thread summarizing paper on X:<br><a href=\"https://x.com/ethanCaballero/status/2059686905105563907\" rel=\"nofollow\">https://x.com/ethanCaballero/status/2059686905105563907</a></p>\n","updatedAt":"2026-06-02T14:40:03.752Z","author":{"_id":"6a1e715024826072a7a66495","avatarUrl":"/avatars/64d3e15abe73b6b72e326b7643c97324.svg","fullname":"Ethan Caballero","name":"ethan-caballero","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6189423203468323},"editors":["ethan-caballero"],"editorAvatarUrls":["/avatars/64d3e15abe73b6b72e326b7643c97324.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.26248","authors":[{"_id":"6a1e7267808ddbc3c7d43eac","name":"Ethan Caballero","hidden":false},{"_id":"6a1e7267808ddbc3c7d43ead","name":"Priyank Jaini","hidden":false},{"_id":"6a1e7267808ddbc3c7d43eae","name":"David Krueger","hidden":false},{"_id":"6a1e7267808ddbc3c7d43eaf","name":"Irina Rish","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/6a1e715024826072a7a66495/QC6uLirK1vSG1GKCx7EzE.mp4"],"publishedAt":"2026-05-25T00:00:00.000Z","submittedOnDailyAt":"2026-06-02T00:00:00.000Z","title":"Unified Neural Scaling Laws","submittedOnDailyBy":{"_id":"6a1e715024826072a7a66495","avatarUrl":"/avatars/64d3e15abe73b6b72e326b7643c97324.svg","isPro":false,"fullname":"Ethan Caballero","user":"ethan-caballero","type":"user","name":"ethan-caballero"},"summary":"We present a functional form (that we refer to as a Unified Neural Scaling Law (UNSL)) that accurately models and extrapolates the scaling behaviors of deep neural networks as multiple dimensions all vary simultaneously (i.e. how the evaluation metric of interest varies as one simultaneously varies the number of model parameters, training dataset size, number of training steps, number of inference steps, amount of compute, and various hyperparameters) for various architectures and for each of various tasks within a varied set of upstream and downstream tasks. This set includes large-scale vision, language, math, and reinforcement learning. When compared to other functional forms for neural scaling, this functional form yields extrapolations of scaling behavior that are considerably more accurate on this set.","upvotes":2,"discussionId":"6a1e7267808ddbc3c7d43eb0","ai_summary":"A Unified Neural Scaling Law is presented that accurately models and extrapolates deep neural network scaling behaviors across multiple simultaneous dimensions including parameters, dataset size, training steps, and compute across diverse architectures and tasks.","ai_keywords":["Unified Neural Scaling Law","deep neural networks","scaling behaviors","model parameters","training dataset size","training steps","inference steps","compute","hyperparameters","architectures","upstream tasks","downstream tasks","vision","language","math","reinforcement learning"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"60f6cbb2852126bac698c89e","name":"deepmind","fullname":"Deepmind","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1638956859875-5f1158120c833276f61f1a84.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a1e715024826072a7a66495","avatarUrl":"/avatars/64d3e15abe73b6b72e326b7643c97324.svg","isPro":false,"fullname":"Ethan Caballero","user":"ethan-caballero","type":"user"},{"_id":"680f8fc61e7bfb412c671dd5","avatarUrl":"/avatars/405d70f31e9a4a97a1463f91297a623b.svg","isPro":false,"fullname":"Rishit Dagli","user":"rdagli","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"60f6cbb2852126bac698c89e","name":"deepmind","fullname":"Deepmind","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1638956859875-5f1158120c833276f61f1a84.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.26248.md"}">
Unified Neural Scaling Laws
Abstract
A Unified Neural Scaling Law is presented that accurately models and extrapolates deep neural network scaling behaviors across multiple simultaneous dimensions including parameters, dataset size, training steps, and compute across diverse architectures and tasks.
We present a functional form (that we refer to as a Unified Neural Scaling Law (UNSL)) that accurately models and extrapolates the scaling behaviors of deep neural networks as multiple dimensions all vary simultaneously (i.e. how the evaluation metric of interest varies as one simultaneously varies the number of model parameters, training dataset size, number of training steps, number of inference steps, amount of compute, and various hyperparameters) for various architectures and for each of various tasks within a varied set of upstream and downstream tasks. This set includes large-scale vision, language, math, and reinforcement learning. When compared to other functional forms for neural scaling, this functional form yields extrapolations of scaling behavior that are considerably more accurate on this set.
Community
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2605.26248 in a model README.md to link it from this page.
Cite arxiv.org/abs/2605.26248 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2605.26248 in a Space README.md 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.