All code, datasets and model weights are publicly available.<br><a href=\"https://huggingface.co/collections/P1n3/sdg\">https://huggingface.co/collections/P1n3/sdg</a><br>github：<a href=\"https://github.com/nianbai006/SDG\" rel=\"nofollow\">https://github.com/nianbai006/SDG</a></p>\n","updatedAt":"2026-06-12T09:50:02.809Z","author":{"_id":"670e63cd41b894977b30c244","avatarUrl":"/avatars/8cef866d528cf0e3a0d3b45f319b94aa.svg","fullname":"Huaisong Zhang","name":"P1n3","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":2,"identifiedLanguage":{"language":"en","probability":0.8711233139038086},"editors":["P1n3"],"editorAvatarUrls":["/avatars/8cef866d528cf0e3a0d3b45f319b94aa.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.06113","authors":[{"_id":"6a28e18be7d78ea7587e54f2","user":{"_id":"670e63cd41b894977b30c244","avatarUrl":"/avatars/8cef866d528cf0e3a0d3b45f319b94aa.svg","isPro":false,"fullname":"Huaisong Zhang","user":"P1n3","type":"user","name":"P1n3"},"name":"Huaisong Zhang","status":"claimed_verified","statusLastChangedAt":"2026-06-12T07:48:22.082Z","hidden":false},{"_id":"6a28e18be7d78ea7587e54f3","name":"Hao Yu","hidden":false},{"_id":"6a28e18be7d78ea7587e54f4","name":"Yuxuan Zhang","hidden":false},{"_id":"6a28e18be7d78ea7587e54f5","name":"Jiahe Wang","hidden":false},{"_id":"6a28e18be7d78ea7587e54f6","name":"Xinrui Chen","hidden":false},{"_id":"6a28e18be7d78ea7587e54f7","name":"Haoxiang Cao","hidden":false},{"_id":"6a28e18be7d78ea7587e54f8","name":"Feng Lu","hidden":false},{"_id":"6a28e18be7d78ea7587e54f9","name":"Wendong Zhang","hidden":false},{"_id":"6a28e18be7d78ea7587e54fa","name":"Changqian Yu","hidden":false},{"_id":"6a28e18be7d78ea7587e54fb","name":"Chun Yuan","hidden":false}],"publishedAt":"2026-06-04T00:00:00.000Z","submittedOnDailyAt":"2026-06-12T00:00:00.000Z","title":"Where, What, Why, and Importance: Structured Defect Grounding for Text-to-Image Feedback","submittedOnDailyBy":{"_id":"670e63cd41b894977b30c244","avatarUrl":"/avatars/8cef866d528cf0e3a0d3b45f319b94aa.svg","isPro":false,"fullname":"Huaisong Zhang","user":"P1n3","type":"user","name":"P1n3"},"summary":"Despite generating increasingly photorealistic images, text-to-image (T2I) models still exhibit localized, subtle, and structurally complex failures. Diagnosing these failures requires instance-level feedback that answers where a defect occurs, what type it is, why it is defective, and its importance to overall image quality. While recent dense-feedback methods move beyond scalar supervision, their heatmap-centric representations still formulate diagnosis as pixel-field regression, making it difficult to localize variable-cardinality defects and bind semantic reasons to individual failures. To address this representation bottleneck, we propose Structured Defect Grounding (SDG), which casts T2I diagnosis as structured set prediction by modeling each defect as a (location, type, reason, importance) tuple. To make this formulation trainable and measurable, we introduce SDG-30K, a 30K-image dataset with box-grounded annotations across four modern T2I generators, together with a dedicated evaluation protocol, SDG-Eval. Building on this structured representation, we further present a diagnosis-to-alignment framework in which a Vision-Language Model (VLM) serves as the SDG detector, and BoxFlow-GRPO converts predicted defect sets into box-derived, importance-weighted spatial rewards for diffusion model alignment. Extensive experiments show that our SDG detector outperforms leading proprietary VLMs on structured defect grounding, while SDG-guided rewards consistently improve T2I alignment and support localized image refinement. These results establish SDG as a unified, instance-level interface for diagnosing, evaluating, and enhancing modern generative models.","upvotes":12,"discussionId":"6a28e18be7d78ea7587e54fc","projectPage":"https://huggingface.co/collections/P1n3/sdg","githubRepo":"https://github.com/nianbai006/SDG","githubRepoAddedBy":"user","ai_summary":"Structured Defect Grounding (SDG) addresses limitations in text-to-image model diagnosis by modeling defects as structured sets and using vision-language models for detection and reward-based alignment.","ai_keywords":["text-to-image models","structured set prediction","defect grounding","Vision-Language Model","diffusion model alignment","box-derived rewards","spatial rewards","instance-level feedback","dense-feedback methods","pixel-field regression","variable-cardinality defects"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":3,"organization":{"_id":"665f02ce9f9e5b38d0a256a8","name":"Kwai-Kolors","fullname":"Kolors Team, Kuaishou Technology","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f0babaef9cc6810cec02ff/sVnELkcfVo5kxg5308rkr.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"670e63cd41b894977b30c244","avatarUrl":"/avatars/8cef866d528cf0e3a0d3b45f319b94aa.svg","isPro":false,"fullname":"Huaisong Zhang","user":"P1n3","type":"user"},{"_id":"6460ffcc1db65f878514f685","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/aEehC1DK6AI8PAKPa9UOv.png","isPro":false,"fullname":"xiaoyuan","user":"xiaoyuaner","type":"user"},{"_id":"68ce3eb08db3012f60069099","avatarUrl":"/avatars/5e02fb7fa7e3dfa2a3d4deccc35a44da.svg","isPro":false,"fullname":"Jinglin Wang","user":"wang-0538","type":"user"},{"_id":"64f19422b4344f592fb59f28","avatarUrl":"/avatars/d8c2c41d01367cd08a5c6d9979e2d4b2.svg","isPro":false,"fullname":"hhhhhhhhhh","user":"zk-guo","type":"user"},{"_id":"669205f1ccca14aa8f13f770","avatarUrl":"/avatars/11ce274e93345fe3790ac9fa687e2bcb.svg","isPro":false,"fullname":"Hao Yu","user":"Longin-Yu","type":"user"},{"_id":"63b91ec5f270ad02f61327b0","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1673076377130-noauth.jpeg","isPro":false,"fullname":"Yunhui Liu","user":"Cloudy1225","type":"user"},{"_id":"691e5f168f82cd99d66df74d","avatarUrl":"/avatars/ca614ef49da66cdb3f1d5ac07118ed9f.svg","isPro":true,"fullname":"Perry the Platypus","user":"AgPerry","type":"user"},{"_id":"63b908d0e3c78740d8e950d0","avatarUrl":"/avatars/3e80075e92aebdfea712f70b00d5ec7d.svg","isPro":false,"fullname":"Yuxuan Zhang","user":"Reacherx","type":"user"},{"_id":"68a1734a3183dbfc402fbd45","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/jWrxuNNzjDh2boJqR_gEU.png","isPro":false,"fullname":"Miller","user":"Bengioalin","type":"user"},{"_id":"674599197a49660f7f611612","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/25CXDJh4ccCiFltqN4NW5.jpeg","isPro":false,"fullname":"SujieHu","user":"SujieHu","type":"user"},{"_id":"6691ea89286045211d622cce","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6691ea89286045211d622cce/xMxx4nCNGlL1dQT0i0uSR.jpeg","isPro":false,"fullname":"Vegetabot","user":"Vegetabot","type":"user"},{"_id":"681d7bc1e36b83d25c8c4581","avatarUrl":"/avatars/9af72c1b7446ec9ecd18ea0c164e28d8.svg","isPro":false,"fullname":"Happy AI","user":"happyiai","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"665f02ce9f9e5b38d0a256a8","name":"Kwai-Kolors","fullname":"Kolors Team, Kuaishou Technology","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/62f0babaef9cc6810cec02ff/sVnELkcfVo5kxg5308rkr.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.06113.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.06113 in a Space README.md to link it from this page.

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