Elamur Research⌕
☰
home / papers / 2025 / Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons

Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons

Paper · 2025 · Yongqi Leng and Deyi Xiong

Summary

Large language models handle many tasks after instruction tuning, but it is still unclear how they share and separate task knowledge internally. Leng and Xiong apply neuron-level analysis, previously used to find language-specific neurons in multilingual models, to multi-task learning. Each FFN neuron gets a score based on how much the task loss would change if its output were set to zero. This change is estimated with a first-order Taylor (gradient attribution) approximation, and the top k% of neurons are labelled task-specific. Deactivating these neurons, or fine-tuning only them, shows they are strongly linked to their tasks.

On classification and generation tasks such as SST-2, AmazonFood and XSum, the authors find two signals of cross-task generalization:

  • how much the neuron sets of the training and test tasks overlap;
  • how similar task-specific neuron parameters are at certain layers.

From this they derive NCFT, a continual fine-tuning scheme that updates only the current task's neurons, and W-NCFT, which weights parameters by task similarity. Both strongly reduce catastrophic forgetting without adding extra parameters.

Key points

  • Neuron definition and scoring:
    • A neuron is one column of an FFN weight matrix (W1 or W2).
    • Its relevance is the absolute loss change when its output is zeroed, approximated by output × gradient.
    • Task-specific neurons are the top-k% scorers; k was set empirically, not tuned.
  • Generalization analysis:
    • Only the training task's neurons are fine-tuned, at several proportions.
    • In-domain and out-of-domain results are measured with Accuracy (classification) and Rouge-L (generation).
    • Results are related to neuron overlap and parameter similarity using PCCs and SROCC.
    • Higher overlap generally means better transfer, but not deterministically.
  • Specialization trade-off: multi-task training can hurt single-task performance because parameters interfere across tasks. The authors see the same interference as the source of catastrophic forgetting.
  • Continual learning results:
    • On the second benchmark (Large Number of Tasks), W-NCFT averages 72.0, ahead of O-LoRA (71.6), EPI (42.6), SeqLoRA (47.7) and SeqFT (34.6).
    • On the first benchmark, O-LoRA leads (76.1 vs. 73.3).
    • Per-Task FT upper bounds are 77.2 and 84.5.
  • Limitations:
    • A gap to Per-Task FT remains.
    • W-NCFT uses static rather than dynamic weighting.
    • The analysis covers only classification and generation tasks.

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

🎓Tianjin UniversityTianjin · China41 works · 1,128 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 Yongqi Leng and Deyi Xiong. 2025. Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons. Proceedings of the 31st International Conference on Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{leng2025understanding,
  title = {{Towards Understanding Multi-Task Learning (Generalization) of LLMs via Detecting and Exploring Task-Specific Neurons}},
  author = {Yongqi Leng and Deyi Xiong},
  booktitle = {Proceedings of the 31st International Conference on Computational Linguistics},
  year = {2025},
  url = {https://aclanthology.org/2025.coling-main.200/}
}

References

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

  1. Ekin Akyürek, Dale Schuurmans, Jacob Andreas, Tengyu Ma, and Denny Zhou. 2023. What learning algorithm is in-context learning? investigations with linear models. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net.
  2. Rahaf Aljundi, Francesca Babiloni, Mohamed Elhoseiny, Marcus Rohrbach, and Tinne Tuytelaars. 2018. Memory aware synapses: Learning what (not) to forget. In Computer Vision - ECCV 2018 - 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part III, volume 11207 of Lecture No… find
  3. 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
  4. Anwoy Chatterjee, Eshaan Tanwar, Subhabrata Dutta, and Tanmoy Chakraborty. 2024. Language models can exploit cross-task in-context learning for data-scarce novel tasks. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangk… find
  5. Ruizhe Chen, Tianxiang Hu, Yang Feng, and Zuozhu Liu. 2024a. Learnable privacy neurons localization in language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, ACL 2024 - Short Papers, Bangkok, Thailand, August 11-16, 2024, pages 256–264. Associati… find
  6. Yuheng Chen, Pengfei Cao, Yubo Chen, Kang Liu, and Jun Zhao. 2024b. Journey to the center of the knowledge neurons: Discoveries of language-independent knowledge neurons and degenerate knowledge neurons. In Thirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Conference… find
  7. Damai Dai, Li Dong, Yaru Hao, Zhifang Sui, Baobao Chang, and Furu Wei. 2022. Knowledge neurons in pretrained transformers. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2022, Dublin, Ireland, May 22-27, 2022, pages 8493– 8502.… find
  8. Ondrej Dusek, Jekaterina Novikova, and Verena Rieser. 2020. Evaluating the state-of-the-art of end-to-end natural language generation: The E2E NLG challenge. Comput. Speech Lang., 59:123–156. find
  9. Roger B. Grosse, Juhan Bae, Cem Anil, Nelson Elhage, Alex Tamkin, Amirhossein Tajdini, Benoit Steiner, Dustin Li, Esin Durmus, Ethan Perez, Evan Hubinger, Kamile Lukosiute, Karina Nguyen, Nicholas Joseph, Sam McCandlish, Jared Kaplan, and Samuel R. Bowman. 2023. Studying large language model general… find
  10. Jian Gu, Chunyang Chen, and Aldeida Aleti. 2023. Neuron patching: Neuron-level model editing on code generation and llms. CoRR, abs/2312.05356. — [arXiv:2312.05356](https://arxiv.org/abs/2312.05356) find
  11. Karl Moritz Hermann, Tomás Kociský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. Teaching machines to read and comprehend. In Advances in Neural Information Processing Systems 28: Annual Conference on Neural Information Processing Systems 2015, December 7-…
  12. Yuncheng Huang, Qianyu He, Yipei Xu, Jiaqing Liang, and Yanghua Xiao. 2024. Laying the foundation first? investigating the generalization from atomic skills to complex reasoning tasks. CoRR, abs/2403.09479. — [arXiv:2403.09479](https://arxiv.org/abs/2403.09479) find
  13. Dieuwke Hupkes, Mario Giulianelli, Verna Dankers, Mikel Artetxe, Yanai Elazar, Tiago Pimentel, Christos Christodoulopoulos, Karim Lasri, Naomi Saphra, Arabella Sinclair, Dennis Ulmer, Florian Schottmann, Khuyagbaatar Batsuren, Kaiser Sun, Koustuv Sinha, Leila Khalatbari, Maria Ryskina, Rita Frieske,… find
  14. Zixuan Ke and Bing Liu. 2022. Continual learning of natural language processing tasks: A survey. CoRR, abs/2211.12701. find
  15. Phillip Keung, Yichao Lu, György Szarvas, and Noah A. Smith. 2020. The multilingual amazon reviews corpus. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online, November 16-20, 2020, pages 4563–4568. Association for Computational Linguistics.
  16. Bill Yuchen Lin, Wangchunshu Zhou, Ming Shen, Pei Zhou, Chandra Bhagavatula, Yejin Choi, and Xiang Ren. 2020. Commongen: A constrained text generation challenge for generative commonsense reasoning. In Findings of the Association for Computational Linguistics: EMNLP 2020, Online Event, 16-20 Novembe… find
  17. Haoyan Luo and Lucia Specia. 2024. From understanding to utilization: A survey on explainability for large language models. CoRR, abs/2401.12874. — [arXiv:2401.12874](https://arxiv.org/abs/2401.12874) find
  18. Shashi Narayan, Shay B. Cohen, and Mirella Lapata. 2018. Don’t give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Brussels, Belgium, October 31 -Novembe…
  19. OpenAI. 2023. GPT-4 technical report. abs/2303.08774.
  20. CoRR, Alec Radford, Rafal Józefowicz, and Ilya Sutskever. 2017. Learning to generate reviews and discovering sentiment. CoRR, abs/1704.01444.
  21. Anastasia Razdaibiedina, Yuning Mao, Rui Hou, Madian Khabsa, Mike Lewis, and Amjad Almahairi. 2023. Progressive prompts: Continual learning for language models. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net. find
  22. Irina Rish. 2021. Continual learning with deep architectures. find
  23. Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal V. Nayak, Debajyoti Datta, Jonatha… — Multitask Prompted Training Enables Zero-Shot Task Generalization
  24. Teven Le Scao, Angela Fan, Christopher Akiki, Ellie Pavlick, Suzana Ilic, Daniel Hesslow, Roman Castagné, Alexandra Sasha Luccioni, François Yvon, Matthias Gallé, Jonathan Tow, Alexander M. Rush, Stella Biderman, Albert Webson, Pawan Sasanka Ammanamanchi, Thomas Wang, Benoît Sagot, Niklas Muennighof… find
  25. Thomas Scialom, Tuhin Chakrabarty, and Smaranda Muresan. 2022. Fine-tuned language models are continual learners. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, EMNLP 2022, Abu Dhabi, United Arab Emirates, December 7-11, 2022, pages 6107–6122. Association… find
  26. Tianhao Shen, Renren Jin, Yufei Huang, Chuang Liu, Weilong Dong, Zishan Guo, Xinwei Wu, Yan Liu, and Deyi Xiong. 2023. Large language model alignment: A survey. CoRR, abs/2309.15025.
  27. Dan Shi, Renren Jin, Tianhao Shen, Weilong Dong, Xinwei Wu, and Deyi Xiong. 2024. IRCAN: mitigating knowledge conflicts in LLM generation via identifying and reweighting context-aware neurons. CoRR, abs/2406.18406. — [arXiv:2406.18406](https://arxiv.org/abs/2406.18406) find
  28. Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2014. Deep inside convolutional networks: Visualising image classification models and saliency maps. In 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Workshop Track Proceedings.
  29. 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…
  30. Xin Su, Shangqi Guo, Tian Tan, and Feng Chen. 2020. Generative memory for lifelong learning. IEEE Trans. Neural Networks Learn. Syst., 31(6):1884– 1898. find
  31. Tianyi Tang, Wenyang Luo, Haoyang Huang, Dongdong Zhang, Xiaolei Wang, Xin Zhao, Furu Wei, and Ji-Rong Wen. 2024. Language-specific neurons: The key to multilingual capabilities in large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volu…
  32. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton-Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller,… — Llama 2: Open Foundation and Fine-Tuned Chat Models
  33. Johannes von Oswald, Eyvind Niklasson, Ettore Randazzo, João Sacramento, Alexander Mordvintsev, Andrey Zhmoginov, and Max Vladymyrov. 2023. Transformers learn in-context by gradient descent. In International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, volume 20…
  34. Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R. Bowman. 2019. GLUE: A multi-task benchmark and analysis platform for natural language understanding. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenRevie…
  35. Liyuan Wang, Xingxing Zhang, Hang Su, and Jun Zhu. 2024. A comprehensive survey of continual learning: Theory, method and application. IEEE Trans. Pattern Anal. Mach. Intell., 46(8):5362–5383. find
  36. Xiao Wang, Tianze Chen, Qiming Ge, Han Xia, Rong Bao, Rui Zheng, Qi Zhang, Tao Gui, and Xuanjing Huang. 2023a. Orthogonal subspace learning for language model continual learning. In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023, pages 10658– 10… find
  37. Xiaozhi Wang, Kaiyue Wen, Zhengyan Zhang, Lei Hou, Zhiyuan Liu, and Juanzi Li. 2022. Finding skill neurons in pre-trained transformer-based language models. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, EMNLP 2022, Abu Dhabi, United Arab Emirates, Decembe…
  38. Zhicheng Wang, Yufang Liu, Tao Ji, Xiaoling Wang, Yuanbin Wu, Congcong Jiang, Ye Chao, Zhencong Han, Ling Wang, Xu Shao, and Wenqiu Zeng. 2023b. Rehearsal-free continual language learning via efficient parameter isolation. In Proceedings of the 61st Annual Meeting of the Association for Computationa… find
  39. Jason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V. Le. 2022. Finetuned language models are zero-shot learners. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview…
  40. Johannes Welbl, Nelson F. Liu, and Matt Gardner. 2017. Crowdsourcing multiple choice science questions. In Proceedings of the 3rd Workshop on Noisy User-generated Text, NUT@EMNLP 2017, Copenhagen, Denmark, September 7, 2017, pages 94–106. Association for Computational Linguistics.
  41. Adina Williams, Nikita Nangia, and Samuel R. Bowman. 2018. A broad-coverage challenge corpus for sentence understanding through inference. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2018…
  42. Xinwei Wu, Weilong Dong, Shaoyang Xu, and Deyi Xiong. 2024. Mitigating privacy seesaw in large language models: Augmented privacy neuron editing via activation patching. In Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024,… find
  43. Xinwei Wu, Junzhuo Li, Minghui Xu, Weilong Dong, Shuangzhi Wu, Chao Bian, and Deyi Xiong. 2023a. DEPN: detecting and editing privacy neurons in pre-trained language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10… find
  44. Yang Wu, Yanyan Zhao, Zhongyang Li, Bing Qin, and Kai Xiong. 2023b. Improving cross-task generalization with step-by-step instructions. CoRR, abs/2305.04429. — [arXiv:2305.04429](https://arxiv.org/abs/2305.04429) find
  45. Wanying Xie, Yang Feng, Shuhao Gu, and Dong Yu. 2021. Importance-based neuron allocation for multilingual neural machine translation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing,… find
  46. Yiran Zhao, Wenxuan Zhang, Guizhen Chen, Kenji Kawaguchi, and Lidong Bing. 2024. How do large language models handle multilingualism? CoRR, abs/2402.18815.
  47. Chunting Zhou, Pengfei Liu, Puxin Xu, Srinivasan Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu, Susan Zhang, Gargi Ghosh, Mike Lewis, Luke Zettlemoyer, and Omer Levy. 2023. LIMA: less is more for alignment. In Advances in Neural Information Processing Systems 36: Annual Confere… find
  48. Shaolin Zhu, Leiyu Pan, Bo Li, and Deyi Xiong. 2024. Landermt: Dectecting and routing language-aware neurons for selectively finetuning llms to machine translation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok,… find
  49. Wenhan Xiong, Jiawei Wu, Hong Wang, Vivek Kulkarni, Mo Yu, Shiyu Chang, Xiaoxiao Guo, and William Yang Wang. 2019. TWEETQA: A social media focused question answering dataset. In Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy, July 28- A… find
  50. Haoyun Xu, Runzhe Zhan, Derek F. Wong, and Lidia S. Chao. 2024. Let’s focus on neuron: Neuron-level supervised fine-tuning for large language model. CoRR, abs/2403.11621.
  51. Haoran Yang, Yumeng Zhang, Jiaqi Xu, Hongyuan Lu, Pheng-Ann Heng, and Wai Lam. 2024. Unveiling the generalization power of fine-tuned large language models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologi… find
  52. Yuan Zhang, Jason Baldridge, and Luheng He. 2019. PAWS: paraphrase adversaries from word scrambling. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2019, Minneapolis, MN, USA, June 2-7, 2019…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/leng-2025-understanding-multi-task-learning-generalization/