The paper introduces Representation Forcing, a method enabling bottleneck-free unified multimodal models by autoregressively predicting visual representation tokens before pixels, removing reliance on frozen external VAEs for image generation.</p>\n","updatedAt":"2026-06-01T02:10:06.690Z","author":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","fullname":"taesiri","name":"taesiri","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":309,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7344756126403809},"editors":["taesiri"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.31604","authors":[{"_id":"6a1ce9e4808ddbc3c7d433ff","user":{"_id":"63ea23b9dedfeebe54d02bdf","avatarUrl":"/avatars/4d9f9a546aa8c63e277161ea700075c4.svg","isPro":false,"fullname":"Yuqing Wang","user":"Epiphqny","type":"user","name":"Epiphqny"},"name":"Yuqing Wang","status":"claimed_verified","statusLastChangedAt":"2026-06-01T09:33:36.812Z","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43400","name":"Zhijie Lin","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43401","name":"Ceyuan Yang","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43402","name":"Yang Zhao","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43403","name":"Fei Xiao","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43404","name":"Hao He","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43405","name":"Qi Zhao","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43406","name":"Zihan Ding","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43407","name":"Fuyun Wang","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43408","name":"Shuai Wang","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d43409","name":"Youliang Zhang","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d4340a","name":"Haoqi Fan","hidden":false},{"_id":"6a1ce9e4808ddbc3c7d4340b","name":"Xihui Liu","hidden":false}],"publishedAt":"2026-05-29T00:00:00.000Z","submittedOnDailyAt":"2026-06-01T00:00:00.000Z","title":"Representation Forcing for Bottleneck-Free Unified Multimodal Models","submittedOnDailyBy":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user","name":"taesiri"},"summary":"Unified multimodal models (UMMs) aim to handle perception and generation in a single model. Yet existing UMMs still rely on a frozen, separately pretrained VAE for image generation, imposing a structural bottleneck. Naively removing it introduces a quality gap, as the model must learn both high-level structure and low-level details from raw pixels. In this paper, we propose Representation Forcing (RF), a technique that closes this gap by making representation prediction a native capability of the model. Concretely, RF forces the decoder to autoregressively predict visual representations as intermediate tokens before pixels; these tokens then stay in context to guide pixel diffusion within the same backbone. By turning representations from perception outputs into generation targets, RF eliminates the need for any external generative latent space. We find that RF benefits both understanding and generation. On image generation, our pixel-space model with RF matches state-of-the-art VAE-based unified models. On image understanding, pixel-space RF generally outperforms its VAE-based variant. Together, these results offer an effective step toward end-to-end, bottleneck-free UMMs.","upvotes":36,"discussionId":"6a1ce9e4808ddbc3c7d4340c","projectPage":"https://yuqingwang1029.github.io/RepresentationForcing/","ai_summary":"Representation Forcing enables unified multimodal models to perform both perception and generation tasks end-to-end without relying on external latent spaces, matching state-of-the-art performance in image generation while improving understanding capabilities.","ai_keywords":["unified multimodal models","visual representations","representation forcing","pixel space","autoregressive prediction","visual tokens","pixel diffusion","end-to-end training","bottleneck-free models"],"organization":{"_id":"67d1140985ea0644e2f14b99","name":"ByteDance-Seed","fullname":"ByteDance Seed","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/flkDUqd_YEuFsjeNET3r-.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"63ea23b9dedfeebe54d02bdf","avatarUrl":"/avatars/4d9f9a546aa8c63e277161ea700075c4.svg","isPro":false,"fullname":"Yuqing Wang","user":"Epiphqny","type":"user"},{"_id":"66615c855fd9d736e670e0a9","avatarUrl":"/avatars/0ff3127b513552432a7c651e21d7f283.svg","isPro":false,"fullname":"wangshuai","user":"wangsssssss","type":"user"},{"_id":"65d851096769b3a9c9376134","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65d851096769b3a9c9376134/j_d2RSa-3rRmSAmRICSk0.jpeg","isPro":false,"fullname":"ZehongMa","user":"zehongma","type":"user"},{"_id":"657c03a5538666d04cd47461","avatarUrl":"/avatars/00a7686e08207915ade05b52a84d8e26.svg","isPro":false,"fullname":"Chen Yang","user":"Y-Sisyphus","type":"user"},{"_id":"66ff5d997d722f0879fd95a3","avatarUrl":"/avatars/cadcd49cbf6c56896dc5b570bc0bc49b.svg","isPro":false,"fullname":"tao","user":"Tonystock","type":"user"},{"_id":"66ffe645a771e3600e57d77f","avatarUrl":"/avatars/7a5777533e4d610ce157fd64a187a954.svg","isPro":false,"fullname":"楼天成","user":"ohyeahhhhhh","type":"user"},{"_id":"665f059a8947302aa2c63afe","avatarUrl":"/avatars/50f560285946532321a0bd526494148d.svg","isPro":false,"fullname":"hanlin wang","user":"hlwang06","type":"user"},{"_id":"63f1d16fbe95ed4c9a9418fe","avatarUrl":"/avatars/a1bdfa97323693808f2f16ec74698ed3.svg","isPro":false,"fullname":"Yang Yue","user":"yueyang2000","type":"user"},{"_id":"6319b118615c77c25d659279","avatarUrl":"/avatars/a11d064945056134aec4b146edbd1f9a.svg","isPro":false,"fullname":"Skirrey","user":"Skirrey","type":"user"},{"_id":"627d2723401f42c57b6b7c0c","avatarUrl":"/avatars/6ff754e56aaee63d8572881a6a966171.svg","isPro":false,"fullname":"Yuanzhi Zhu","user":"Yuanzhi","type":"user"},{"_id":"670a3bc3ada59c956f18cc17","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/670a3bc3ada59c956f18cc17/57oBwS0V9m9SImYHtDb5f.jpeg","isPro":false,"fullname":"SII-sqs","user":"groundhogLLM","type":"user"},{"_id":"66320391060ab1f666819e9c","avatarUrl":"/avatars/42856c3066f49ff183b10b5239daa6ab.svg","isPro":false,"fullname":"Ruopeng Gao","user":"HELLORPG","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":2,"organization":{"_id":"67d1140985ea0644e2f14b99","name":"ByteDance-Seed","fullname":"ByteDance Seed","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/6535c9e88bde2fae19b6fb25/flkDUqd_YEuFsjeNET3r-.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.31604.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2605.31604 in a model README.md to link it from this page.

Cite arxiv.org/abs/2605.31604 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2605.31604 in a Space README.md to link it from this page.