{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:03:55Z","timestamp":1777889035404,"version":"3.51.4"},"reference-count":50,"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\/iccv51701.2025.01679","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"18068-18077","source":"Crossref","is-referenced-by-count":0,"title":["Rethinking DPO-Style Diffusion Aligning Frameworks"],"prefix":"10.1109","author":[{"given":"Xun","family":"Wu","sequence":"first","affiliation":[{"name":"Microsoft Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shaohan","family":"Huang","sequence":"additional","affiliation":[{"name":"Microsoft Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Lingjie","family":"Jiang","sequence":"additional","affiliation":[{"name":"Peking University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Furu","family":"Wei","sequence":"additional","affiliation":[{"name":"Microsoft Research"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Training diffusion models with reinforcement learning","author":"Black","year":"2023","journal-title":"arXiv preprint"},{"key":"ref2","article-title":"Training diffusion models with reinforcement learning","volume-title":"ICLR. OpenReview.net","author":"Black","year":"2024"},{"key":"ref3","article-title":"Simple hierarchical planning with diffusion","volume-title":"ICLR. OpenReview.net","author":"Chen","year":"2024"},{"key":"ref4","article-title":"Diffusion posterior sampling for general noisy inverse problems","volume-title":"ICLR. OpenReview.net","author":"Chung","year":"2023"},{"key":"ref5","article-title":"Directly fine-tuning diffusion models on differentiable rewards","author":"Clark","year":"2023","journal-title":"arXiv preprint"},{"key":"ref6","first-page":"8780","article-title":"Diffusion models beat gans on image synthesis","volume-title":"NeurIPS","author":"Dhariwal","year":"2021"},{"key":"ref7","first-page":"11808","article-title":"Nerfdiff: Single-image view synthesis with nerf-guided distillation from 3d-aware diffusion","volume-title":"ICML","author":"Gu","year":"2023"},{"key":"ref8","article-title":"Gradient guidance for diffusion models: An optimization perspective","volume":"abs\/2404.14743","author":"Guo","year":"2024","journal-title":"CoRR"},{"key":"ref9","article-title":"Denoising diffusion probabilistic models","volume-title":"NeurIPS","author":"Ho","year":"2020"},{"key":"ref10","article-title":"Imagen video: High definition video generation with diffusion models","volume":"abs\/2210.02303","author":"Ho","year":"2022","journal-title":"CoRR"},{"key":"ref11","first-page":"47:1","article-title":"Cascaded diffusion models for high fidelity image generation","volume":"23","author":"Ho","year":"2022","journal-title":"J. Mach. Learn. Res."},{"key":"ref12","article-title":"Lora: Low-rank adaptation of large language models","volume-title":"ICLR. OpenReview.net","author":"Hu","year":"2022"},{"key":"ref13","first-page":"9902","article-title":"Planning with diffusion for flexible behavior synthesis","volume-title":"ICML","author":"Janner","year":"2022"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00813"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1594"},{"key":"ref16","article-title":"Reward design with language models","volume-title":"The Eleventh International Conference on Learning Representations","author":"Kwon","year":"2022"},{"key":"ref17","article-title":"Step-dpo: Step-wise preference optimization for long-chain reasoning of 11 ms","author":"Lai","year":"2024","journal-title":"arXiv preprint"},{"key":"ref18","article-title":"Aligning text-to-image models using human feedback","author":"Lee","year":"2023","journal-title":"arXiv preprint"},{"key":"ref19","article-title":"Step-aware preference optimization: Aligning preference with denoising performance at each step","volume":"abs\/2406.04314","author":"Liang","year":"2024","journal-title":"CoRR"},{"issue":"5","key":"ref20","first-page":"7","article-title":"Step-aware preference optimization: Aligning preference with denoising performance at each step","volume":"2","author":"Liang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref21","first-page":"21450","article-title":"Audioldm: Text-to-audio generation with latent diffusion models","volume-title":"ICML","author":"Liu","year":"2023"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.00750"},{"key":"ref23","first-page":"16784","article-title":"GLIDE: towards photorealistic image generation and editing with text-guided diffusion models","volume-title":"ICML","author":"Nichol","year":"2022"},{"key":"ref24","article-title":"Dreamfusion: Text-to-3d using 2d diffusion","volume-title":"ICLR. OpenReview.net","author":"Poole","year":"2023"},{"key":"ref25","article-title":"Aligning text-to-image diffusion models with reward backpropagation","author":"Prabhudesai","year":"2023","journal-title":"arXiv preprint"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2338"},{"key":"ref27","article-title":"Direct preference optimization: Your language model is secretly a reward model","volume-title":"NeurIPS","author":"Rafailov","year":"2023"},{"issue":"2","key":"ref28","article-title":"Hierarchical text-conditional image generation with clip latents","volume":"1","author":"Ramesh","year":"2022","journal-title":"arXiv preprint"},{"key":"ref29","article-title":"Refining alignment framework for diffusion models with intermediate-step preference ranking","author":"Ren","year":"2025","journal-title":"arXiv preprint"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.344"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2643"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2643"},{"key":"ref36","article-title":"Make-a-video: Text-to-video generation without text-video data","volume-title":"ICLR. OpenReview.net","author":"Singer","year":"2023"},{"key":"ref37","first-page":"2256","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"ICML","author":"Sohl-Dickstein","year":"2015"},{"key":"ref38","article-title":"Denoising diffusion implicit models","author":"Song","year":"2020","journal-title":"arXiv preprint"},{"key":"ref39","article-title":"Denoising diffusion implicit models","author":"Song","year":"2021","journal-title":"ICLR. OpenReview.net"},{"key":"ref40","article-title":"Zerothorder optimization meets human feedback: Provable learning via ranking oracles","author":"Tang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref41","article-title":"Diffusion model alignment using direct preference optimization","author":"Wallace","year":"2023","journal-title":"arXiv preprint"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00786"},{"key":"ref43","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":"ref44","article-title":"Multimodal large language model is a human-aligned annotator for text-toimage generation","author":"Wu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref45","article-title":"Imagereward: Learning and evaluating human preferences for text-to-image generation","author":"Xu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref46","volume-title":"Rlcd: Reinforcement learning from contrast distillation for language model alignment","author":"Yang","year":"2023"},{"key":"ref47","article-title":"Using human feedback to fine-tune diffusion models without any reward model","author":"Yang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00854"},{"key":"ref49","article-title":"A dense reward view on aligning text-to-image diffusion with preference","author":"Yang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref50","article-title":"Instructvideo: Instructing video diffusion models with human feedback","author":"Yuan","year":"2023","journal-title":"arXiv preprint"}],"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\/11446059.pdf?arnumber=11446059","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:05:48Z","timestamp":1777611948000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11446059\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":50,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01679","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}