betaFeedback

WorksComputer Science

A Numerical Refinement of an Exact Discrete Fourier Transform Bound

Abstract

We refine the numerical bounds in Kartal's instantiation of Sussman's exact discrete Fourier transform construction. For every integer n >= 1, the resulting algorithm computes the length-n transform with work O(n (log n)^(1 - delta)), where delta = 0.0030201659. The bound holds in an exact complex-arithmetic RAM model with a supplied canonical root of unity. This final exponent saving exceeds the earlier certified parameter C_0 = 0.00302016 by 5.9 x 10^(-9). We retain the underlying circuit, tighten rational bounds for sums of powers of its block and target ranks, and increase the fixed tensor power from 7760 to 11640. These bounds establish the inequalities for Sussman's recursion at the intermediate certificate parameter C = 0.00302016595. The gap C - delta absorbs the overhead of recursion and Fourier reduction. The resulting transform theorem is proved in Lean and checked against the reference problem statement.

Introduction

We consider the arithmetic work needed to compute the discrete Fourier transform for every integer input length n≥1n \ge1. Cooley and Tukey’s factorization gives the familiar O(nlog⁡n)O(n \log n) bound, while Bluestein’s reduction extends this order of growth to arbitrary lengths [4, 2]. More recent work has reduced leading constants: Alman and Rao obtain 154nlog⁡2n+O(n)\frac{15}{4}n \log_2 n + O(n) real arithmetic operations for lengths that are powers of two [1].

Another line of work studies a saving in the exponent of the logarithm under exact complex arithmetic. OpenAI’s construction for arbitrary lengths gives such a saving in a uniform RAM model with charged scalar preparation and a supplied root of unity [11]. Subsequent finite circuits and formalizations have increased the explicit saving. The present work concerns the instantiation by Kartal [9], which uses Sussman’s mixed-target recursion [12] and the controller and tensor constructions developed by Davis and the integer-multiplication circuit community [5, 6, 7].

The exponent in this construction is controlled by a finite set of inequalities involving the sizes and counts of its component computations. We call the exact numerical bounds that verify these inequalities a numerical certificate. Kartal’s submission states a final Fourier saving of 0.0030103, while its certificate establishes the stronger parameter C0=4719/1562500=0.00302016C_0 = 4719/1562500 = 0.00302016 [9]. The recursion and Fourier reduction permit any fixed saving 0<δ<C00 < \delta< C_0: the gap below C0C_0 absorbs their asymptotic overhead. Thus, for example, a saving of 0.00302 already follows from the earlier argument. We improve the numerical certificate to obtain a saving beyond C0C_0.

Theorem 1.1. In the exact complex-arithmetic RAM model described in § 2, there are fixed deterministic programs that compute the length-nn discrete Fourier transform for every integer n≥1n \ge1, with work

W(n)=O(n(log⁡n)1−δ),δ=302016591010=0.0030201659.W(n) = O\left(n(\log n)^{1-\delta}\right), \qquad\delta= \frac{30201659}{10^{10}} = 0.0030201659.

The work bound includes the root-order computation, one supplied-root operation, scalar preparation, and array and index operations. Integer values and addresses are bounded by a fixed polynomial in nn.

The improvement is a numerical refinement of the existing algorithm. We use the same raw circuit, the same sizes and counts of its component computations, and the same reduction to the Fourier transform. The changes are tighter rational estimates of the rank moments and a larger fixed tensor power. Our final DFT saving is δ=0.0030201659\delta= 0.0030201659, exceeding the earlier certified parameter C0=0.00302016C_0 = 0.00302016 by 5.9×10−95.9 \times10^{-9}. Earlier circuit analysis placed the raw moment balance point near 0.003020165977. Our final saving remains below this value to accommodate the overhead of the tensor construction, recursion, and Fourier reduction. Conditional integer-multiplication constructions had also used rates close to that point [8, 3].

The proof below gives the finite arithmetic behind the refinement and then applies Sussman’s recursion and the Fourier reduction. The accompanying Lean proof verifies the resulting transform algorithm and its work bound in the reference model.

Computational model and the finite unit

For n≥1n \ge1, write ζn=exp⁡(2πi/n)\zeta_n = \exp(2\pi i/n). We use the positive-sign, unnormalized transform

(DFTnx)j=∑k=0n−1ζnjkxk,0≤j<n.(1)(\mathrm{DFT}_n x)_j = \sum_{k=0}^{n-1} \zeta_n^{jk}x_k,\qquad0 \le j < n. \tag*{(1)}

All logarithms in complexity bounds are natural logarithms.

We retain OpenAI’s exact RAM model [11]. A fixed program first computes a positive integer D(n)D(n) from nn. A root primitive then supplies the exact complex number ζD(n)=exp⁡(2πi/D(n))\zeta_{D(n)} = \exp(2\pi i/D(n)), after which a second fixed program computes (1). The order D(n)D(n) may differ from the transform length nn. The formal specification requires D(n)<1024n3D(n) < 1024n^3 and bounds all integer values and addresses by (n+2)c(n+2)^c for one constant cc. Its cost includes both programs and the root primitive. Exact scalar arithmetic and integer-word, array, and index operations are charged according to the RAM semantics. Scalar coefficients have unrestricted magnitude. The work bound counts exact arithmetic and RAM operations.

The finite unit used here is Kartal’s P16A8Z instance [9]. It combines a 16-pair controller, with prefix parameter eight, and an 18-stage geometry with 145 label coordinates. Its supplier comes from the patched circuit in Davis’s joint-helper construction [7]. In the cost analysis, a charged block is a component computation whose work is counted, and a target job specifies a computation that the circuit carries out. The rank of a block or job is the number of coordinate axes on which it acts. After dividing out a common multiplicity factor, the ranks and their counts form the two multisets in Table 1. For example, the entry (2,6984)(2, 6984) describes 6984 charged blocks of rank two. We use the same two multisets as Kartal’s construction.

Charged blocks P\mathcal{P}Target jobs T\mathcal{T}
RankCountRankCount
1419464504
269847126
388928846
4264695868
564817448
814425448
1644833448
2444841448
3244849448
4044857448
4844865448
5644873224
64448145224
72224——

Table 1. Ranks and counts of charged blocks and target jobs in the raw P16A8Z construction, with the common multiplicity factor omitted.

For a finite multiset AA of positive integers, define its moment by

Mt(A)=∑r∈Art,(2)M_t(A) = \sum_{r \in A} r^t, \tag*{(2)}

where the sum counts multiplicity. Let P\mathcal{P} and T\mathcal{T} be the charged-block and target multisets, and abbreviate Pt=Mt(P)P_t=M_t(\mathcal{P}) and Tt=Mt(T)T_t=M_t(\mathcal{T}). Direct summation gives

P1=239134,T1=239886,∣T∣=10928.(3)P_1 = 239134,\qquad T_1 = 239886,\qquad|\mathcal{T}| = 10928. \tag*{(3)}

The recursion compares PtP_t with TtT_t at an exponent t<1t<1. It also needs every charged-block rank to be smaller than every target rank. The raw multisets do not have this separation: some block ranks exceed the smallest target rank. Tensor products, target completion, and block splitting establish it in Section 4.

The refined numerical bounds

Fix the exact rational numbers

C=6040331920000000000=0.00302016595,t=1−C,ε=271012.(4)C = \frac{60403319}{20000000000} = 0.00302016595,\qquad t = 1-C,\qquad\varepsilon= \frac{27}{10^{12}}. \tag*{(4)}

Our first task is to prove Pt≤(1−ε)TtP_t \le(1-\varepsilon)T_t. The gap is small enough that the earlier exponential estimates lose the required precision. The following elementary bounds suffice.

Lemma 3.1. Let S6(x)=∑j=05xj/j!S_6(x)=\sum_{j=0}^{5}x^j/j!. For x≥0x\ge0, S6(x)≤exS_6(x)\le e^x. For 0≤x≤10\le x\le1,

ex≤S6(x)+74320x6.(5)e^x \le S_6(x) + \frac{7}{4320}x^6. \tag*{(5)}

Proof. The lower bound discards nonnegative terms of the exponential series. For the upper bound, its tail starts with x6/6!x^6/6!. Each subsequent term is at most 1/71/7 times the preceding term when x≤1x\le1. The tail is therefore at most (x6/720)/(1−1/7)=7x6/4320(x^6/720)/(1-1/7)=7x^6/4320.

We normalize at the largest target rank, 145. For a rank rr with multiplicity mm, its contribution becomes

m(r145)1−C=mr145exp⁡(Clog⁡145r).(6)m\left(\frac{r}{145}\right)^{1-C}=\frac{mr}{145}\exp\left(C\log\frac{145}{r}\right). \tag*{(6)}

Every argument of the exponential lies in [0,0.016][0,0.016]. Rational bounds for log⁡(145/r)\log(145/r) can therefore be substituted into Lemma 3.1 with little loss. Appendix A gives the bounds for every term, together with the finite-series tests that certify the logarithms.

Proposition 3.2. At the parameters in (3),

Pt145t≤16614830130147711012=:U,(7)\frac{P_t}{145^t}\leq\frac{1661483013014771}{10^{12}}=:U, \tag*{(7)}
Tt145t≥16614830130615371012=:L.(8)\frac{T_t}{145^t}\geq\frac{1661483013061537}{10^{12}}=:L. \tag*{(8)}

Consequently, Pt/Tt≤1−εP_t/T_t\leq1-\varepsilon.

Proof. For each charged-block rank, use an upper logarithm bound in (5) and then (4). For each target rank, use a lower logarithm bound and the lower polynomial S6S_6. The rank-145 target term is exactly 224. Adding the respective rational enclosures in Table 2 gives (6) and (7). Both sums are positive, and exact rational arithmetic gives

Rank rrLogarithm bound ℓr\ell_rNNMoment bound brb_r
Charged blocks: upper bounds
14.97673374242129293.663682394904
24.2835865618612797.585379104181
33.87812145375326186.139872110444
43.5904393813012573.788927475228
53.3672958299872422.573229270017
82.897292200741238.014652284989
162.2041450201812149.764660119895
241.7986799120731974.555635518839
321.5109978396211899.321181555482
401.28785428830717124.067835657887
481.10553273151316148.799445098402
560.95138205168615173.518550400695
640.81785065906115198.226956274002
720.70006762340514111.463005749806
Target jobs: lower bounds
43.5904393813001214.055033804805
73.030823593365116.138693311557
82.8972922007401147.086082174304
92.77950916508411367.291038096216
172.1435203983641052.865271049404
251.7578579175521077.652546021820
331.48022618095410102.415450004130
411.2631616757169127.160048134959
491.0849134443099151.889974497414
570.9336824745869176.607655012455
650.8023464725248201.314825634830
730.6862743012728113.006395319643
1450—224

Table 2. Exact rational enclosures for the normalized moment terms. The charged-block rows give upper bounds; the target rows give lower bounds. NN is the number of terms in the logarithm test described in the text.

1−UL=467661661483013061537>271012.1-\frac{U}{L}=\frac{46766}{1661483013061537}>\frac{27}{10^{12}}.

Since Pt/Tt≤U/LP_t/T_t\leq U/L, the stated gap follows.

The completion and splitting estimates also use the scaled target moment 20CTt20^C T_t. Applying the same upper exponential bound gives

(14520)C≤E:=251500221699250000000000.(9)\left(\frac{145}{20}\right)^C\leq E:=\frac{251500221699}{250000000000}. \tag*{(9)}

Hence

20CTt=145(145/20)CTt145t≥145LE.(10)20^C T_t=\frac{145}{(145/20)^C}\frac{T_t}{145^t}\geq\frac{145L}{E}. \tag*{(10)}

Set

γ=99857100000,η=958310000.(11)\gamma=\frac{99857}{100000},\qquad\eta=\frac{9583}{10000}. \tag*{(11)}

The rational inequalities

239134≤γ145LE,10928⋅21≤η145LE239134\leq\gamma\frac{145L}{E},\qquad10928\cdot21\leq\eta\frac{145L}{E}

and (2) imply

P120CTt≤γ,21∣T∣20CTt≤η.(12)\frac{P_1}{20^C T_t}\leq\gamma,\qquad\frac{21|\mathcal{T}|}{20^C T_t}\leq\eta. \tag*{(12)}

All numbers used in these comparisons are rational.

A separated tensor construction

We use the completion and splitting argument already present in the earlier construction [5, 9]. The moment estimate is recalled here to make the role of the larger tensor power explicit.

For multisets AA, BB, let A⊗BA \otimes B contain every product abab, with multiplicity, for a∈Aa \in A and b∈Bb \in B. Moments multiply:

Mt(A⊗v)=Mt(A)v.(13)M_t\left(A^{\otimes v}\right)=M_t(A)^v. \tag*{(13)}

In the vv-fold tensor unit, choose positive integers J<BJ < B as the maximum block rank and minimum target rank. A target of rank K<BK < B is completed to rank BB by adding a charged block of rank B−KB-K. Targets of rank at least BB are unchanged. Each charged block of rank R>JR > J is then split into blocks of rank JJ and one remainder block, if needed. Write P′P' and T′T' for the resulting multisets. Every charged-block rank is at most JJ, and every target rank is at least BB.

Lemma 4.1 (Completion and splitting). Let 0≤t≤10 \le t \le1 and v≥1v \ge1. Then

Mt(P′)≤Ptv+2Jt−1P1v+2∣T∣vBJt−1,Mt(T′)≥Ttv.(14)M_t(P') \le P_t^v + 2J^{t-1}P_1^v + 2|T|^vBJ^{t-1}, \qquad M_t(T') \ge T_t^v. \tag*{(14)}

Proof. When R>JR > J, the sum of the ttth powers of its pieces is at most

⌊RJ⌋Jt+(R mod J)t≤RJt−1+Jt≤2RJt−1.\left\lfloor\frac{R}{J}\right\rfloor J^t + (R \bmod J)^t \le RJ^{t-1}+J^t \le2RJ^{t-1}.

For R≤JR \le J, the contribution is RtR^t. Thus a bound valid in both cases is Rt+2RJt−1R^t+2RJ^{t-1}. Summing over the charged blocks in the tensor unit and using (12) gives the first two terms in (13). There is at most one completion block for each target in T⊗vT^{\otimes v}, and its rank is less than BB. After splitting, its contribution is at most 2BJt−12BJ^{t-1}. If its rank is already at most JJ, its contribution is at most Jt≤BJt−1J^t \le BJ^{t-1}. There are ∣T∣v|T|^v possible completion blocks. Finally, completion only increases target ranks, so their tt-moment is at least TtvT_t^v. □\square

Take J=20vJ=20^v and B=21vB=21^v. At t=1−Ct=1-C, division by TtvT_t^v in Lemma 4.1 reduces the desired comparison to

(PtTt)v+2(P120CTt)v+2(21∣T∣20CTt)v<1.(15)\left(\frac{P_t}{T_t}\right)^v+2\left(\frac{P_1}{20^C T_t}\right)^v+2\left(\frac{21|T|}{20^C T_t}\right)^v<1. \tag*{(15)}

By Proposition 3.2 and (11), it is enough to prove (1−ε)v+2γv+2ηv<1(1-\varepsilon)^v+2\gamma^v+2\eta^v<1.

Proposition 4.2. For v=11640v=11640, the completed and split tensor unit has Mt(P′)<Mt(T′)M_t(P')<M_t(T'), and every charged-block rank is smaller than every target rank.

Proof. Exact rational exponentiation gives

γ97≤544625,(544625)5≤12,η40≤12.\gamma^{97}\le\frac{544}{625},\qquad\left(\frac{544}{625}\right)^5\le\frac{1}{2},\qquad\eta^{40}\le\frac{1}{2}.

Since 11640=485⋅24=40⋅29111640=485\cdot24=40\cdot291, γ11640≤2−24\gamma^{11640}\le2^{-24} and η11640≤2−291\eta^{11640}\le2^{-291}. Also, Bernoulli’s inequality gives (1−ε)v≤(1+vε)−1(1-\varepsilon)^v\le(1+v\varepsilon)^{-1}. Therefore

(1−ε)11640+2γ11640+2η11640≤11+11640ε+2−23+2−290<1.(16)(1-\varepsilon)^{11640}+2\gamma^{11640}+2\eta^{11640}\le\frac{1}{1+11640\varepsilon}+2^{-23}+2^{-290}<1. \tag*{(16)}

The last strict inequality is rational; its slack exceeds 195/109195/10^9. Equations (14)–(15) now give the moment comparison. Rank separation follows from 2011640<211164020^{11640}<21^{11640}. □\square

The earlier certificate used v=7760v = 7760. At the new rate the available moment gap is much smaller, so increasing vv lets the two tail terms fall below the accumulated gain in the first term.

Proof of Theorem 1.1. The P16A8Z certificate realizes the raw rank multisets by an exact circuit [9]. Its tensor construction realizes the completed and split multisets used above. Proposition 4.2 supplies the moment comparison and rank separation required by the mixed-target recursion of Sussman [12]. That recursion gives the kernel program used in the reduction to arbitrary transform lengths at any fixed exponent strictly above 1−C1-C.

Choose the intermediate saving

β=30201659251012=0.003020165925.\beta= \frac{3020165925}{10^{12}} = 0.003020165925.

Then δ<β<C\delta< \beta< C. With kernel exponent 1−β1-\beta, the Fourier reduction, including its preparation costs, gives

W(n)=O(n(log⁡n)1−β(log⁡log⁡n)2).W(n) = O\left(n(\log n)^{1-\beta}(\log\log n)^2\right).

For the positive constant β−δ\beta-\delta, (log⁡log⁡n)2=o((log⁡n)β−δ)(\log\log n)^2 = o((\log n)^{\beta-\delta}). This proves the asserted work bound. The fixed-program, root-order, and polynomial integer bounds are supplied by the same reduction. □

Corollary 4.3. In the corresponding exact RAM model allowing multiplication of the two input data streams, the linear convolution of two vectors of length nn can be computed in O(n(log⁡n)0.9969798341)O(n(\log n)^{0.9969798341}) work.

Proof. Apply the convolution reduction to transform length 2n−12n-1. The scalar and pointwise multiplication work is included in that reduction’s cost bound, and 2n−1=Θ(n)2n-1 = \Theta(n). □

Verification and numerical comparison

Theorem 1.1 and the convolution corollary are verified in Lean 4.34.1. Using the development wrapper of Comparator [10], a statement comparison checked the transform theorem against the reference exact-DFT specification at the new rational exponent and rechecked the proof with Lean’s default kernel. The dependency audit reports propositional extensionality, classical choice, and quotient soundness for the transform, convolution, and finite tensor moment theorems. The artifact records these checks and the source changes to the numerical bounds, tensor power, and exponent parameters.

An independent calculation checks the numerical inequalities using exact rational interval arithmetic. It uses an atanh series for logarithms and a separately implemented exponential-series bound, rather than the generator’s logarithm certificates. This check covers the raw moment gap, both tail ratios, and (16).

The interval calculation also resolves the raw balance point more closely. Let F(c)=P1−c−T1−cF(c) = P_{1-c} - T_{1-c}. With

c−=0.00302016597701562048,c+=0.00302016597701562050,c_- = 0.00302016597701562048,\qquad c_+ = 0.00302016597701562050,

the certified signs are F(c−)<0<F(c+)F(c_-) < 0 < F(c_+). Continuity therefore places a zero in (c−,c+)(c_-,c_+). The same calculation gives F(0.003020166)>0F(0.003020166) > 0, so that parameter fails the raw moment test.

The final Fourier saving remains strictly below this interval because the proof uses slack at both the tensor and Fourier stages. The large fixed tensor power also affects the hidden constants in the work bound.

Acknowledgments

OpenAI ChatGPT Codex was used throughout the research and preparation of this paper. Its contributions included analysis of prior constructions, exploration of candidate improvements, derivation of the numerical bounds, development of Lean proofs and verification scripts, execution and analysis of verification checks, preparation of the reproducibility artifact, and drafting and revision of the manuscript. This work was supported in part by a grant of access to OpenAI models through the ChatGPT for Academic Researchers program.

Artifact availability

The accompanying artifact contains the Lean development, exact rational certificates, an independent arithmetic checker, source provenance, and verification logs. Zenodo DOI: https://doi.org/10.5281/zenodo.23843356.

References

Finite rational bounds

Table 2 lists the logarithm and moment enclosures used in Proposition 3.2. Every displayed decimal in the table is an exact rational with denominator dividing 101210^{12}. The multiplicity mrm_r for each row is given in Table 1.

For N≥1N \ge1, put

SN(x)=∑j=0N−1xjj!,UN(x)=SN(x)+N+1NN!xN.S_N(x) = \sum_{j=0}^{N-1} \frac{x^j}{j!}, \qquad U_N(x) = S_N(x) + \frac{N+1}{N N!}x^N.

For 0≤x≤10 \le x \le1, the same tail argument as in Lemma 3.1 gives SN(x)≤ex≤UN(x)S_N(x) \le e^x \le U_N(x). An upper logarithm entry ℓr\ell_r is certified by 145/r≤SN(ℓr)145/r \le S_N(\ell_r); the lower-series inequality holds for all ℓr≥0\ell_r \ge0. A lower logarithm entry is certified by UN(ℓr/8)8≤145/rU_N(\ell_r/8)^8 \le145/r. In the latter case, 0≤ℓr/8≤10 \le\ell_r/8 \le1, and the test implies eℓr≤145/re^{\ell_r} \le145/r. The charged-block moment bound brb_r satisfies mrr145U6(Cℓr)≤br\frac{m_r^r}{145}U_6(C\ell_r) \le b_r. The target bound satisfies br≤r145S6(Cℓr)b_r \le\frac{r}{145}S_6(C\ell_r). All these tests reduce to integer arithmetic after clearing denominators.

For (9), use the upper logarithm bound log⁡(145/20)≤1.981001468867\log(145/20) \le1.981001468867, certified by 145/20≤S20(1.981001468867)145/20 \le S_{20}(1.981001468867), followed by U6U_6 at C⋅1.981001468867C \cdot1.981001468867.

The independent checker obtains logarithm intervals differently. After writing x=2kzx = 2^k z with 1≤z≤21 \le z \le2, it uses

log⁡z=2∑j=0N−1y2j+12j+1+RN,y=z−1z+1,0≤RN≤2y2N+1(2N+1)(1−y2).\log z = 2 \sum_{j=0}^{N-1} \frac{y^{2j+1}}{2j+1} + R_N, \qquad y = \frac{z-1}{z+1}, \qquad0 \le R_N \le\frac{2y^{2N+1}}{(2N+1)(1-y^2)}.

Here N=40N = 40 and 0≤y≤1/30 \le y \le1/3; the same formula encloses log⁡2\log2. For an exponential argument 0≤x≤10 \le x \le1, the checker uses S16(x)S_{16}(x) and the upper tail x16/(16!(1−x/17))x^{16}/(16!(1-x/17)). These bounds also certify the signs reported in § 5 without floating-point acceptance tests.

References

  1. [1]Josh Alman and Kevin Rao. Faster Walsh–Hadamard and discrete Fourier transforms from matrix non-rigidity. arXiv:2211.06459v2, 2023. https://arxiv.org/abs/2211.06459v2.
  2. [2]Leo I. Bluestein. A linear filtering approach to the computation of discrete Fourier transform. IEEE Transactions on Audio and Electroacoustics, 18(4):451–455, 1970. https://doi.org/10.1109/TAU.1970.1162132.
  3. [3]Chrisbryan17. Tighten conditional critical-root transfer to κ = 0.0030201659769 (exact arithmetic). CrocSwap integer-mult-bounds, pull request 4707, 2026. https://github.com/CrocSwap/integer-mult-bounds/pull/4707.
  4. [4]James W. Cooley and John W. Tukey. An algorithm for the machine calculation of complex Fourier series. Mathematics of Computation, 19(90):297–301, 1965. https://doi.org/10.1090/S0025-5718-1965-0178586-1.
  5. [5]Cinna Davis. Conditional κ = 0.00268647 via paid target padding after native composition. CrocSwap integer-mult-bounds, pull request 4416, 2026. https://github.com/CrocSwap/integer-mult-bounds/pull/4416.
  6. [6]Cinna Davis. Conditional κ = 0.00296 via query-thinned productive controllers. CrocSwap integer-mult-bounds, pull request 4440, 2026. https://github.com/CrocSwap/integer-mult-bounds/pull/4440.
  7. [7]Cinna Davis. Conditional κ = 0.00302015: Joint helper boundaries. CrocSwap integer-mult-bounds, pull request 4470, 2026. https://github.com/CrocSwap/integer-mult-bounds/pull/4470.
  8. [8]hexapode. Conditional κ = 0.00302016597. CrocSwap integer-mult-bounds, pull request 4568, 2026. https://github.com/CrocSwap/integer-mult-bounds/pull/4568.
  9. [9]Halit Ziya Kartal. Exact DFT submission: halitziyav4-logception. Big O Speedrun verifier, pull request 41, 2026. Source snapshot 859d6949ef4bdcab4ba661d09cd7fd720d0c4960, directory halitziyarkartal-halitziyav4-logception. https://github.com/bigospeedrun/verifier/pull/41.

Paper details

vlnc:W_kjw7cj5b
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Codex

View statement
Looking for related papers…
Contents