{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,7]],"date-time":"2026-04-07T21:11:04Z","timestamp":1775596264865,"version":"3.50.1"},"reference-count":75,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"4","license":[{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T00:00:00Z","timestamp":1775001600000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"Research Grants Council (RGC) of the Hong Kong SAR"},{"name":"General Research Fund","award":["17203023"],"award-info":[{"award-number":["17203023"]}]},{"name":"Collaborative Research Fund","award":["C5052-23G"],"award-info":[{"award-number":["C5052-23G"]}]},{"name":"NSFC\/RGC Collaborative Research Scheme","award":["CRS_HKU703\/24"],"award-info":[{"award-number":["CRS_HKU703\/24"]}]},{"name":"UBTECH Robotics"},{"name":"Jockey Club STEM Lab of Multimedia and Machine Learning funded by The Hong Kong Jockey Club Charities Trust"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Circuits Syst. Video Technol."],"published-print":{"date-parts":[[2026,4]]},"DOI":"10.1109\/tcsvt.2025.3637139","type":"journal-article","created":{"date-parts":[[2025,11,25]],"date-time":"2025-11-25T18:30:48Z","timestamp":1764095448000},"page":"5093-5105","source":"Crossref","is-referenced-by-count":0,"title":["Complexity-Aware Policy via Heterogeneous Experts for Robotic Manipulation"],"prefix":"10.1109","volume":"36","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-4277-7702","authenticated-orcid":false,"given":"Ming","family":"Yang","sequence":"first","affiliation":[{"name":"School of Software, Beihang University, Beijing, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4895-7720","authenticated-orcid":false,"given":"Yunzhe","family":"Hu","sequence":"additional","affiliation":[{"name":"School of Computing and Data Science, The University of Hong Kong, Pokfulam, SAR, Hong Kong"}]},{"ORCID":"https:\/\/orcid.org\/0009-0005-1734-5616","authenticated-orcid":false,"given":"Ge","family":"Yuan","sequence":"additional","affiliation":[{"name":"School of Computing and Data Science, The University of Hong Kong, Pokfulam, SAR, Hong Kong"}]},{"given":"Difan","family":"Zou","sequence":"additional","affiliation":[{"name":"School of Computing and Data Science, The University of Hong Kong, Pokfulam, SAR, Hong Kong"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3985-5802","authenticated-orcid":false,"given":"Jianxin","family":"Pang","sequence":"additional","affiliation":[{"name":"UBTech Robotics Corporation, Shenzhen, China"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2775-9730","authenticated-orcid":false,"given":"Dong","family":"Xu","sequence":"additional","affiliation":[{"name":"School of Computing and Data Science, The University of Hong Kong, Pokfulam, SAR, Hong Kong"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1177\/02783649241273668"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.15607\/RSS.2023.XIX.028"},{"key":"ref3","first-page":"1949","article-title":"3D diffuser actor: Policy diffusion with 3D scene representations","volume-title":"Proc. Conf. Robot Learn.","author":"Ke"},{"key":"ref4","first-page":"9902","article-title":"Planning with diffusion for flexible behavior synthesis","volume-title":"Proc. 39th Int. Conf. Mach. Learn.","author":"J\u00e4nner"},{"key":"ref5","first-page":"1","article-title":"Is conditional generative modeling all you need for decision making?","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Ajay"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"ref7","first-page":"1","article-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Shazeer"},{"issue":"120","key":"ref8","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2021","journal-title":"J. Mach. Learn. Res."},{"key":"ref9","article-title":"DeepSeek-V3 technical report","volume-title":"arXiv:2412.19437","author":"Liu","year":"2024"},{"key":"ref10","article-title":"Mixtral of experts","author":"Jiang","year":"2024","journal-title":"arXiv:2401.04088"},{"key":"ref11","first-page":"649","article-title":"Sparse diffusion policy: A sparse, reusable, and flexible policy for robot learning","volume-title":"Proc. Conf. Robot Learn.","author":"Wang"},{"key":"ref12","first-page":"1","article-title":"Efficient diffusion transformer policies with mixture of expert denoisers for multitask learning","volume-title":"Proc. 13th Int. Conf. Learn. Represent.","author":"Reuss"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.15607\/RSS.2024.XX.090"},{"key":"ref14","first-page":"1","article-title":"Zero-shot robotic manipulation with pre-trained image-editing diffusion models","volume-title":"Proc. 12th Int. Conf. Learn. Represent.","author":"Black"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00684"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01712"},{"key":"ref17","first-page":"2323","article-title":"Chaineddiffuser: Unifying trajectory diffusion and keypose prediction for robotic manipulation","volume-title":"Proc. Conf. Robot Learn.","author":"Xian"},{"key":"ref18","first-page":"1","article-title":"Plan-seq-learn: Language model guided RL for solving long horizon robotics tasks","volume-title":"Proc. 12th Int. Conf. Learn. Represent.","author":"Dalal"},{"key":"ref19","first-page":"1","article-title":"Flow straight and fast: Learning to generate and transfer data with rectified flow","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Liu"},{"key":"ref20","first-page":"1","article-title":"Flow matching for generative modeling","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Lipman"},{"key":"ref21","first-page":"1","article-title":"Building normalizing flows with stochastic interpolants","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Albergo"},{"key":"ref22","first-page":"2256","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sohl-Dickstein"},{"key":"ref23","first-page":"6840","article-title":"Denoising diffusion probabilistic models","volume-title":"Proc. NIPS","author":"Ho"},{"key":"ref24","first-page":"11918","article-title":"Generative modeling by estimating gradients of the data distribution","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Song"},{"key":"ref25","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"Proc. Int. Conf. Mach. Learn.","volume":"139","author":"Radford"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01100"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2643"},{"key":"ref29","first-page":"16784","article-title":"GLIDE: Towards photorealistic image generation and editing with text-guided diffusion models","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Nichol"},{"key":"ref30","first-page":"8633","article-title":"Video diffusion models","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"35","author":"Ho"},{"key":"ref31","article-title":"Stable video diffusion: Scaling latent video diffusion models to large datasets","author":"Blattmann","year":"2023","journal-title":"arXiv:2311.15127"},{"key":"ref32","first-page":"1","article-title":"DreamFusion: Text-to-3D using 2D diffusion","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Poole"},{"key":"ref33","first-page":"3766","article-title":"Scaling up and distilling down: Language-guided robot skill acquisition","volume-title":"Proc. Conf. Robot Learn.","author":"Ha"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.15607\/RSS.2024.XX.121"},{"key":"ref35","first-page":"2012","article-title":"PlayFusion: Skill acquisition via diffusion from language-annotated play","volume-title":"Proc. Conf. Robot Learn.","author":"Chen"},{"key":"ref36","first-page":"6948","article-title":"Exploring the adversarial vulnerabilities of vision-language-action models in robotics","volume-title":"Proc. IEEE\/CVF Int. Conf. Comput. Vis.","author":"Wang"},{"key":"ref37","first-page":"12606","article-title":"Scaling rectified flow transformers for high-resolution image synthesis","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Esser"},{"key":"ref38","article-title":"Movie gen: A cast of media foundation models","author":"Polyak","year":"2024","journal-title":"arXiv:2410.13720"},{"key":"ref39","first-page":"41052","article-title":"MusicFlow: Cascaded flow matching for text guided music generation","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Prajwal"},{"key":"ref40","first-page":"107198","article-title":"Trajectory flow matching with applications to clinical time series modelling","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Bengio"},{"key":"ref41","first-page":"33007","article-title":"Sequence-augmented SE(3)-flow matching for conditional protein generation","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Akhound-Sadegh"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.15607\/RSS.2024.XX.122"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.15607\/rss.2025.xxi.010"},{"key":"ref44","first-page":"1","article-title":"Robot manipulation with flow matching","volume-title":"Proc. CoRL Workshop Mastering Robot Manipulation World Abundant Data","author":"Zhang"},{"key":"ref45","first-page":"5547","article-title":"GLaM: Efficient scaling of language models with mixture-of-experts","volume-title":"Proc. Int. Conf. Mach. Learn.","volume":"162","author":"Du"},{"key":"ref46","first-page":"1","article-title":"GShard: Scaling giant models with conditional computation and automatic sharding","volume-title":"Proc. Int. Conf. Learn. Represent.","author":"Lepikhin"},{"key":"ref47","article-title":"No language left behind: Scaling human-centered machine translation","author":"Team","year":"2022","journal-title":"arXiv:2207.04672"},{"key":"ref48","first-page":"1","article-title":"Mixture of LoRA experts","volume-title":"Proc. 12th Int. Conf. Learn. Represent.","author":"Wu"},{"key":"ref49","first-page":"8583","article-title":"Scaling vision with sparse mixture of experts","volume-title":"Proc. NIPS","author":"Riquelme"},{"key":"ref50","first-page":"269","article-title":"Tutel: Adaptive mixture-of-experts at scale","volume-title":"Proc. Mach. Learn. Syst.","author":"Hwang"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2025.3577027"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2025.3556422"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2025.3531003"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/tcsvt.2025.3586550"},{"key":"ref55","first-page":"9564","article-title":"Multimodal contrastive learning with LIMoE: The language-image mixture of experts","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Mustafa"},{"key":"ref56","first-page":"11329","article-title":"Scaling vision-language models with sparse mixture of experts","volume-title":"Proc. Findings Assoc. Comput. Linguistics, EMNLP","author":"Shen"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73397-0_18"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/tcsvt.2025.3587703"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2024.3474101"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0515"},{"key":"ref61","first-page":"29335","article-title":"DSelect-k: Differentiable selection in the mixture of experts with applications to multi-task learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Hazimeh"},{"key":"ref62","first-page":"1","article-title":"Lory: Fully differentiable mixture-of-experts for autoregressive language model pre-training","volume-title":"Proc. 1st Conf. Lang. Model.","author":"Zhong"},{"key":"ref63","first-page":"18332","article-title":"DeepSpeed-MoE: Advancing mixture-of-experts inference and training to power next-generation AI scale","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Rajbhandari"},{"key":"ref64","first-page":"1","article-title":"Visual agents as fast and slow thinkers","volume-title":"Proc. 13th Int. Conf. Learn. Represent.","author":"Sun"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.1109\/LRA.2022.3180108"},{"key":"ref66","first-page":"6572","article-title":"Neural ordinary differential equations","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"31","author":"Chen"},{"key":"ref67","article-title":"Sparse mixture-of-experts for compositional generalization: Empirical evidence and theoretical foundations of optimal sparsity","author":"Zhao","year":"2024","journal-title":"arXiv:2410.13964"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1017\/CBO9781107298019"},{"key":"ref69","first-page":"1","article-title":"Classifier-free diffusion guidance","volume-title":"Proc. NeurIPS Workshop Deep Generative Models Downstream Appl.","author":"Ho"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1939"},{"key":"ref71","first-page":"3705","article-title":"Evaluating real-world robot manipulation policies in simulation","volume-title":"Proc. Conf. Robot Learn.","author":"Li"},{"key":"ref72","article-title":"Open X-embodiment: Robotic learning datasets and RT-X models","author":"O\u2019Neill","year":"2023","journal-title":"arXiv:2310.08864"},{"key":"ref73","first-page":"1","article-title":"Vision-language foundation models as effective robot imitators","volume-title":"Proc. 12th Int. Conf. Learn. Represent.","author":"Li"},{"key":"ref74","first-page":"1","article-title":"Unleashing large-scale video generative pre-training for visual robot manipulation","volume-title":"Proc. 12th Int. Conf. Learn. Represent.","author":"Wu"},{"key":"ref75","first-page":"4062","article-title":"QueST: Self-supervised skill abstractions for learning continuous control","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"37","author":"Mete"}],"container-title":["IEEE Transactions on Circuits and Systems for Video Technology"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/76\/11475579\/11268499.pdf?arnumber=11268499","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,7]],"date-time":"2026-04-07T20:03:30Z","timestamp":1775592210000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11268499\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4]]},"references-count":75,"journal-issue":{"issue":"4"},"URL":"https:\/\/doi.org\/10.1109\/tcsvt.2025.3637139","relation":{},"ISSN":["1051-8215","1558-2205"],"issn-type":[{"value":"1051-8215","type":"print"},{"value":"1558-2205","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4]]}}}