{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,21]],"date-time":"2026-03-21T07:14:13Z","timestamp":1774077253479,"version":"3.50.1"},"reference-count":16,"publisher":"Association for Natural Language Processing","issue":"1","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Journal of Natural Language Processing"],"published-print":{"date-parts":[[2026]]},"DOI":"10.5715\/jnlp.33.388","type":"journal-article","created":{"date-parts":[[2026,3,14]],"date-time":"2026-03-14T22:13:03Z","timestamp":1773526383000},"page":"388-394","source":"Crossref","is-referenced-by-count":0,"title":["Massive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment Quality"],"prefix":"10.5715","volume":"33","author":[{"given":"Yuto","family":"Harada","sequence":"first","affiliation":[{"name":"NII LLMC"},{"name":"The University of Tokyo"}]},{"given":"Yusuke","family":"Yamauchi","sequence":"additional","affiliation":[{"name":"NII LLMC"},{"name":"The University of Tokyo"}]}],"member":"3685","reference":[{"key":"1","unstructured":"Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter, C., Hesse, C., Chen, M., Sigler, E., Litwin, M., Gray, S., Chess, B., Clark, J., Berner, C., McCandlish, S., Radford, A., Sutskever, I., and Amodei, D. (2020). \u201cLanguage Models are Few-Shot Learners.\u201d In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M., and Lin, H. (Eds.), <i>Advances in Neural Information Processing Systems<\/i>, Vol. 33, pp. 1877\u20131901. Curran Associates, Inc."},{"key":"2","unstructured":"Chen, L., Li, S., Yan, J., Wang, H., Gunaratna, K., Yadav, V., Tang, Z., Srinivasan, V., Zhou, T., Huang, H., et al. (2023). \u201cAlpagasus: Training a Better Alpaca with Fewer Data.\u201d <i>arXiv preprint arXiv:2307.08701<\/i>."},{"key":"3","unstructured":"DeepSeek-AI (2025). \u201cDeepSeek-V3.2: Pushing the Frontier of Open Large Language Models.\u201d <i>arXiv preprint arXiv:2512.02556<\/i>."},{"key":"4","doi-asserted-by":"crossref","unstructured":"Dong, G., Yuan, H., Lu, K., Li, C., Xue, M., Liu, D., Wang, W., Yuan, Z., Zhou, C., and Zhou, J. (2024). \u201cHow Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition.\u201d In Ku, L.-W., Martins, A., and Srikumar, V. (Eds.), <i>Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)<\/i>, pp. 177\u2013198, Bangkok, Thailand. Association for Computational Linguistics.","DOI":"10.18653\/v1\/2024.acl-long.12"},{"key":"5","doi-asserted-by":"crossref","unstructured":"Guan, M. Y., Joglekar, M., Wallace, E., Jain, S., Barak, B., Helyar, A., Dias, R., Vallone, A., Ren, H., Wei, J., Chung, H. W., Toyer, S., Heidecke, J., Beutel, A., and Glaese, A. (2025). \u201cDeliberative Alignment: Reasoning Enables Safer Language Models.\u201d <i>arXiv preprint arXiv:2412.16339<\/i>.","DOI":"10.70777\/si.v2i3.15159"},{"key":"6","doi-asserted-by":"crossref","unstructured":"Harada, Y., Yamauchi, Y., Oda, Y., Oseki, Y., Miyao, Y., and Takagi, Y. (2025). \u201cMassive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment Quality.\u201d In Christodoulopoulos, C., Chakraborty, T., Rose, C., and Peng, V. (Eds.), <i>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing<\/i>, pp. 22360\u201322381, Suzhou, China. Association for Computational Linguistics.","DOI":"10.18653\/v1\/2025.emnlp-main.1138"},{"key":"7","unstructured":"Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., and Chen, W. (2022). \u201cLoRA: Low-Rank Adaptation of Large Language Models.\u201d In <i>International Conference on Learning Representations<\/i>."},{"key":"8","unstructured":"Jin, X. and Ren, X. (2024). \u201cDemystifying Language Model Forgetting with Low-Rank Example Associations.\u201d <i>arXiv preprint arXiv:2406.14026<\/i>."},{"key":"9","doi-asserted-by":"crossref","unstructured":"Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. (2022). \u201cTraining Language Models to Follow Instructions with Human Feedback.\u201d <i>Advances in Neural Information Processing Systems<\/i>, 35, pp. 27730\u201327744.","DOI":"10.52202\/068431-2011"},{"key":"10","unstructured":"Ruis, L., Mozes, M., Bae, J., Kamalakara, S. R., Talupuru, D., Locatelli, A., Kirk, R., Rockt\u00e4schel, T., Grefenstette, E., and Bartolo, M. (2024). \u201cProcedural Knowledge in Pretraining Drives Reasoning in Large Language Models.\u201d <i>arXiv preprint arXiv:2411.12580<\/i>."},{"key":"11","unstructured":"Wu, C.-C., Tam, Z. R., Lin, C.-Y., yi Lee, H., and Chen, Y.-N. (2025). \u201cClear Minds Think Alike: What Makes LLM Fine-tuning Robust? A Study of Token Perplexity.\u201d <i>arXiv preprint arXiv:2501.14315<\/i>."},{"key":"12","unstructured":"Yang, A., Li, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., Gao, C., Huang, C., Lv, C., Zheng, C., Liu, D., Zhou, F., Huang, F., Hu, F., Ge, H., Wei, H., Lin, H., Tang, J., Yang, J., Tu, J., Zhang, J., Yang, J., Yang, J., Zhou, J., Zhou, J., Lin, J., Dang, K., Bao, K., Yang, K., Yu, L., Deng, L., Li, M., Xue, M., Li, M., Zhang, P., Wang, P., Zhu, Q., Men, R., Gao, R., Liu, S., Luo, S., Li, T., Tang, T., Yin, W., Ren, X., Wang, X., Zhang, X., Ren, X., Fan, Y., Su, Y., Zhang, Y., Zhang, Y., Wan, Y., Liu, Y., Wang, Z., Cui, Z., Zhang, Z., Zhou, Z., and Qiu, Z. (2025). \u201cQwen3 Technical Report.\u201d <i>arXiv preprint arXiv:2505.09388<\/i>."},{"key":"13","unstructured":"Zhang, B., Liu, Z., Cherry, C., and Firat, O. (2024). \u201cWhen Scaling Meets Llm Finetuning: The Effect of Data, Model and Finetuning Method.\u201d <i>arXiv preprint arXiv:2402.17193<\/i>."},{"key":"14","unstructured":"Zhao, H., Andriushchenko, M., Croce, F., and Flammarion, N. (2024). \u201cLong is More for Alignment: A Simple but Tough-to-beat Baseline for Instruction Fine-tuning.\u201d <i>arXiv preprint arXiv:2402.04833<\/i>."},{"key":"15","doi-asserted-by":"crossref","unstructured":"Zhou, C., Liu, P., Xu, P., Iyer, S., Sun, J., Mao, Y., Ma, X., Efrat, A., Yu, P., Yu, L., et al. (2024). \u201cLima: Less is more for alignment.\u201d <i>Advances in Neural Information Processing Systems<\/i>, 36, pp. 55006\u201355021.","DOI":"10.52202\/075280-2400"},{"key":"16","unstructured":"Zhuo, T. Y., Zebaze, A., Suppattarachai, N., von Werra, L., de Vries, H., Liu, Q., and Muennighoff, N. (2024). \u201cAstraios: Parameter-Efficient Instruction Tuning Code Large Language Models.\u201d <i>arXiv preprint arXiv:2401.00788<\/i>."}],"container-title":["Journal of Natural Language Processing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.jstage.jst.go.jp\/article\/jnlp\/33\/1\/33_388\/_pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,21]],"date-time":"2026-03-21T03:53:50Z","timestamp":1774065230000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.jstage.jst.go.jp\/article\/jnlp\/33\/1\/33_388\/_article\/-char\/ja\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":16,"journal-issue":{"issue":"1","published-print":{"date-parts":[[2026]]}},"URL":"https:\/\/doi.org\/10.5715\/jnlp.33.388","relation":{},"ISSN":["1340-7619","2185-8314"],"issn-type":[{"value":"1340-7619","type":"print"},{"value":"2185-8314","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026]]}}}