{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,20]],"date-time":"2026-10-20T18:39:14Z","timestamp":1792521554109,"version":"4.5.0"},"reference-count":45,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2027,1,1]],"date-time":"2027-01-01T00:00:00Z","timestamp":1798761600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/tdm\/userlicense\/1.0\/"},{"start":{"date-parts":[[2027,1,1]],"date-time":"2027-01-01T00:00:00Z","timestamp":1798761600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/legal\/tdmrep-license"},{"start":{"date-parts":[[2026,9,9]],"date-time":"2026-09-09T00:00:00Z","timestamp":1788912000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100006196","name":"Oulun Yliopisto","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100006196","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002341","name":"Research Council of Finland","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100002341","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["elsevier.com","sciencedirect.com"],"crossmark-restriction":true},"short-container-title":["Neural Networks"],"published-print":{"date-parts":[[2027,1]]},"DOI":"10.1016\/j.neunet.2026.109617","type":"journal-article","created":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T15:10:37Z","timestamp":1789139437000},"page":"109617","update-policy":"https:\/\/doi.org\/10.1016\/elsevier_cm_policy","source":"Crossref","is-referenced-by-count":0,"special_numbering":"PC","title":["MoEP: Compact and efficient sparsity with modular expert paths"],"prefix":"10.1016","volume":"205","author":[{"given":"Joonas","family":"Tapaninaho","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4422-8723","authenticated-orcid":false,"given":"Mourad","family":"Oussalah","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"10.1016\/j.neunet.2026.109617_bib0001","series-title":"Advances in neural information processing systems","article-title":"Mixture-of-recursions: Learning dynamic recursive depths for adaptive token-level computation","author":"Bae","year":"2025"},{"key":"10.1016\/j.neunet.2026.109617_bib0002","series-title":"Proceedings of the 40th international conference on machine learning","first-page":"2397","article-title":"Pythia: A suite for analyzing large language models across training and scaling","volume":"vol. 202","author":"Biderman","year":"2023"},{"key":"10.1016\/j.neunet.2026.109617_bib0003","unstructured":"Charpentier, L., Choshen, L., & Cotterell, R., et al. (2025). BabyLM turns 3: Call for papers for the 2025 babyLM workshop. arXiv: 2502.10645."},{"key":"10.1016\/j.neunet.2026.109617_bib0004","series-title":"Proceedings of the 2nd babyLM challenge workshop","article-title":"GPT or BERT: Why not both?","author":"Charpentier","year":"2024"},{"issue":"9","key":"10.1016\/j.neunet.2026.109617_bib0005","doi-asserted-by":"crossref","first-page":"1229","DOI":"10.1016\/S0893-6080(99)00043-X","article-title":"Improved learning algorithms for mixture of experts in multiclass classification","volume":"12","author":"Chen","year":"1999","journal-title":"Neural Networks"},{"key":"10.1016\/j.neunet.2026.109617_bib0006","first-page":"1","article-title":"PaLM: Scaling language modeling with pathways","volume":"24","author":"Chowdhery","year":"2023","journal-title":"Journal of Machine Learning Research"},{"key":"10.1016\/j.neunet.2026.109617_bib0007","unstructured":"CSC, LUMI Supercomputer. Lumi supercomputer. https:\/\/lumi-supercomputer.eu\/. Accessed: 2026-06-16."},{"key":"10.1016\/j.neunet.2026.109617_bib0008","unstructured":"CSC, Puhti Supercomputer. Puhti supercomputer. https:\/\/docs.csc.fi\/computing\/systems-puhti\/. Accessed: 2025-08-06."},{"key":"10.1016\/j.neunet.2026.109617_bib0009","unstructured":"DeepSeek-AI (2024). Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model. arXiv: 2405.04434."},{"key":"10.1016\/j.neunet.2026.109617_bib0010","unstructured":"DeepSeek-AI (2025). Deepseek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv: 2501.12948."},{"key":"10.1016\/j.neunet.2026.109617_bib0011","doi-asserted-by":"crossref","unstructured":"Dou, S., Zhou, E., Liu, Y., Gao, S., Shen, W., Xiong, L., Zhou, Y., Wang, X., Xi, Z., Fan, X., Pu, S., Zhu, J., Zheng, R., Gui, T., Zhang, Q., & Huang, X. (2024). LoRAMoE: Alleviate world knowledge forgetting in large language models via MoE-style plugin. https:\/\/aclanthology.org\/2024.acl-long.106\/. 10.18653\/v1\/2024.acl-long.106.","DOI":"10.18653\/v1\/2024.acl-long.106"},{"key":"10.1016\/j.neunet.2026.109617_bib0012","series-title":"Proceedings of the 39th international conference on machine learning","first-page":"5547","article-title":"GLaM: Efficient scaling of language models with mixture-of-experts","volume":"vol. 162","author":"Du","year":"2022"},{"key":"10.1016\/j.neunet.2026.109617_bib0013","series-title":"Proceedings of the 2025 conference on empirical methods in natural language processing","first-page":"14201","article-title":"Mixture-of-clustered-experts: Advancing expert specialization and generalization in instruction tuning","author":"Eo","year":"2025"},{"issue":"120","key":"10.1016\/j.neunet.2026.109617_bib0014","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"Journal of Machine Learning Research"},{"key":"10.1016\/j.neunet.2026.109617_bib0015","unstructured":"Fu, et al. (2024). Moa: Mixture of sparse attention for automatic large language model compression. arXiv: 2406.14909."},{"key":"10.1016\/j.neunet.2026.109617_bib0016","unstructured":"Grattafiori, et al. (2024). The Llama 3 herd of models. arXiv: 2407.21783."},{"key":"10.1016\/j.neunet.2026.109617_bib0017","article-title":"Towards efficient mixture of experts: A holistic study of compression techniques","author":"He","year":"2025","journal-title":"Transactions on Machine Learning Research"},{"key":"10.1016\/j.neunet.2026.109617_bib0018","series-title":"Proceedings of the international conference on learning representations","article-title":"LoRA: Low-rank adaptation of large language models","author":"Hu","year":"2022"},{"issue":"1","key":"10.1016\/j.neunet.2026.109617_bib0019","doi-asserted-by":"crossref","first-page":"79","DOI":"10.1162\/neco.1991.3.1.79","article-title":"Adaptive mixtures of local experts","volume":"3","author":"Jacobs","year":"1991","journal-title":"Neural Computation"},{"key":"10.1016\/j.neunet.2026.109617_bib0020","unstructured":"Jiang, et al. (2024). Mixtral of experts. arXiv: 2401.04088."},{"issue":"3","key":"10.1016\/j.neunet.2026.109617_bib0021","doi-asserted-by":"crossref","first-page":"987","DOI":"10.1214\/aos\/1018031265","article-title":"Hierarchical mixtures-of-experts for exponential family regression models: Approximation and maximum likelihood estimation","volume":"27","author":"Jiang","year":"1999","journal-title":"The Annals of Statistics"},{"issue":"2","key":"10.1016\/j.neunet.2026.109617_bib0022","doi-asserted-by":"crossref","first-page":"181","DOI":"10.1162\/neco.1994.6.2.181","article-title":"Hierarchical mixtures of experts and the EM algorithm","volume":"6","author":"Jordan","year":"1994","journal-title":"Neural Computation"},{"issue":"9","key":"10.1016\/j.neunet.2026.109617_bib0023","doi-asserted-by":"crossref","first-page":"1409","DOI":"10.1016\/0893-6080(95)00014-3","article-title":"Convergence results for the EM approach to mixtures of experts architectures","volume":"8","author":"Jordan","year":"1995","journal-title":"Neural Networks"},{"key":"10.1016\/j.neunet.2026.109617_bib0024","doi-asserted-by":"crossref","unstructured":"Lu, E., et al. (2025). MoBA: Mixture of block attention for long-context LLMs. 10.48550\/arXiv.2502.13189.","DOI":"10.52202\/085713-0602"},{"key":"10.1016\/j.neunet.2026.109617_bib0025","unstructured":"Meta-AI (2025). The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation. https:\/\/ai.meta.com\/blog\/llama-4-multimodal-intelligence\/. Accessed: 2025-09-20."},{"key":"10.1016\/j.neunet.2026.109617_bib0026","series-title":"Proceedings of the international conference on learning representations","article-title":"OLMoE: Open mixture-of-experts language models","author":"Muennighoff","year":"2025"},{"key":"10.1016\/j.neunet.2026.109617_bib0027","unstructured":"OpenAI (2025). GPT-OSS-120b & GPT-OSS-20b model card. arXiv: 2508.10925."},{"key":"10.1016\/j.neunet.2026.109617_bib0028","unstructured":"Pan, B., Shen, Y., Liu, H., Mishra, M., Zhang, G., Oliva, A., Raffel, C., & Panda, R. (2024). Dense training, sparse inference: Rethinking training of mixture-of-experts language models. arXiv: 2404.05567."},{"key":"10.1016\/j.neunet.2026.109617_bib0029","series-title":"Proceedings of the 39th international conference on machine learning","first-page":"17627","article-title":"Branchformer: Parallel MLP-attention architectures to capture local and global context for speech recognition and understanding","volume":"vol. 162","author":"Peng","year":"2022"},{"key":"10.1016\/j.neunet.2026.109617_bib0030","unstructured":"Raposo, D., Ritter, S., Richards, B., Lillicrap, T., Humphreys, P. C., & Santoro, A. (2024). Mixture-of-depths: Dynamically allocating compute in transformer-based language models. 10.48550\/arXiv.2404.02258."},{"key":"10.1016\/j.neunet.2026.109617_bib0031","unstructured":"Shen, Y., Guo, Z., Cai, T., & Qin, Z. (2024). JetMoE: Reaching Llama2 performance with 0.1m dollars. arXiv: 2404.07413."},{"key":"10.1016\/j.neunet.2026.109617_bib0032","series-title":"Proceedings of the first babyLM workshop","first-page":"540","article-title":"MoEP: Modular expert paths for sample-efficient language modeling","author":"Tapaninaho","year":"2025"},{"key":"10.1016\/j.neunet.2026.109617_bib0033","unstructured":"Tapaninaho, J., & Oussalah, M. (2025). Papaformer: Language model from pre-trained parallel paths. arXiv: 2508.00544."},{"key":"10.1016\/j.neunet.2026.109617_bib0034","unstructured":"G. Team (2024a). Gemma 2: Improving open language models at a practical size. arXiv: 2408.00118."},{"key":"10.1016\/j.neunet.2026.109617_bib0035","unstructured":"G. Team (2024b). Gemma: Open models based on gemini research and technology. arXiv: 2403.08295."},{"key":"10.1016\/j.neunet.2026.109617_bib0036","unstructured":"Touvron, et al. (2023a). Llama 2: Open foundation and fine-tuned chat models. arXiv: 2307.09288."},{"key":"10.1016\/j.neunet.2026.109617_bib0037","unstructured":"Touvron, et al. (2023b). LLaMA: Open and efficient foundation language models. https:\/\/arxiv.org\/abs\/2302.13971."},{"key":"10.1016\/j.neunet.2026.109617_bib0038","series-title":"Advances in neural information processing systems","article-title":"Attention is all you need","volume":"vol. 30","author":"Vaswani","year":"2017"},{"key":"10.1016\/j.neunet.2026.109617_bib0039","series-title":"Proceedings of the 2022 conference on empirical methods in natural language processing","first-page":"5744","article-title":"AdaMix: Mixture-of-adaptations for parameter-efficient model tuning","author":"Wang","year":"2022"},{"key":"10.1016\/j.neunet.2026.109617_bib0040","series-title":"Proceedings of the international conference on learning representations","article-title":"Mixture of LoRA experts","author":"Wu","year":"2024"},{"key":"10.1016\/j.neunet.2026.109617_bib0041","series-title":"Proceedings of the international conference on learning representations","article-title":"Pushing mixture of experts to the limit: Extremely parameter efficient MoE for instruction tuning","author":"Zadouri","year":"2024"},{"key":"10.1016\/j.neunet.2026.109617_bib0042","doi-asserted-by":"crossref","unstructured":"Zhang, Z., Xia, Y., Wang, H., Yang, D., Hu, C., Zhou, X., & Cheng, D. (2024). MpMoE: Memory efficient moe for pre-trained models with adaptive pipeline parallelism. 10.1109\/TPDS.2024.3385639.","DOI":"10.1109\/TPDS.2024.3385639"},{"key":"10.1016\/j.neunet.2026.109617_bib0043","series-title":"Advances in neural information processing systems","article-title":"Mixture-of-experts with expert choice routing","volume":"vol. 35","author":"Zhou","year":"2022"},{"key":"10.1016\/j.neunet.2026.109617_bib0044","unstructured":"Zhu, Y., Wichers, N., Lin, C.-C., Wang, X., Chen, T., Shu, L., Lu, H., Liu, C., Luo, L., Chen, J., & Meng, L. (2023). SiRA: Sparse mixture of low rank adaptation. arXiv: 2311.09179."},{"key":"10.1016\/j.neunet.2026.109617_bib0045","series-title":"Proceedings of the international conference on learning representations","article-title":"Taming sparsely activated transformer with stochastic experts","author":"Zuo","year":"2022"}],"container-title":["Neural Networks"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026010749?httpAccept=text\/xml","content-type":"text\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026010749?httpAccept=text\/plain","content-type":"text\/plain","content-version":"vor","intended-application":"text-mining"}],"deposited":{"date-parts":[[2026,10,20]],"date-time":"2026-10-20T17:43:07Z","timestamp":1792518187000},"score":1,"resource":{"primary":{"URL":"https:\/\/linkinghub.elsevier.com\/retrieve\/pii\/S0893608026010749"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2027,1]]},"references-count":45,"alternative-id":["S0893608026010749"],"URL":"https:\/\/doi.org\/10.1016\/j.neunet.2026.109617","relation":{},"ISSN":["0893-6080"],"issn-type":[{"value":"0893-6080","type":"print"}],"subject":[],"published":{"date-parts":[[2027,1]]},"assertion":[{"value":"Elsevier","name":"publisher","label":"This article is maintained by"},{"value":"MoEP: Compact and efficient sparsity with modular expert paths","name":"articletitle","label":"Article Title"},{"value":"Neural Networks","name":"journaltitle","label":"Journal Title"},{"value":"https:\/\/doi.org\/10.1016\/j.neunet.2026.109617","name":"articlelink","label":"CrossRef DOI link to publisher maintained version"},{"value":"article","name":"content_type","label":"Content Type"},{"value":"\u00a9 2026 The Author(s). Published by Elsevier Ltd.","name":"copyright","label":"Copyright"}],"article-number":"109617"}}