{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,3]],"date-time":"2026-11-03T13:37:26Z","timestamp":1793713046244,"version":"4.6.1"},"reference-count":107,"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"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.02129","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"22934-22946","source":"Crossref","is-referenced-by-count":6,"title":["Cycle Consistency as Reward: Learning Image-Text Alignment Without Human Preferences"],"prefix":"10.1109","author":[{"given":"Hyojin","family":"Bahng","sequence":"first","affiliation":[{"name":"MIT CSAIL"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Caroline","family":"Chan","sequence":"additional","affiliation":[{"name":"MIT CSAIL"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Fr\u00e9do","family":"Durand","sequence":"additional","affiliation":[{"name":"MIT CSAIL"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Phillip","family":"Isola","sequence":"additional","affiliation":[{"name":"MIT CSAIL"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46454-1_24"},{"issue":"2","key":"ref2","article-title":"Qwen-vl: A frontier large vision-language model with versatile abilities","volume":"1","author":"Bai","year":"2023","journal-title":"arXiv preprint"},{"issue":"3","key":"ref3","article-title":"Improving image generation with better captions","volume":"2","author":"Betker","year":"2023","journal-title":"Computer Science"},{"key":"ref4","article-title":"Training diffusion models with reinforcement learning","volume-title":"The Twelfth International Conference on Learning Representations","author":"Black","year":"2024"},{"key":"ref5","volume-title":"Announcing black forest labs","year":"2024"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1177\/135910457000100301"},{"key":"ref7","article-title":"Clair: Evaluating image captions with large language models","volume-title":"EMNLP","author":"David","year":"2023"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_22"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1007\/s11432-024-4231-5"},{"key":"ref10","article-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021","journal-title":"arXiv preprint"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2142"},{"key":"ref12","article-title":"Less is more: Improving llm alignment via preference data selection","author":"Deng","year":"2025","journal-title":"arXiv preprint"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.798"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.18653\/vl\/N19-142"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.212"},{"key":"ref16","article-title":"Benchmarking and improving detail image caption","author":"Dong","year":"2024","journal-title":"arXiv preprint"},{"key":"ref17","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2021","journal-title":"ICLR"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3327"},{"key":"ref19","article-title":"The llama 3 herd of models","author":"Dubey","year":"2024","journal-title":"arXiv preprint"},{"key":"ref20","article-title":"Scaling rectified flow transformers for high-resolution image synthesis","volume-title":"Forty-first International Conference on Machine Learning","author":"Esser","year":"2024"},{"key":"ref21","article-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models","author":"Fu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref22","article-title":"Dreamsim: Learning new dimensions of human visual similarity using synthetic data","author":"Fu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.6"},{"key":"ref24","article-title":"Gemini: a family of highly capable multimodal models","volume-title":"arXiv preprint","year":"2023"},{"key":"ref25","first-page":"270","article-title":"Unsupervised monocular depth estimation with leftright consistency","volume-title":"Proceedings of the IEEE conference on computer vision and pattern recognition","author":"Godard","year":"2017"},{"key":"ref26","article-title":"Text-to-image-totext translation using cycle consistent adversarial networks","author":"Krishna Gorti","year":"2018","journal-title":"arXiv preprint"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i16.29771"},{"key":"ref28","article-title":"Dual learning for machine translation","author":"He","year":"2016","journal-title":"Advances in neural information processing systems"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.595"},{"key":"ref30","first-page":"1989","article-title":"Cycada: Cycle-consistent adversarial domain adaptation","volume-title":"International conference on machine learning","author":"Hoffman","year":"2018"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.01354"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1007\/978-981-96-2071-5_30"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3443"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1111\/cgf.12184"},{"key":"ref35","article-title":"The platonic representation hypothesis","volume-title":"International Conference on Machine Learning","author":"Huh","year":"2024"},{"key":"ref36","first-page":"19545","article-title":"Space-time correspondence as a contrastive random walk","volume":"33","author":"Jabri","year":"2020","journal-title":"Advances in neural information processing systems"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00263"},{"key":"ref38","article-title":"Regularized best-of-n sampling to mitigate reward hacking for language model alignment","volume-title":"ICML 2024 Workshop on Models of Human Feedback for AI Alignment","author":"Jinnai","year":"2024"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/ICPR.2010.675"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1594"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.3115\/1626355.1626389"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.205"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/cvprw63382.2024.00538"},{"key":"ref44","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref45","article-title":"Alice: Towards understanding adversarial learning for joint distribution matching","volume":"30","author":"Li","year":"2017","journal-title":"Advances in neural information processing systems"},{"key":"ref46","first-page":"12888","article-title":"Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation","volume-title":"International conference on machine learning","author":"Li","year":"2022"},{"key":"ref47","first-page":"19730","article-title":"Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language.models","volume-title":"International conference on machine learning","author":"Li","year":"2023"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.358"},{"key":"ref49","article-title":"Leveraging unpaired data for vision-language generative models via cycle consistency","volume-title":"The Twelfth International Conference on Learning Representations","author":"Li","year":"2024"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.20"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72673-6_20"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1516"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02484"},{"key":"ref55","article-title":"Visual instruction tuning","volume":"36","author":"Liu","year":"2024","journal-title":"Advances in neural information processing systems"},{"key":"ref56","article-title":"Decoupled weight decay regularization","volume-title":"7th International Conference on Learning Representations (ICLR)","author":"Loshchilov","year":"2019"},{"key":"ref57","article-title":"Decoupled weight decay regularization","volume-title":"International Conference on Learning Representations","author":"Loshchilov","year":"2019"},{"key":"ref58","article-title":"Inference-time scaling for diffusion models beyond scaling denoising steps","author":"Ma","year":"2025","journal-title":"arXiv preprint"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/LRA.2022.3145053"},{"key":"ref60","article-title":"Webgpt: Browser-assisted question-answering with human feedback","author":"Nakano","year":"2021","journal-title":"arXiv preprint"},{"key":"ref61","volume-title":"Hello gpt-4o","year":"2024"},{"key":"ref62","volume-title":"Gpt-4v(ision) system card","year":"2023"},{"key":"ref63","article-title":"OpenGVLab","volume-title":"Internvl-2.0","year":"2024"},{"key":"ref64","article-title":"Dinov2: Learning robust visual features without supervision","author":"Oquab","year":"2024","journal-title":"Transactions on Machine Learning Research"},{"key":"ref65","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2011"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.3115\/1073083.1073135"},{"key":"ref67","article-title":"Sdxl: Improving latent diffusion models for high-resolution image synthesis","author":"Podell","year":"2023","journal-title":"arXiv preprint"},{"key":"ref68","author":"Prabhudesai","year":"2023","journal-title":"Aligning text-to-image diffusion models with reward backpropagation"},{"key":"ref69","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"International conference on machine learning","author":"Radford","year":"2021"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2338"},{"key":"ref71","article-title":"Selfeval: Leveraging discriminative nature of generative models for evaluation","author":"Saketh Rambhatla","year":"2025","journal-title":"Transactions on Machine Learning Research"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D19-1410"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2643"},{"key":"ref75","article-title":"Adversarial diffusion distillation","author":"Sauer","year":"2023","journal-title":"arXiv preprint"},{"key":"ref76","article-title":"Proximal policy optimization algorithms","author":"Schulman","year":"2017","journal-title":"arXiv preprint"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00681"},{"key":"ref78","article-title":"Deepseekmath: Pushing the limits of mathematical reasoning in open language models","author":"Shao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref79","article-title":"Synth2: Boosting visual-language models with synthetic captions and image embeddings","author":"Sharifzadeh","year":"2024","journal-title":"arXiv preprint"},{"key":"ref80","first-page":"3008","article-title":"Christiano. Learning to summarize with human feedback","volume":"33","author":"Stiennon","year":"2020","journal-title":"Advances in neural information processing systems"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.775"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-642-15549-9_32"},{"key":"ref83","article-title":"Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context","author":"Team","year":"2024","journal-title":"arXiv preprint"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.02521"},{"key":"ref85","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7299087"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00786"},{"key":"ref87","doi-asserted-by":"publisher","DOI":"10.1109\/iccv.2013.110"},{"key":"ref88","author":"Wang","year":"2023","journal-title":"Cogvlm: Visual expert for pretrained language models"},{"key":"ref89","article-title":"Human preference score v2: A solid benchmark for evaluating human preferences of text-to-image synthesis","author":"Wu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00200"},{"key":"ref91","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v40i13.38107"},{"key":"ref92","article-title":"Imagereward: Learning and evaluating human.preferences for text-to-image generation","volume":"36","author":"Xu","year":"2024","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref93","article-title":"Painting with words: Elevating detailed image captioning with benchmark and alignment learning","volume-title":"The Thirteenth International Conference on Learning Representations","author":"Ye","year":"2025"},{"key":"ref94","article-title":"How reliable is human feedback for aligning large language models?","author":"Yeh","year":"2024","journal-title":"arXiv preprint"},{"key":"ref95","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.310"},{"key":"ref96","article-title":"Scaling autoregressive models for content-rich text-to-image generation","author":"Yu","year":"2022","journal-title":"Transactions on Machine Learning Research"},{"key":"ref97","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01310"},{"key":"ref98","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.01861"},{"key":"ref99","article-title":"When and why visionlanguage.models behave like bags-of-words, and what to.do about it?","volume-title":"The Eleventh International Conference on Learning Representations","author":"Yuksekgonul","year":"2023"},{"key":"ref100","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2010.5539801"},{"key":"ref101","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2018.00068"},{"key":"ref102","article-title":"Bertscore: Evaluating text generation with bert","author":"Zhang","year":"2019","journal-title":"arXiv preprint"},{"key":"ref103","article-title":"Beyond hallucinations: Enhancing lvlms through hallucination-aware direct preference optimization","author":"Zhao","year":"2023","journal-title":"arXiv preprint"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2015.7298723"},{"key":"ref105","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr.2016.20"},{"key":"ref106","article-title":"Aligning modalities in vision large language models via preference fine-tuning","volume-title":"ICLR 2024 Workshop on Reliable and Responsible Foundation.Models","author":"Zhou","year":"2024"},{"key":"ref107","doi-asserted-by":"publisher","DOI":"10.1109\/iccv.2017.244"}],"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\/11443599.pdf?arnumber=11443599","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T06:22:51Z","timestamp":1777530171000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11443599\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":107,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.02129","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}