiliang Tian, Xin Niu, Dongsheng Li, and Dacheng Tao. 2023. Self-evolution learning for mixup: Enhance data augmentation on few-shot text classification tasks. CoRR, abs/2305.13547. https://doi.org/10 .18653/v1/2023.emnlp-main.555 find
  54. Zhongjie Yu and Sebastian Raschka. 2020. Looking back to lower-level information in few-shot learning. CoRR, abs/2005.13638. https://doi.org/10.3390/info11070345 — [doi:10.3390/info11070345](https://doi.org/10.3390/info11070345) ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bayer-2026-activellm-large-language-model-based/