Hugging Face Daily Papers · · 4 min read

Safe Few-Step Generation via Velocity Editing

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Project Page: <a href=\"https://uzn36.github.io/VESFlow/\" rel=\"nofollow\">https://uzn36.github.io/VESFlow/</a></p>\n","updatedAt":"2026-06-23T03:34:25.266Z","author":{"_id":"652066649004117947e46ed6","avatarUrl":"/avatars/972c97df6f26d2c3d6ce71ec579984bb.svg","fullname":"Jaehong Yoon","name":"jaehong31","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5055934190750122},"editors":["jaehong31"],"editorAvatarUrls":["/avatars/972c97df6f26d2c3d6ce71ec579984bb.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.23267","authors":[{"_id":"6a39fa9ffdcd3514343bb536","name":"Yujin Choi","hidden":false},{"_id":"6a39fa9ffdcd3514343bb537","user":{"_id":"652066649004117947e46ed6","avatarUrl":"/avatars/972c97df6f26d2c3d6ce71ec579984bb.svg","isPro":false,"fullname":"Jaehong Yoon","user":"jaehong31","type":"user","name":"jaehong31"},"name":"Jaehong Yoon","status":"claimed_verified","statusLastChangedAt":"2026-06-23T13:56:46.305Z","hidden":false}],"publishedAt":"2026-06-22T00:00:00.000Z","submittedOnDailyAt":"2026-06-23T00:00:00.000Z","title":"Safe Few-Step Generation via Velocity Editing","submittedOnDailyBy":{"_id":"652066649004117947e46ed6","avatarUrl":"/avatars/972c97df6f26d2c3d6ce71ec579984bb.svg","isPro":false,"fullname":"Jaehong Yoon","user":"jaehong31","type":"user","name":"jaehong31"},"summary":"Flow matching has recently emerged as a strong paradigm for state-of-the-art text-to-image (T2I) generation, enabling high-quality generation with a small number of sampling steps. As these models are increasingly integrated into real-world applications, ensuring safe and non-sensitive content generation has become a critical requirement. However, adapting safety and concept removal methods to this new generation framework remains an open challenge. Specifically, prior methods largely rely on iterative trajectory steering across a number of denoising steps or on CLIP-centric prompt embedding manipulation. These design assumptions pose fundamental bottlenecks for safety in flow matching-based T2I generation, where limited sampling steps constrain iterative correction and modern context-aware text encoders diminish the effectiveness of embedding-level interventions. In this paper, we propose VESFlow, a training-free safety method tailored to flow matching with extremely few sampling steps. Leveraging the fact that flow matching models learn the marginal velocity, we directly edit the velocity field via a safe-conditional posterior. VESFlow steers the trajectory toward safe outputs while leaving the conditioning prompt unchanged. Building on the observation that VESFlow leaves outputs unchanged under benign prompts, we further introduce a risk score-based filtering that bypasses velocity editing to reduce computational cost while preserving benign prompt generation. Based on this filtering, we propose VESFlow+, a stronger variant of VESFlow that not only edits the velocity toward the safe direction, but also pushes it away from the unsafe direction. Experimental results show that VESFlow+ removes the target concept, reducing the attack success rate by NudeNet to 6.3% on Ring-A-Bell and 6.8% on MMA-Diffusion on the 4-step MeanFlow model, while preserving fidelity on benign prompts.","upvotes":7,"discussionId":"6a39faa0fdcd3514343bb538","projectPage":"https://uzn36.github.io/VESFlow/","ai_summary":"VESFlow is a training-free safety method for flow matching-based text-to-image generation that edits velocity fields to ensure safe output while maintaining prompt integrity.","ai_keywords":["flow matching","text-to-image generation","safety","concept removal","marginal velocity","velocity field","safe-conditional posterior","risk score-based filtering","attack success rate","NudeNet","MeanFlow"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"6371470aafbe42caa5a76208","name":"nanyang-technological-university-singapore","fullname":"Nanyang Technological University Singapore","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/637146c5afbe42caa5a75e1b/sZyHSA1AQaAS4nrGan682.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"652066649004117947e46ed6","avatarUrl":"/avatars/972c97df6f26d2c3d6ce71ec579984bb.svg","isPro":false,"fullname":"Jaehong Yoon","user":"jaehong31","type":"user"},{"_id":"676c04f44464f476aaa53d1c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/k488J1893F3JGwEMvaeuh.png","isPro":false,"fullname":"Chong Xia","user":"xiac24","type":"user"},{"_id":"63946778dda2f4142a3526d0","avatarUrl":"/avatars/9162a41f8c6611bc1258e38475b1d098.svg","isPro":false,"fullname":"Yujin Choi","user":"uzn","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"65f3d7ebc2d214f88485bc7d","avatarUrl":"/avatars/d5724567e69e39ec557045a2da237bdd.svg","isPro":false,"fullname":"RagMaster","user":"ragmaster1","type":"user"},{"_id":"64103f66928400b4164308f0","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64103f66928400b4164308f0/5ZikDcmC1qBCWEP6YJeCx.jpeg","isPro":false,"fullname":"Uday Allu","user":"udayallu","type":"user"},{"_id":"68500910db82ecfed88cd168","avatarUrl":"/avatars/50f337d480e3ba9a697fa3f19ac9d4fa.svg","isPro":false,"fullname":"Akarsh 48","user":"v-tri-ai-eng-yellow","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6371470aafbe42caa5a76208","name":"nanyang-technological-university-singapore","fullname":"Nanyang Technological University Singapore","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/637146c5afbe42caa5a75e1b/sZyHSA1AQaAS4nrGan682.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.23267.md","query":{}}">
Papers
arxiv:2606.23267

Safe Few-Step Generation via Velocity Editing

Published on Jun 22
· Submitted by
Jaehong Yoon
on Jun 23
Authors:
,

Abstract

VESFlow is a training-free safety method for flow matching-based text-to-image generation that edits velocity fields to ensure safe output while maintaining prompt integrity.

Flow matching has recently emerged as a strong paradigm for state-of-the-art text-to-image (T2I) generation, enabling high-quality generation with a small number of sampling steps. As these models are increasingly integrated into real-world applications, ensuring safe and non-sensitive content generation has become a critical requirement. However, adapting safety and concept removal methods to this new generation framework remains an open challenge. Specifically, prior methods largely rely on iterative trajectory steering across a number of denoising steps or on CLIP-centric prompt embedding manipulation. These design assumptions pose fundamental bottlenecks for safety in flow matching-based T2I generation, where limited sampling steps constrain iterative correction and modern context-aware text encoders diminish the effectiveness of embedding-level interventions. In this paper, we propose VESFlow, a training-free safety method tailored to flow matching with extremely few sampling steps. Leveraging the fact that flow matching models learn the marginal velocity, we directly edit the velocity field via a safe-conditional posterior. VESFlow steers the trajectory toward safe outputs while leaving the conditioning prompt unchanged. Building on the observation that VESFlow leaves outputs unchanged under benign prompts, we further introduce a risk score-based filtering that bypasses velocity editing to reduce computational cost while preserving benign prompt generation. Based on this filtering, we propose VESFlow+, a stronger variant of VESFlow that not only edits the velocity toward the safe direction, but also pushes it away from the unsafe direction. Experimental results show that VESFlow+ removes the target concept, reducing the attack success rate by NudeNet to 6.3% on Ring-A-Bell and 6.8% on MMA-Diffusion on the 4-step MeanFlow model, while preserving fidelity on benign prompts.

Community

Paper author Paper submitter about 21 hours ago
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2606.23267
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2606.23267 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2606.23267 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2606.23267 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from Hugging Face Daily Papers