Hugging Face Daily Papers · · 4 min read

GPTQ-2D: Cubic-Time Two-Sided Adaptive Rounding

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

Adaptive rounding methods such as GPTQ, or equivalently Babai's nearest plane algorithm, round a real matrix to integers under a quadratic metric. They process the entries in a fixed order, one at a time, propagating each rounding error to the entries not yet processed through a triangular feedback matrix. We study the two-sided version of this task, in which fixed nonsingular basis matrices act on both the left and the right of the residual; the familiar one-sided case is the special case of an identity right basis. Vectorizing the matrix turns the two-sided objective into a quadratic metric whose Gram matrix is a Kronecker product, so the one-dimensional algorithm applies verbatim, but takes quartic time in the matrix dimension. We present GPTQ-2D, which produces the identical rounded matrix in cubic time. It rounds the entries anti-diagonal by anti-diagonal; entries on the same anti-diagonal are independent and are rounded in parallel.</p>\n","updatedAt":"2026-08-04T08:44:13.233Z","author":{"_id":"6298d8dab58e71e2ac9e2967","avatarUrl":"/avatars/99fec4ba78ab61c5952f51e6ebf03ffa.svg","fullname":"Jiale Chen","name":"softmax","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9164726138114929},"editors":["softmax"],"editorAvatarUrls":["/avatars/99fec4ba78ab61c5952f51e6ebf03ffa.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.27042","authors":[{"_id":"6a6b2bbcb2106777884ab098","user":{"_id":"6298d8dab58e71e2ac9e2967","avatarUrl":"/avatars/99fec4ba78ab61c5952f51e6ebf03ffa.svg","isPro":false,"fullname":"Jiale Chen","user":"softmax","type":"user","name":"softmax"},"name":"Jiale Chen","status":"claimed_verified","statusLastChangedAt":"2026-07-30T16:45:04.682Z","hidden":false},{"_id":"6a6b2bbcb2106777884ab099","name":"Torsten Hoefler","hidden":false},{"_id":"6a6b2bbcb2106777884ab09a","name":"Dan Alistarh","hidden":false}],"publishedAt":"2026-07-29T00:00:00.000Z","submittedOnDailyAt":"2026-08-04T00:00:00.000Z","title":"GPTQ-2D: Cubic-Time Two-Sided Adaptive Rounding","submittedOnDailyBy":{"_id":"6298d8dab58e71e2ac9e2967","avatarUrl":"/avatars/99fec4ba78ab61c5952f51e6ebf03ffa.svg","isPro":false,"fullname":"Jiale Chen","user":"softmax","type":"user","name":"softmax"},"summary":"Adaptive rounding methods such as GPTQ, or equivalently Babai's nearest plane algorithm, round a real matrix to integers under a quadratic metric. They process the entries in a fixed order, one at a time, propagating each rounding error to the entries not yet processed through a triangular feedback matrix. We study the two-sided version of this task, in which fixed nonsingular basis matrices act on both the left and the right of the residual; the familiar one-sided case is the special case of an identity right basis. Vectorizing the matrix turns the two-sided objective into a quadratic metric whose Gram matrix is a Kronecker product, so the one-dimensional algorithm applies verbatim, but takes quartic time in the matrix dimension. We present GPTQ-2D, which produces the identical rounded matrix in cubic time. It rounds the entries anti-diagonal by anti-diagonal; entries on the same anti-diagonal are independent and are rounded in parallel.","upvotes":16,"discussionId":"6a6b2bbcb2106777884ab09b","organization":{"_id":"64d0ffde9cff738203a50e9b","name":"ISTA-DASLab","fullname":" IST Austria Distributed Algorithms and Systems Lab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/628e0ce4e53bbd334577fcb0/TRPtgtSavYjDJOK3S1I8M.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6298d8dab58e71e2ac9e2967","avatarUrl":"/avatars/99fec4ba78ab61c5952f51e6ebf03ffa.svg","isPro":false,"fullname":"Jiale Chen","user":"softmax","type":"user"},{"_id":"6887b2d4a41d7844decbcbce","avatarUrl":"/avatars/a1a47b1eba0602f6c61f2f8793120d77.svg","isPro":false,"fullname":"Ping Tuo","user":"abbytuo","type":"user"},{"_id":"623753b5eddd7763adc9346a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/623753b5eddd7763adc9346a/290JrM0VdY-H98Ws40ITK.jpeg","isPro":false,"fullname":"Andrei Panferov","user":"BlackSamorez","type":"user"},{"_id":"632a2e325f2ff1958c0103be","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/632a2e325f2ff1958c0103be/Tb0ql9e4LcaFktTK1hzqe.jpeg","isPro":false,"fullname":"Rush Tabesh","user":"soroushtabesh","type":"user"},{"_id":"67e2bc43c007fa6bfd52be07","avatarUrl":"/avatars/e1c68a32f5a1599ea32abcb3b9d92359.svg","isPro":false,"fullname":"Max Kleinegger","user":"mkleinegger","type":"user"},{"_id":"64ef52c2718f94ae8e78a5e7","avatarUrl":"/avatars/d169f4ee62786a3eb4a3fa9d1fec52e9.svg","isPro":false,"fullname":"Alistarh","user":"d-alistarh","type":"user"},{"_id":"607992183a565c15675055a9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/607992183a565c15675055a9/gN748QT4PjTKTTp3qp07O.jpeg","isPro":false,"fullname":"Eugenia Iofinova","user":"jen","type":"user"},{"_id":"67e20a0096307b65a50a6db1","avatarUrl":"/avatars/c3a107f2c529e9a23596aab4b985d8e6.svg","isPro":false,"fullname":"sergei","user":"sergeipa","type":"user"},{"_id":"65f08bd773f2e1e291244d86","avatarUrl":"/avatars/4d66171163094024ebc82bb59859d9c5.svg","isPro":false,"fullname":"Alexandra","user":"alexandraww","type":"user"},{"_id":"69bd0ab90b2fb84daf2cf0ba","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/MQ8ftzLqyMtmOonGK4mRl.png","isPro":false,"fullname":"Zhang Wenxuan","user":"mateor367","type":"user"},{"_id":"6a6a93d2bbca071c7189619a","avatarUrl":"/avatars/2d133bd4af457635809bf72d2c8ca3b4.svg","isPro":false,"fullname":"Mary Hernandez","user":"Frost-Mary","type":"user"},{"_id":"6a6c840261da43cdbfda1203","avatarUrl":"/avatars/3c6b0a713733718b0ccb485d3e2591fb.svg","isPro":false,"fullname":"Edward Davis","user":"roy-4640947","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"64d0ffde9cff738203a50e9b","name":"ISTA-DASLab","fullname":" IST Austria Distributed Algorithms and Systems Lab","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/628e0ce4e53bbd334577fcb0/TRPtgtSavYjDJOK3S1I8M.png"},"query":{}}">
Papers
arxiv:2607.27042

GPTQ-2D: Cubic-Time Two-Sided Adaptive Rounding

Published on Jul 29
· Submitted by
Jiale Chen
on Aug 4
Authors:

Abstract

Adaptive rounding methods such as GPTQ, or equivalently Babai's nearest plane algorithm, round a real matrix to integers under a quadratic metric. They process the entries in a fixed order, one at a time, propagating each rounding error to the entries not yet processed through a triangular feedback matrix. We study the two-sided version of this task, in which fixed nonsingular basis matrices act on both the left and the right of the residual; the familiar one-sided case is the special case of an identity right basis. Vectorizing the matrix turns the two-sided objective into a quadratic metric whose Gram matrix is a Kronecker product, so the one-dimensional algorithm applies verbatim, but takes quartic time in the matrix dimension. We present GPTQ-2D, which produces the identical rounded matrix in cubic time. It rounds the entries anti-diagonal by anti-diagonal; entries on the same anti-diagonal are independent and are rounded in parallel.

Community

Paper author Paper submitter about 4 hours ago

Adaptive rounding methods such as GPTQ, or equivalently Babai's nearest plane algorithm, round a real matrix to integers under a quadratic metric. They process the entries in a fixed order, one at a time, propagating each rounding error to the entries not yet processed through a triangular feedback matrix. We study the two-sided version of this task, in which fixed nonsingular basis matrices act on both the left and the right of the residual; the familiar one-sided case is the special case of an identity right basis. Vectorizing the matrix turns the two-sided objective into a quadratic metric whose Gram matrix is a Kronecker product, so the one-dimensional algorithm applies verbatim, but takes quartic time in the matrix dimension. We present GPTQ-2D, which produces the identical rounded matrix in cubic time. It rounds the entries anti-diagonal by anti-diagonal; entries on the same anti-diagonal are independent and are rounded in parallel.

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

Models citing this paper

No model linking this paper

Cite arxiv.org/abs/2607.27042 in a model README.md to link it from this page.

Datasets citing this paper

No dataset linking this paper

Cite arxiv.org/abs/2607.27042 in a dataset README.md to link it from this page.

Spaces citing this paper

No Space linking this paper

Cite arxiv.org/abs/2607.27042 in a Space README.md to link it from this page.

Collections including this paper

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