Project Page: <a href=\"https://renderformer.github.io/v2/\" rel=\"nofollow\">https://renderformer.github.io/v2/</a></p>\n<p>A pretrained transformer that turns a sequence of mixed scene primitives into a globally illuminated image.</p>\n","updatedAt":"2026-09-09T20:11:27.232Z","author":{"_id":"60747cbf3ea03830676542b5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60747cbf3ea03830676542b5/wGr1Jzz520JM9nZ-UcLyb.png","fullname":"Chong Zeng","name":"NCJ","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":7,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8230963349342346},"editors":["NCJ"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/60747cbf3ea03830676542b5/wGr1Jzz520JM9nZ-UcLyb.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2609.05738","authors":[{"_id":"6aa10d98d0174964227beeb6","name":"Chong Zeng","hidden":false},{"_id":"6aa10d98d0174964227beeb7","name":"Yue Dong","hidden":false},{"_id":"6aa10d98d0174964227beeb8","name":"Pieter Peers","hidden":false},{"_id":"6aa10d98d0174964227beeb9","name":"Lvmin Zhang","hidden":false},{"_id":"6aa10d98d0174964227beeba","name":"Maneesh Agrawala","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/DeC8RPaZVVsXSf2mLHHjV.mp4","https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/jzYe7oQOWc9xtFw3ieTDr.jpeg","https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/vozWevzYX92A6je3RRqb8.jpeg","https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/JYlcSj0CgdDi0gVrJMcsl.png","https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/tigCPV5ud3aGimAxja9j4.png","https://cdn-uploads.huggingface.co/production/uploads/60747cbf3ea03830676542b5/zf_LRYLJBJ1vbCB9uUTnu.png"],"publishedAt":"2026-09-04T00:00:00.000Z","submittedOnDailyAt":"2026-09-09T00:00:00.000Z","title":"RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives","submittedOnDailyBy":{"_id":"60747cbf3ea03830676542b5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60747cbf3ea03830676542b5/wGr1Jzz520JM9nZ-UcLyb.png","isPro":false,"fullname":"Chong Zeng","user":"NCJ","type":"user","name":"NCJ"},"summary":"We present 'RenderFormer-V2', a unified learned transformer-based neural rendering model, complementary to modern physics-based rendering systems, that can handle diverse light-transport effects such as caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials without per-scene training or specialized code. RenderFormer-V2 models global light transport as a sequence-to-sequence transformation. Following its predecessor, RenderFormer-V2 also employs a two stage process: a view-independent stage that resolves intra-scene primitive to primitive transport, and a view-dependent stage that transforms the internal neural scene representation into image pixels. Different from RenderFormer, our model employs a novel combined windowed-attention and rendering-informed attention sink in the view-independent stage to improve scalability while maintaining render accuracy. To further improve versatility, RenderFormerV2 supports heterogeneous scene primitives, including environment maps and participating media, and it employs a material encoding independent of the underlying surface reflectance model that encodes material appearance via a novel neural embedding. We demonstrate the versatility of RenderFormer-V2 on a variety of scenes and perform an extensive ablation of the improved attention mechanism.","upvotes":1,"discussionId":"6aa10d98d0174964227beebb","projectPage":"https://renderformer.github.io/v2/","githubRepo":"https://github.com/iamNCJ/RenderFormer-Studio","githubRepoAddedBy":"user","ai_summary":"RenderFormer-V2 is a transformer-based neural rendering model that handles diverse light-transport effects via a two-stage sequence-to-sequence architecture with improved attention and heterogeneous scene support.","ai_keywords":["transformer-based neural rendering","sequence-to-sequence transformation","windowed-attention","attention sink","neural scene representation","heterogeneous scene primitives","neural embedding","material encoding","global light transport"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":1,"organization":{"_id":"636025b83605bd411c1889d9","name":"Stanford","fullname":"Stanford AI","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/604d2f473050a33ebb17ef51/Z-vDTyG_6-yZhzfXklqAK.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"60747cbf3ea03830676542b5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/60747cbf3ea03830676542b5/wGr1Jzz520JM9nZ-UcLyb.png","isPro":false,"fullname":"Chong Zeng","user":"NCJ","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"636025b83605bd411c1889d9","name":"Stanford","fullname":"Stanford AI","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/604d2f473050a33ebb17ef51/Z-vDTyG_6-yZhzfXklqAK.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2609/2609.05738.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2609.05738 in a model README.md to link it from this page.

Cite arxiv.org/abs/2609.05738 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2609.05738 in a Space README.md to link it from this page.

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