Deprecated: The each() function is deprecated. This message will be suppressed on further calls in /home/zhenxiangba/zhenxiangba.com/public_html/phproxy-improved-master/index.php on line 456 Paper page - Gaussian Mixture Flow Matching Models
arXiv] [GitHub]\n","updatedAt":"2025-04-08T02:43:17.998Z","author":{"_id":"638067fcb334960c987fbeda","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/638067fcb334960c987fbeda/63ZgNyCXQjQLhRd6poRK-.png","fullname":"Hansheng Chen","name":"Lakonik","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":36,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7727221250534058},"editors":["Lakonik"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/638067fcb334960c987fbeda/63ZgNyCXQjQLhRd6poRK-.png"],"reactions":[{"reaction":"👍","users":["search-facility"],"count":1}],"isReport":false}},{"id":"67f7207511f2381b46c78781","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false},"createdAt":"2025-04-10T01:35:49.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [CFG-Zero*: Improved Classifier-Free Guidance for Flow Matching Models](https://huggingface.co/papers/2503.18886) (2025)\n* [Diffusion Models without Classifier-free Guidance](https://huggingface.co/papers/2502.12154) (2025)\n* [FlowDPS: Flow-Driven Posterior Sampling for Inverse Problems](https://huggingface.co/papers/2503.08136) (2025)\n* [Inductive Moment Matching](https://huggingface.co/papers/2503.07565) (2025)\n* [Inference-Time Scaling for Flow Models via Stochastic Generation and Rollover Budget Forcing](https://huggingface.co/papers/2503.19385) (2025)\n* [Towards Hierarchical Rectified Flow](https://huggingface.co/papers/2502.17436) (2025)\n* [Enhanced Diffusion Sampling via Extrapolation with Multiple ODE Solutions](https://huggingface.co/papers/2504.01855) (2025)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
\n
The following papers were recommended by the Semantic Scholar API
Please give a thumbs up to this comment if you found it helpful!
\n
If you want recommendations for any Paper on Hugging Face checkout this Space
\n
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend
\n","updatedAt":"2025-04-10T01:35:49.112Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6847865581512451},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2504.05304","authors":[{"_id":"67f48c11412c65a9d4e3e552","user":{"_id":"638067fcb334960c987fbeda","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/638067fcb334960c987fbeda/63ZgNyCXQjQLhRd6poRK-.png","isPro":false,"fullname":"Hansheng Chen","user":"Lakonik","type":"user"},"name":"Hansheng Chen","status":"claimed_verified","statusLastChangedAt":"2025-04-08T06:51:39.675Z","hidden":false},{"_id":"67f48c11412c65a9d4e3e553","name":"Kai Zhang","hidden":false},{"_id":"67f48c11412c65a9d4e3e554","name":"Hao Tan","hidden":false},{"_id":"67f48c11412c65a9d4e3e555","user":{"_id":"65565b1b8616d8d740abb976","avatarUrl":"/avatars/c245b19ed05292bf1bf4a9d8cf2a409c.svg","isPro":false,"fullname":"Zexiang Xu","user":"zexiangxu","type":"user"},"name":"Zexiang Xu","status":"admin_assigned","statusLastChangedAt":"2025-04-08T10:38:32.849Z","hidden":false},{"_id":"67f48c11412c65a9d4e3e556","name":"Fujun Luan","hidden":false},{"_id":"67f48c11412c65a9d4e3e557","name":"Leonidas Guibas","hidden":false},{"_id":"67f48c11412c65a9d4e3e558","user":{"_id":"6694e583ac96ca2c17131505","avatarUrl":"/avatars/6e7a31f257e36cf301da6f879dc0a122.svg","isPro":false,"fullname":"Gordon Wetzstein","user":"wetzste1","type":"user"},"name":"Gordon Wetzstein","status":"admin_assigned","statusLastChangedAt":"2025-04-08T10:38:08.838Z","hidden":false},{"_id":"67f48c11412c65a9d4e3e559","user":{"_id":"66303560597dc7c535168f07","avatarUrl":"/avatars/a69d52a4a198c6c8451c1d137031bf07.svg","isPro":false,"fullname":"Sai Bi","user":"saibi","type":"user"},"name":"Sai Bi","status":"admin_assigned","statusLastChangedAt":"2025-04-08T10:38:03.075Z","hidden":false}],"publishedAt":"2025-04-07T17:59:42.000Z","submittedOnDailyAt":"2025-04-08T01:13:17.992Z","title":"Gaussian Mixture Flow Matching Models","submittedOnDailyBy":{"_id":"638067fcb334960c987fbeda","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/638067fcb334960c987fbeda/63ZgNyCXQjQLhRd6poRK-.png","isPro":false,"fullname":"Hansheng Chen","user":"Lakonik","type":"user"},"summary":"Diffusion models approximate the denoising distribution as a Gaussian and\npredict its mean, whereas flow matching models reparameterize the Gaussian mean\nas flow velocity. However, they underperform in few-step sampling due to\ndiscretization error and tend to produce over-saturated colors under\nclassifier-free guidance (CFG). To address these limitations, we propose a\nnovel Gaussian mixture flow matching (GMFlow) model: instead of predicting the\nmean, GMFlow predicts dynamic Gaussian mixture (GM) parameters to capture a\nmulti-modal flow velocity distribution, which can be learned with a KL\ndivergence loss. We demonstrate that GMFlow generalizes previous diffusion and\nflow matching models where a single Gaussian is learned with an L_2 denoising\nloss. For inference, we derive GM-SDE/ODE solvers that leverage analytic\ndenoising distributions and velocity fields for precise few-step sampling.\nFurthermore, we introduce a novel probabilistic guidance scheme that mitigates\nthe over-saturation issues of CFG and improves image generation quality.\nExtensive experiments demonstrate that GMFlow consistently outperforms flow\nmatching baselines in generation quality, achieving a Precision of 0.942 with\nonly 6 sampling steps on ImageNet 256times256.","upvotes":11,"discussionId":"67f48c13412c65a9d4e3e5d7","githubRepo":"https://github.com/Lakonik/GMFlow","githubRepoAddedBy":"user","ai_summary":"A Gaussian mixture flow matching model addresses few-step sampling errors and over-saturation in diffusion and flow matching models, improving image generation quality.","ai_keywords":["diffusion models","flow matching models","Gaussian distribution","flow velocity","dynamic Gaussian mixture","KL divergence loss","GM-SDE/ODE solvers","classifier-free guidance","GMFlow"],"githubStars":180},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6351e5bb3734c6e8a5c1bec1","avatarUrl":"/avatars/a784a51b369b197398575c3afbd5ceab.svg","isPro":false,"fullname":"Han-Bit Kang","user":"hbkang","type":"user"},{"_id":"6342796a0875f2c99cfd313b","avatarUrl":"/avatars/98575092404c4197b20c929a6499a015.svg","isPro":false,"fullname":"Yuseung \"Phillip\" Lee","user":"phillipinseoul","type":"user"},{"_id":"630652803aed65d34e98eee3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/630652803aed65d34e98eee3/XG_PuVFA6ziGQZd3UUZSF.jpeg","isPro":false,"fullname":"Nicolas Dufour","user":"nicolas-dufour","type":"user"},{"_id":"6761500fe5d10c2b311f8c3b","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/d1oji8SjgngoSvD8K44YS.png","isPro":false,"fullname":"Julius Duin","user":"duinamit","type":"user"},{"_id":"65454d7c117ecae648892170","avatarUrl":"/avatars/83a7091a24bf86801176ca85234b417a.svg","isPro":false,"fullname":"Yusuf Dalva","user":"ydalva","type":"user"},{"_id":"66b57c77778c98d29446c8ec","avatarUrl":"/avatars/c176bb7c072f3093f6a0786c87d384d8.svg","isPro":false,"fullname":"Taekyung Ki","user":"taekyungki","type":"user"},{"_id":"651c240a37fecec1fe96c60b","avatarUrl":"/avatars/5af52af97b7907e138efecac0f20799b.svg","isPro":false,"fullname":"S.F.","user":"search-facility","type":"user"},{"_id":"65a4567e212d6aca9a3e8f5a","avatarUrl":"/avatars/ed944797230b5460381209bf76e4a0e4.svg","isPro":false,"fullname":"Catherine Liu","user":"Liu12uiL","type":"user"},{"_id":"63041b541dd5d3c62486c294","avatarUrl":"/avatars/a5286d562f7b9082730f760e66c3bf29.svg","isPro":true,"fullname":"Ho Kei Cheng","user":"hkchengrex","type":"user"},{"_id":"64513261938967fd069d2340","avatarUrl":"/avatars/e4c3c435f6a4cda57d0e2f16ec1cda6e.svg","isPro":false,"fullname":"sdtana","user":"sdtana","type":"user"},{"_id":"663ccbff3a74a20189d4aa2e","avatarUrl":"/avatars/83a54455e0157480f65c498cd9057cf2.svg","isPro":false,"fullname":"Nguyen Van Thanh","user":"NguyenVanThanhHust","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">
A Gaussian mixture flow matching model addresses few-step sampling errors and over-saturation in diffusion and flow matching models, improving image generation quality.
AI-generated summary
Diffusion models approximate the denoising distribution as a Gaussian and
predict its mean, whereas flow matching models reparameterize the Gaussian mean
as flow velocity. However, they underperform in few-step sampling due to
discretization error and tend to produce over-saturated colors under
classifier-free guidance (CFG). To address these limitations, we propose a
novel Gaussian mixture flow matching (GMFlow) model: instead of predicting the
mean, GMFlow predicts dynamic Gaussian mixture (GM) parameters to capture a
multi-modal flow velocity distribution, which can be learned with a KL
divergence loss. We demonstrate that GMFlow generalizes previous diffusion and
flow matching models where a single Gaussian is learned with an L_2 denoising
loss. For inference, we derive GM-SDE/ODE solvers that leverage analytic
denoising distributions and velocity fields for precise few-step sampling.
Furthermore, we introduce a novel probabilistic guidance scheme that mitigates
the over-saturation issues of CFG and improves image generation quality.
Extensive experiments demonstrate that GMFlow consistently outperforms flow
matching baselines in generation quality, achieving a Precision of 0.942 with
only 6 sampling steps on ImageNet 256times256.
GMFlow is an extension of diffusion/flow matching models.
Gaussian Mixture Output: GMFlow expands the network's output layer to predict a Gaussian Mixture (GM) distribution of flow velocity. Standard diffusion/flow matching models are special cases of GMFlow with a single Gaussian component.
Precise Few-Step Sampling: GMFlow introduces novel GM-SDE and GM-ODE solvers that leverage analytic denoising distributions and velocity fields for precise few-step sampling.
Improved Classifier-Free Guidance (CFG): GMFlow introduces a probabilistic guidance scheme that mitigates the over-saturation issues of CFG and improves image generation quality.
Efficiency: GMFlow maintains similar training and inference costs to standard diffusion/flow matching models.