{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T03:31:23Z","timestamp":1777865483861,"version":"3.51.4"},"reference-count":52,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100004395","name":"Institute for Information & Communications Technology Planning & Evaluation (IITP)","doi-asserted-by":"publisher","award":["IITP-2025-RS-2023-00254129,RS-2024-00425354,RS-202400436934"],"award-info":[{"award-number":["IITP-2025-RS-2023-00254129,RS-2024-00425354,RS-202400436934"]}],"id":[{"id":"10.13039\/501100004395","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.01870","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"20105-20115","source":"Crossref","is-referenced-by-count":0,"title":["Task Vector Quantization for Memory-Efficient Model Merging"],"prefix":"10.1109","author":[{"given":"Youngeun","family":"Kim","sequence":"first","affiliation":[{"name":"Yale University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Seunghwan","family":"Lee","sequence":"additional","affiliation":[{"name":"Sungkyunkwan University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Aecheon","family":"Jung","sequence":"additional","affiliation":[{"name":"Sungkyunkwan University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Bogon","family":"Ryu","sequence":"additional","affiliation":[{"name":"Sungkyunkwan University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sungeun","family":"Hong","sequence":"additional","affiliation":[{"name":"Sungkyunkwan University"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","first-page":"3091","article-title":"Pareto-optimal quantized resnet is mostly 4-bit","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"AmirAli","year":"2021"},{"key":"ref2","volume-title":"Qreg: Onregularization effects of quantization","author":"MohammadHossein","year":"2022"},{"key":"ref3","article-title":"Post training 4-bit quantization of convolutional networks for rapid deployment","volume":"32","author":"Ron","year":"2019","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1023\/a:1007379606734"},{"key":"ref5","volume-title":"Pact: Parameterized clipping activation for quantized neural networks","author":"Jungwook","year":"2018"},{"key":"ref6","first-page":"217","article-title":"Intra-inter modal attention blocks for rgb-d semantic segmentation","volume-title":"Proc. of Int\u2019l Conf. on Multimedia Retrieval (ICMR)","author":"Soyun","year":"2023"},{"key":"ref7","first-page":"270","article-title":"Model breadcrumbs: Scaling multi-task model merging with sparse masks","volume-title":"Proc. of European Conf. on Computer Vision (ECCV)","author":"MohammadReza","year":"2024"},{"key":"ref8","author":"Steven K","year":"2019","journal-title":"Learned step size quantization"},{"key":"ref9","first-page":"69","article-title":"Post-training piecewise linear quantization for deep neural networks","volume-title":"Computer Vision\u2013ECCV 2020: 16th European Conference","author":"Jun","year":"2020"},{"key":"ref10","volume-title":"Fighting quantization bias with bias","author":"Alexander","year":"2019"},{"key":"ref11","first-page":"18695","article-title":"Task singular vectors: Reducing task interference in model merging","volume-title":"Proceedings of the Computer Vision and Pattern Recognition Conference","author":"Antonio Andrea","year":"2025"},{"key":"ref12","article-title":"Loss surfaces, mode connectivity, and fast ensembling of dnns","volume-title":"Proc. of Neural Information Processing Systems (NeurIPS)","volume":"31","author":"Timur","year":"2018"},{"key":"ref13","first-page":"770","article-title":"Deep residual learning for image recognition","volume-title":"Proc. of Computer Vision and Pattern Recognition (CVPR)","author":"Kaiming","year":"2016"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.52202\/079017-3900"},{"issue":"(1)","key":"ref15","first-page":"6869","article-title":"Quantized neural networks: Training neural networks with low precision weights and activations","volume":"18","author":"Itay","year":"2017","journal-title":"The Journal of Machine Learning Research"},{"key":"ref16","volume-title":"Improving post training neural quantization: Layer-wise calibration and integer programming","author":"Itay","year":"2020"},{"key":"ref17","article-title":"Editing models with task arithmetic","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Gabriel","year":"2023"},{"key":"ref18","article-title":"Model stock: All we need is just a few fine-tuned models","volume-title":"Proc. of European Conf. on Computer Vision (ECCV)","author":"Dong-Hwan","year":"2024"},{"key":"ref19","first-page":"1","article-title":"Introduction to nvidia jetson nano","volume-title":"IoT Projects with NVIDIA Jetson Nano: AI-Enabled Internet of Things Projects for Beginners","author":"Agus","year":"2021"},{"key":"ref20","article-title":"Brecq: Pushing the limit of post-training quantization by block reconstruction","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Yuhang","year":"2021"},{"key":"ref21","volume-title":"Roberta: A robustly optimized bert pretraining approach","author":"Y","year":"1907"},{"key":"ref22","first-page":"28092","article-title":"Post-training quantization for vision transformer","volume":"34","author":"Zhenhua","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref23","volume-title":"Llm-qat: Data-free quantization aware training for large language models","author":"Zechun","year":"2023"},{"key":"ref24","volume-title":"Kivi: A tuning-free asymmetric 2bit quantization for kv cache","author":"Zirui","year":"2024"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2505"},{"key":"ref26","first-page":"379","article-title":"Magmax: Leveraging model merging for seamless continual learning","volume-title":"Proc. of European Conf. on Computer Vision (ECCV)","author":"Daniel","year":"2024"},{"key":"ref27","first-page":"7197","article-title":"Up or down? adaptive rounding for post-training quantization","volume-title":"International Conference on Machine Learning","author":"Markus","year":"2020"},{"issue":"(11-12)","key":"ref28","first-page":"3245","article-title":"Loss aware post-training quantization","volume":"110","author":"Yury","year":"2021","journal-title":"Machine Learning"},{"key":"ref29","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Alec","year":"2021"},{"key":"ref30","first-page":"746","article-title":"Indoor segmentation and support inference from rgbd images","volume-title":"Proc. of European Conf. on Computer Vision (ECCV)","author":"Nathan","year":"2012"},{"key":"ref31","volume-title":"Fusionbench: A comprehensive benchmark of deep model fusion","author":"Anke","year":"2024"},{"key":"ref32","article-title":"Merging multi-task models via weight ensembling mixture of experts","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Anke","year":"2024"},{"issue":"(7)","key":"ref33","first-page":"3614","article-title":"Multi-task learning for dense prediction tasks: A survey","volume":"44","author":"Simon","year":"2021","journal-title":"IEEE Trans. on Pattern Anal. Mach. Intell. (TPAMI)"},{"key":"ref34","article-title":"Glue: Amulti-taskbenchmarkandanalysisplatform for natural language understanding","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Alex","year":"2019"},{"key":"ref35","article-title":"Localizing task information for improved model merging and compression","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Ke","year":"2024"},{"key":"ref36","article-title":"Lines: Post-training layer scaling prevents forgetting and enhances model merging [C]","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Ke","year":"2025"},{"key":"ref37","article-title":"Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Xiuying","year":"2022"},{"key":"ref38","first-page":"23965","article-title":"Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Mitchell","year":"2022"},{"key":"ref39","first-page":"7959","article-title":"Robust fine-tuning of zero-shot models","volume-title":"Proc. of Computer Vision and Pattern Recognition (CVPR)","author":"Mitchell","year":"2022"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2010.5539970"},{"key":"ref41","article-title":"Ties-merging: Resolving interference whenmerging models","volume-title":"Proc. of Neural Information Processing Systems (NeurIPS)","volume":"36","author":"Prateek","year":"2024"},{"key":"ref42","first-page":"5029","article-title":"Learnable companding quantization for accurate low-bit neural networks","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Kohei","year":"2021"},{"key":"ref43","article-title":"Representation surgery for multi-task model merging","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Enneng","year":"2024"},{"key":"ref44","article-title":"Adamerging: Adaptive model merging for multi-task learning","volume-title":"Proc. of Int\u2019l Conf. on Learning Representation (ICLR)","author":"Enneng","year":"2024"},{"key":"ref45","volume-title":"Understanding straight-through estimator in training activation quantized neural nets","author":"Penghang","year":"2019"},{"key":"ref46","volume-title":"How to parameterize asymmetric quantization ranges for quantization-aware training","author":"Jaeseong","year":"2024"},{"key":"ref47","article-title":"Language models are super mario: Absorbing abilities from homologous models as a free lunch","volume-title":"Proc. of Int\u2019l Conf. on Machine Learning (ICML)","author":"Le","year":"2024"},{"key":"ref48","first-page":"365","article-title":"Lq-nets: Learned quantization for highly accurate and compact deep neural networks","volume-title":"Proceedings of the European conference on computer vision (ECCV)","author":"Dongqing","year":"2018"},{"issue":"(2)","key":"ref49","first-page":"305","article-title":"A survey on negative transfer","volume":"10","author":"Wen","year":"2022","journal-title":"IEEE\/CAA Journal of Automatica Sinica"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2025.111376"},{"key":"ref51","first-page":"7543","article-title":"Improving neural network quantization without retraining using outlier channel splitting","volume-title":"International Conference on Machine Learning","author":"Ritchie","year":"2019"},{"key":"ref52","volume-title":"Dorefa-net: Training low bitwidth convolutional neural networks with low bitwidth gradients","author":"Shuchang","year":"2016"}],"event":{"name":"2025 IEEE\/CVF International Conference on Computer Vision (ICCV)","location":"Honolulu, HI, USA","start":{"date-parts":[[2025,10,19]]},"end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/CVF International Conference on Computer Vision (ICCV)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11443115\/11443287\/11446046.pdf?arnumber=11446046","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T06:40:13Z","timestamp":1777531213000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11446046\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":52,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01870","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}