Pith.

Computational Complexity · arXiv:2610.30054 v1 · submitted 21 Oct 2026 · 15 references

Multiplying matrices via polynomials: exposition and barriers

Alimzhan Amanov, Maxim van den Berg, Jeroen Zuiddam

The paper says

“We give an exposition of OpenAI's recent proof that the exponent ωℂ of matrix multiplication over the complex numbers satisfies ωℂ ≤ 9/4. Our version of the proof gives ω𝔽 ≤ 9/4 over every field 𝔽.”

more of the abstract

We also treat the rectangular matrix multiplication exponent ω𝔽(a,b,c). Here we have ω𝔽(1,k,1) = 2 for 0 ≤ k ≤ 1/2 and ω𝔽(1,k,1) ≤ 1 + k + 1/(4k) for k ≥ 1/2, and, for real numbers a ≥ b ≥ c ≥ 0, ω𝔽(a,b,c) = a + b if c(a + b) ≤ ab and ω𝔽(a,b,c) ≤ (a + b + c)/2 + (ab/c + ac/b + bc/a)/4 otherwise. Our exposition is in the language of degeneration inequalities among polynomial multiplication tensors and Strassen's asymptotic spectrum duality. We moreover prove an operational barrier (which can be thought of as coming from an operational asymptotic spectrum) showing that the current analysis of all these upper bounds on ω𝔽 and ω𝔽(a,b,c) is optimal and that in fact the current degeneration ingredients cannot improve the bounds.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 22 Oct 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

The passes disagreed. Reader: holds up. Skeptic: holds up only with conditions. Referee: major revision.

Solid all-fields square proof and a genuinely new barrier, but the rectangular upper bounds rest on a false inequality in Proposition 4.14 and are unproven as written. This line is wrong

This review says the paper's mathematics is wrong. Pith did not check that against the paper's source.

2 major objections · 3 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 15 references say.

The paper says, at Section 4.5, proof of Proposition 4.14

The review did not quote this passage.

Pith says: Objection 1 of 2 · major

The claimed upper bound G_n ≤ (nS)^{2S} is false as written. The proof derives it from φπ(a,b) < a+b, but here a = n^{x_{π(1)}} and b = n^{x_{π(2)}}, so a+b is exponential in n and is not bounded by nS.

more of objection 1

A concrete witness is φ = R^{(1)} with (x1,x2,x3) = (2,1,1): the factors φπ(n^{x_{π(1)}}, n^{x_{π(2)}}) take values of order n, n^2, or n^2+n-1 according to which original flattening rank becomes the first leg, and the product G_n has order at least n^{10}, whereas (nS)^{2S} = (4n)^8 has order n^8. This step is load-bearing: Proposition 4.14 is the only ingredient used to prove Theorem 1.2 and Corollary 4.17, and the upper-bound half of Theorem 5.11 invokes it as well. The square bound Theorem 1.1 via Proposition 4.12 is independent and is not affected.

This review says the paper's mathematics is wrong. Pith did not check that against the paper's source.

The paper says, at Section 5.3, completeness of the formal preorder

The review did not quote this passage.

Pith says: Objection 2 of 2 · major

Theorem 5.4 is a statement about the formal preorder ⪯, but the manuscript proves only the forward direction T⪯S ⇒ T⊴S. The converse containment, namely that every derivation built from ingredients (D0)-(D4) and rules (C1)-(C4), including applications inside partial direct sums, yields a relation in ⪯, is asserted but not shown by induction.

more of objection 2

Because (D3) and (D4) are allowed inside partial direct sums and ⪯ is not globally compatible with ⊕1, this completeness direction is not a purely notational point. Since the introductory reading of the barrier is that no combination of the ingredients can improve the bounds, a short induction over derivation trees should be supplied.

3 minor objections

The paper says, at Section 4.5, paragraph after Lemma 4.15

Pith says: minor objection 1 of 3

There is a typo: "the the maximum" should read "the maximum".

The paper says, at Section 5.8, proof of Theorem 5.11

Pith says: minor objection 2 of 3

The proof of Theorem 5.11 is explicitly labeled a sketch. Since this theorem is presented as a headline characterization of Δ⪯, the upper-bound half should be written out completely, especially because it depends on Proposition 4.14.

show 1 more

The paper says, at Section 5.4, definition of F*_θ

Pith says: minor objection 3 of 3

It would help the reader to state explicitly in Definition 5.5 that the value of F*_θ at a generator is defined only for admissible triples of positive integers, and that the extension to F uses the admissible triples appearing in (D2)-(D4); the current presentation is correct but slightly implicit.

What would settle it

Recompute the disputed step for φ = R^{(1)} and x = (2,1,1), using R^{(1)}(P_{a,b}) = a, R^{(2)}(P_{a,b}) = b, R^{(3)}(P_{a,b}) = a+b−1. This gives G_n ∼ c n^{12} while the proof's asserted upper bound is (nS)^{2S} = (4n)^8. If the computation confirms these asymptotic orders, the proof of Proposition 4.14 is invalid, and since no alternative derivation of Theorem 1.2 appears in the paper, the rectangular claims are currently unsupported.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 22 Oct 2026

  1. Reader holds up confidence high
  2. Skeptic disagrees: says it holds up only with conditions 8 min 23 s · 51,730 tokens read, 60,747 written Proposition 4.14 rests on the false bound G_n ≤ (nS)^{2S}; the rectangular upper bounds are not proved as written.
  3. Editor wrote the verdict above 1 min 23 s · 53,433 tokens read, 998 written
  4. Referee major revision: 2 major objections, 3 minor 2 min 25 s · 52,128 tokens read, 23,561 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

The short version: the all-fields square bound omega_F <= 9/4 is proved cleanly, and the operational barrier in Section 5 is a real, self-contained contribution. But the rectangular part has a load-bearing flaw. Proposition 4.14 is not proved as written, and since Theorem 1.2 and Corollary 4.17 rest on it, those results are unproven in this manuscript. The stress-test note is correct, and the reader's ACCEPT was too fast.

The bad step is the bound Gn <= (nS)2S in the proof of Proposition 4.14. The paper derives it from phi_pi(a,b) <= a+b-1 together with sum_pi xpi(3) = 2S, which treats nx + ny as if it were bounded by n(x+y). It is not. Take x = (2,1,1) and phi = R^(3): each of the six factors in Gn has order n2, so Gn ~ c n12, while (nS)2S = (4n)8 ~ n8. The displayed bound fails by a factor of order n4, and the subsequent step ((n+1)/2)8B/S <= (nS)2S collapses with it. Since Proposition 4.14 is the sole route to Theorem 1.2, the rectangular bounds and alpha_F >= 1/2 are not established here. They may well be true -- Uznanski's paper is cited for them, and a Lean version is mentioned in the concurrent-work section -- but this manuscript does not prove them.

What is actually new and good: the field-general proof of the square bound, with the positive-characteristic issue handled honestly in Section 3, and above all the F*_theta dual certificate for the formal preorder in Section 5. The barrier is stronger than the concurrent square-only GitHub results, and it is independent of the broken Proposition 4.14. The modelling caveat -- the barrier applies to the formal preorder, not to all of degeneration -- is stated plainly by the authors, so it is an honest premise, not a hidden flaw. The field-generality claim is also corroborated by the listed concurrent Lean formalizations, which I treat as real evidence.

Minor items: Theorem 5.11 is explicitly a sketch and Remark 5.12 omits a proof; both are acknowledged, so minor. The reader's free-parameter and circularity scores were right, and the weakest-assumption concern is the authors' own stated premise.

For algebraic complexity readers the barrier is the takeaway; anyone citing the rectangular bounds should wait for the fix. If I had to decide today: not acceptable as is, but clearly worth referee time. The fix is either to repair Proposition 4.14 or to restructure Theorem 1.2 as an application of Uznanski's proof. Either way, the barrier and the square bound justify a serious review.

The strongest objection

In the proof of Proposition 4.14 the paper defines S = x1+x2+x3 and G_n = ∏_{π∈S3} φπ(n^{x_{π(1)}}, n^{x_{π(2)}})^{x_{π(3)}}, then asserts G_n ≤ (nS)^{2S} using φπ(a,b) ≤ a+b−1 and ∑π x_{π(3)} = 2S. The implication is invalid: here a = n^{x_{π(1)}} and b = n^{x_{π(2)}} are exponential in n, and n^{x}+n^{y} is not bounded by nS. A concrete witness is φ = R^{(1)}, the first flattening rank spectral point, with x = (2,1,1), so S = 4. Using R^{(1)}(P_{a,b}) = a, R^{(2)}(P_{a,b}) = b, and R^{(3)}(P_{a,b}) = a+b−1, each of the six factors in G_n has order n^2, hence G_n ∼ c n^{12}, whereas (nS)^{2S} = (4n)^8 ∼ n^8. The displayed bound fails by a factor of order n^4. This step is load-bearing: Proposition 4.14 is the sole route in the manuscript to Theorem 1.2 and Corollary 4.17, and Theorem 5.11 also invokes it. The square bound ω_F ≤ 9/4 via Proposition 4.12 is a separate argument and is not affected by this flaw.

I read the paper in good faith and checked the main square-multiplication argument and the F*θ barrier construction; those parts appear coherent. However, the proof of Proposition 4.14, which is the load-bearing step for the rectangular upper bounds in Theorem 1.2 and Corollary 4.17, contains a concrete false inequality: G_n ≤ (nS)^{2S} fails for the valid spectral point R^{(1)} already at x = (2,1,1). The reader's weakest assumption concerned the representability of all degeneration combinations in the formal preorder; my concern is different and more direct. Because no other proof of Proposition 4.14 is supplied, the manuscript does not currently establish the rectangular claims it advertises. The ω_F ≤ 9/4 square bound and the operational barrier Theorem 5.4 are separate and may well stand; a repair of Proposition 4.14 could restore the rectangular results. Hence I recommend a conditional verdict pending that repair rather than unconditional acceptance.

The referee's report

If the rectangular results were fully supported, this would be a substantial paper: it removes characteristic-zero assumptions from the new 9/4 bound, gives α_F ≥ 1/2 over every field, improves known rectangular bounds, and gives a barrier argument going beyond the irreversibility barrier. The exposition is clear, and the field-general degenerations in Section 3 and the monotonicity analysis of the functions F*_θ in Section 5 are valuable and appear complete. However, the proof of Proposition 4.14, which is the sole route to Theorem 1.2 and Corollary 4.17, contains a false estimate as written. The square bound ω_F ≤ 9/4 via Proposition 4.12 is a separate argument and is not affected by this flaw.

Pith's reading of the paper (a machine's summary, not the authors')

Matrix multiplication is the workhorse behind many computations, and its exponent is the number such that multiplying n by n matrices costs about n to that power operations. For decades the best known exponent stayed above 2.37, and a recent OpenAI proof brought the complex-number exponent down to 9/4. This paper re-proves that result in a way that works over every field, not just the complex numbers. The main objects are polynomial multiplication tensors, which describe multiplying polynomials and are much cheaper than matrices. The proof uses Strassen's asymptotic spectrum, an infinite family of cost functions that are monotone under reductions, to convert degenerations between tensors into inequalities and then optimizes those inequalities. The same machinery gives rectangular bounds, including the fact that multiplying n^2 by n matrices with n by n^2 matrices costs at most O(n^{4+epsilon}) operations. The paper's own new result is an operational barrier. It defines a formal system containing exactly the degeneration ingredients and rules used by the proof, and it constructs a dual certificate function that is monotone in that system and takes the value n^{9/4} at the n by n by n matrix multiplication tensor. Therefore any relation in the formal system that would imply an exponent below 9/4 is impossible, and the analogous statement holds for the rectangular curve. The barrier is carefully scoped: it does not claim that no better bound exists, only that the current polynomial-multiplication ingredients and every combination of them are exhausted.

Core claim. The central load-bearing claim is Theorem 1.1: for every field F, omega_F <= 9/4, together with the rectangular refinement Theorem 1.2 giving omega_F(a,b,c) = a+b when c(a+b) <= ab and the displayed upper bound otherwise. The paper also claims Theorem 5.4: in the formal preorder generated by ingredients (D0)-(D4) and rules (C1)-(C4), every relation <n1,n2,n3> ⪯ <r> forces log r to be at least the Theorem 1.2 bound, and in the square case r >= n^{9/4}; hence no combination of these ingredients improves the bounds.

The premise it rests on. The barrier is proved for a formal preorder ⪯ on a set F of formal expressions (Sections 5.2-5.3). The paper asserts without a separate formalization that every degeneration construction from (D0)-(D4) using (C1)-(C4), including operations inside partial direct sums, is representable as a relation in ⪯. If a valid way of combining the ingredients falls outside this formal system, the barrier would not rule it out. The all-fields upper bounds also inherit Strassen's spectral duality (Theorem 4.2), a standard theorem cited from Strassen that must hold over arbitrary fields.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

We give an exposition of OpenAI's recent proof that the exponent ωℂ of matrix multiplication over the complex numbers satisfies ωℂ ≤ 9/4. Our version of the proof gives ω𝔽 ≤ 9/4 over every field 𝔽. We also treat the rectangular matrix multiplication exponent ω𝔽(a,b,c). Here we have ω𝔽(1,k,1) = 2 for 0 ≤ k ≤ 1/2 and ω𝔽(1,k,1) ≤ 1 + k + 1/(4k) for k ≥ 1/2, and, for real numbers a ≥ b ≥ c ≥ 0, ω𝔽(a,b,c) = a + b if c(a + b) ≤ ab and ω𝔽(a,b,c) ≤ (a + b + c)/2 + (ab/c + ac/b + bc/a)/4 otherwise. Our exposition is in the language of degeneration inequalities among polynomial multiplication tensors and Strassen's asymptotic spectrum duality. We moreover prove an operational barrier (which can be thought of as coming from an operational asymptotic spectrum) showing that the current analysis of all these upper bounds on ω𝔽 and ω𝔽(a,b,c) is optimal and that in fact the current degeneration ingredients cannot improve the bounds.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-10-22

The authors' main claim, as the classifier picked it out 0.68

“We moreover prove an operational barrier (which can be thought of as coming from an operational asymptotic spectrum) showing that the current analysis of all these upper bounds on $\omega_\mathbb{F}$ and $\omega_\mathbb{F}(a,b,c)$ is optimal and that in fact the current degeneration ingredients cannot improve the bounds.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.05 · states limitation 0.79 · human participants 0.01 · controlled comparison 0.01 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.68 · mentions code or data availability 0.02

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
We give an exposition of OpenAI's recent proof that the exponent $\omega_\mathbb{C}$ of matrix multiplication over the complex numbers satisfies $\omega_\mathbb{C}\leq9/4$. contribution 0.81 0.27 0.39 asserted 1.00
Our version of the proof gives $\omega_\mathbb{F}\leq9/4$ over every field $\mathbb{F}$. result 0.92 0.90 0.38 asserted 1.00
We also treat the rectangular matrix multiplication exponent $\omega_\mathbb{F}(a,b,c)$. contribution 0.97 0.72 0.05 asserted 0.81
Here we have $\omega_\mathbb{F}(1,k,1)=2$ for $0\leq k\leq1/2$ and $\omega_\mathbb{F}(1,k,1)\leq1+k+1/(4k)$ for $k\geq1/2$, and, for real numbers $a\geq b\geq c\geq0$, $\omega_\mathbb{F}(a,b,c)=a+b$ if $c(a+b)\leq ab$ and $\omega_\mathbb{F}(a,b,c)\leq(a+b+c)/2+(ab/c+ac/b+bc/a)/4$ otherwise. result 0.99 0.70 0.19 asserted 1.00
Our exposition is in the language of degeneration inequalities among polynomial multiplication tensors and Strassen's asymptotic spectrum duality. method 0.97 0.63 0.04 asserted 1.00
We moreover prove an operational barrier (which can be thought of as coming from an operational asymptotic spectrum) showing that the current analysis of all these upper bounds on $\omega_\mathbb{F}$ and $\omega_\mathbb{F}(a,b,c)$ is optimal and that in fact the current degeneration ingredients cannot improve the bounds. result 0.70 0.95 0.51 asserted 1.00

What kind of paper the classifier reports this is: theoretical result 0.98 · evidence from proof or derivation 1.00

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-22: 1 request, 5,384 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (4)
Figure 1 of the paper
Figure 1. (Left) The upper bound of Theorem 1.2 on ωF(1, b, c) for 1 ≥ b ≥ c ≥ 0 as a blue surface, above the grey plane of the lower bound 1 + b. The black dotted curve c = b/(1 + b) indicates until where upper bound and lower bound match and give ωF(1, b, c) = 1 + b. The orange dotted line c = 0.321334 b shows until where this match was known by previous results from the laser method [WXXZ24]. (Right) The upper bound of Corollary 4.17 on ωF(1, k, 1) for 0 ≤ k ≤ 1, compared with the previously known upper bound from the laser method, as the lower convex hull of the values tabulated in…
Figure 2 of the paper
Figure 2. The direct sum T1 ⊕ T2 (left) and the partial direct sum T1 ⊕1 T2 (right), drawn as three￾dimensional arrays. All coefficients outside the two boxes are zero. We will be working a lot with these partial direct sums ⊕1, but in the end we will need a way to transform these back into the usual world of direct sums ⊕: (D2) Separating partial direct sums into direct sums. For a partial direct sum T1 ⊕1 · · · ⊕1 Tm and every integer t ≥ 1, (T1 ⊕ · · · ⊕ Tm) ⊗ P1,t ⊴ (T1 ⊕1 · · · ⊕1 Tm) ⊗ Pm,t (Proposition 3.5). Note how ingredient (D2) turns a partial direct sum into a direct sum, at the c
Figure 3 of the paper
Figure 3. Lemma 2.3, with the tensors drawn as three-dimensional arrays: the first leg U points into the page, the second leg V1 ⊕ V2 downwards and the third leg W1 ⊕ W2 to the right. The coefficients of Ti,j lie in the box in block row i and block column j, and all other coefficients are zero. The tensor T (left) has no box in block row 2 and block column 1, and degenerates on the second and third legs to the sum T1,1 + T2,2 of its diagonal blocks (right). Lemma 2.3 (Strassen [Str87, Proposition 6.2]). Let U, V1, V2, W1 and W2 be vector spaces. Write T ∈ U ⊗ (V1 ⊕ V2) ⊗ (W1 ⊕ W2) as T = T1,1…
Figure 4 of the paper
Figure 4. P ⊆ log ∆m F ⊆ Q ∩ {w : w1 + w2 + w3 ≥ 2}, for the triangle P on the left, given by Strassen’s support functionals, and the part of Q with w1 + w2 + w3 ≥ 2 on the right, shown with level curves of w1 + w2 + w3. Of course we do not know the position of ωF 3 (1, 1, 1), but it is on the line between 2 3 (1, 1, 1) and 3 4 (1, 1, 1). The bound of Lemma 4.15 has two cases which naturally correspond to whether the maximum is at the boundary of [0, 1]3 or in the interior. If c(a + b) ≤ ab, then the maximum is attained at the vertex (0, 1, 1) of Q. If c(a + b) > ab, the the maximum is…

Figures are the authors', reproduced from the paper.

References (15)
  1. [1] [ADW+25] Josh Alman, Ran Duan, Virginia Vassilevska Williams, Yinzhan Xu, Zixuan Xu, and Renfei Zhou. More Asymmetry Yields Faster Matrix Multiplication. In Yossi Azar and Debmalya Panigrahi, editors,Proceedings of the 2025 Annual ACM-SIAM Symposium on Discrete Algorithms, SOD... on Pith
  2. [11] 2, 31 [Sch81] A. Sch¨ onhage. Partial and total matrix multiplication.SIAM J. Comput., 10(3):434–455, 1981.doi:10.1137/0210032. 4 [SDW26] Keiya Sakabe, Mahmut Levent Do˘ gan, and Michael Walter. Strassen’s support functionals coincide with the quantum functionals, on Pithdoi:10.1137/0210032
  3. [12] arXiv:2601.21553. 21 [Str69] Volker Strassen. Gaussian elimination is not optimal.Numer. Math., 13(4):354– 356, 1969.doi:10.1007/BF02165411. 3 [Str87] Volker Strassen. Relative bilinear complexity and matrix multiplication.J. Reine Angew. Math., 375/376:406–443, on PithPith's reviewarXiv
  4. [14] doi:10.1515/crll.1991. 413.127. 21, 29 [Uzn26] Przemyslaw Uznanski. Rectangular matrix multiplication from polynomial products, 2026.arXiv:2610.10494. 2, 41 42 [WXXZ24] Virginia Vassilevska Williams, Yinzhan Xu, Zixuan Xu, and Renfei Zhou. New bounds for matrix multiplication:... on PithPith's reviewarXiv
  5. [15] URL: https: //staff.fnwi.uva.nl/j.zuiddam/papers/convexity.pdf , doi:10.1090/bu ll/1880. 2, 7, 21 43 on Pithdoi:10.1090/bu
  6. [1983] doi:10.101 6/0304-3975(83)90054-3. 31 [L¨ u26] Houxi L¨ u. Polynomial separation of a shared tensor leg and matrix multiplication in arbitrary characteristic, 2026.arXiv:2610.18914. 41 [Ope26a] OpenAI. An Upper Bound of 9/4 for the Matrix Multiplication Exponent. OpenAI Math R... on PithPith's reviewarXiv
  7. [1987] doi:10.1515/crll.1987.375-376.406. 10, 11, 12, 13, 14, 21, 22, 23, 24, 39 [Str88] Volker Strassen. The asymptotic spectrum of tensors.J. Reine Angew. Math., 384:102–152, 1988.doi:10.1515/crll.1988.384.102. 2, 8, 21, 22, 24, 29 [Str91] Volker Strassen. Degeneration and complexi... on Pithdoi:10.1515/crll.1987.375-376.406
  8. [1991] A first course, Readings in Mathematics.doi:10.1007/978-1-4612-0979-9. 17 [LR83] Grazia Lotti and Francesco Romani. On the asymptotic complexity of rectangular matrix multiplication.Theoret. Comput. Sci., 23(2):171–185, on Pithdoi:10.1007/978-1-4612-0979-9
  9. [1997] doi:10.1007/978-3-662-03338-8 . 3, 12, 15 41 [Bl¨ a13] Markus Bl¨ aser.Fast Matrix Multiplication. Number 5 in Graduate Surveys. Theory of Computing Library, on Pithdoi:10.1007/978-3-662-03338-8
  10. [2013] doi:10.4086/toc.gs.2013.005. 3, 23, 24 [CLGLZ25] Matthias Christandl, Fran¸ cois Le Gall, Vladimir Lysikov, and Jeroen Zuiddam. Barriers for rectangular matrix multiplication.Comput. Complexity, 34(1):Paper No. 4, 38, 2025.doi:10.1007/s00037-025-00264-9. 9 [CVZ21] Matthias Chr... on Pithdoi:10.4086/toc.gs.2013.005
  11. [2015] 9 [Alm19] Josh Alman. Limits on the universal method for matrix multiplication. In Proceedings of the 34th Computational Complexity Conference (CCC 2019), pages 12:1–12:24, 2019.arXiv:1812.08731,doi:10.4230/LIPIcs.CCC.2019.12. 9 [BCS97] Peter B¨ urgisser, Michael Clausen, and ... on PithPith's reviewarXiv
  12. [2021] doi: 10.4086/toc.2021.v017a002. 9 [CVZ23] Matthias Christandl, P´ eter Vrana, and Jeroen Zuiddam. Universal points in the asymptotic spectrum of tensors.J. Amer. Math. Soc., 36(1):31–79, on Pithdoi:10.4086/toc.2021.v017a002
  13. [2023] doi:10.1090/jams/996. 21 [FH91] William Fulton and Joe Harris.Representation theory, volume 129 ofGraduate Texts in Mathematics. Springer-Verlag, New York, on Pithdoi:10.1090/jams/996
  14. [2025] doi:10.1137/1.97 81611978322.63. 2, 3, 31 [AFLG15] Andris Ambainis, Yuval Filmus, and Fran¸ cois Le Gall. Fast matrix multiplica- tion: limitations of the Coppersmith-Winograd method (extended abstract). In STOC’15—Proceedings of the 2015 ACM Symposium on Theory of Computing, ... on Pithdoi:10.1137/1.97
  15. [2026] 2, 41 [Ope26b] OpenAI. Complex Matrix Multiplication Below 2 .258 and Rectangular Bounds. OpenAI Math Release preprint OAI:Complex-Matrix-Multiplication-Below-2.258- and-Rectangular-Bounds-September-24-2026, on Pith
Cited by 10 Pith papers

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment