By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

Latest version

Checks complete

Sync OpenAI Math manuscripts from 4eaee87548250b2682affda7d05d947c05554bae

Committed

All papers

Claims

Findings

The additive indecomposability of the primes

If A,B⊆N0A, B \subseteq \mathbb{N}_0 satisfy ∣A∣,∣B∣≥2|A|, |B| \ge 2, then (A+B)△P(A + B) \triangle \mathcal{P} is infinite; equivalently, every finite modification of P\mathcal{P} is additively indecomposable into two sets with at least two elements each.

There do not exist two infinite sets A,B⊆N0A, B \subseteq \mathbb{N}_0 such that (A+B)△P(A + B) \triangle \mathcal{P} is finite.

Under the assumed eventual decomposition of the primes, ∑T≤log⁡p≤2Tlog⁡ppmax⁡t∈Fp∣Ex∈Spχp(x−t)∣=o(T)\sum_{T \le \log p \le 2T} \frac{\log p}{p} \max_{t \in \mathbb{F}_p} |\mathbb{E}_{x \in S_p} \chi_p(x - t)| = o(T) as T→∞T \to \infty.

For every fixed 0<α<β0 < \alpha < \beta, outside a set of primes of harmonic mass o(L)o(L) in αL≤log⁡log⁡p≤βL\alpha L \le \log\log p \le \beta L, one has uniformly σp=12+o(1)\sigma_p = \frac{1}{2} + o(1) and max⁡a∈Fp,λ multiplicative∣Ev∈Fpgp(v)λ(v)ep(av)∣=o(1)\max_{a \in \mathbb{F}_p, \lambda \text{ multiplicative}} |\mathbb{E}_{v \in \mathbb{F}_p} g_p(v) \lambda(v) e_p(av)| = o(1).

There is an absolute constant δ0>0\delta_0 > 0 such that, for all sufficiently large LL, ∑.05L≤log⁡log⁡p≤.9L1/3≤σp≤2/3γp≥δ01p≥.15L\sum_{\substack{.05L \le \log\log p \le .9L \\ 1/3 \le \sigma_p \le 2/3 \\ \gamma_p \ge \delta_0}} \frac{1}{p} \ge .15L.

In a single finite-field binary-tree diagram of depth ll with independent uniform leaf variables in U=Fq×U = \mathbb{F}_q^\times, E∣W∣2≤3r\mathbb{E}|W|^2 \le 3^r; for two diagrams of the same depth l≥2l \ge 2 paired on b≤3r/4b \le 3r/4 components, ∣EW1W2‾∣≤Cl(εq+q−1/4)|\mathbb{E} W_1 \overline{W_2}| \le C_l(\varepsilon_q + q^{-1/4}).

In either the final or diagonal outer environment, E∑0<∣s∣≤VlG∣Al(s)∣2≤exp⁡(r(Δ0+Cm)+ok(m))\mathbb{E} \sum_{0 < |s| \le V_l} \mathcal{G} |A_l(s)|^2 \le \exp(r(\Delta_0 + Cm) + o_k(m)), and for a pair of assignments with l≥2l \ge 2 and at most 3r/43r/4 overlap components, ∣E∑0<∣s∣≤VlGKarAl(1)(s)Al(2)(s)‾∣≤exp⁡(−ωk(m))|\mathbb{E} \sum_{0 < |s| \le V_l} \mathcal{G} K_{\mathrm{ar}} A_l^{(1)}(s) \overline{A_l^{(2)}(s)}| \le \exp(-\omega_k(m)) where ωk(m)/m→∞\omega_k(m)/m \to \infty.

Methods

Amplified moment with a periodic weight W(n), Poisson summation with arbitrary prime-dependent translations, and a uniform moment estimate over a polynomial-sized family of coefficient arrays via Bonami hypercontractivity, combined with Heath-Brown's quadratic large sieve.

Source excerpts

“An amplified moment and removal of repeated primes.” · Show in paper

“A moment estimate for a polynomial-sized family of arrays.” · Show in paper

“We use Heath-Brown’s quadratic large sieve [18, Theorem 1]” · Show in paper

Repeated Cauchy-Schwarz transfers generating binary-tree frequency histories, combined with small and large anchor primes whose anchor codes distinguish bulk variable positions under permutations.

Source excerpts

“Templates and the transfer identity” · Show in paper

“Anchor codes and the diagonal bound” · Show in paper

Centered tensor coordinate energy identity and contracting local kernel built from sparse additive Fourier spectra, compared against prime-coverage lower bounds via an unweighted primitive-character zero-density estimate.

Source excerpts

“5.1 A budget for centered tensor coordinates” · Show in paper

“5.2 A contracting local kernel from a sparse transform” · Show in paper

“5.4 The prime mean of the weight” · Show in paper

Finite-field binary-tree comparison using cycle reduction on the bipartite multigraph of component sets and bottom quartets, paired with Mellin-character and affine-action estimates.

Source excerpts

“6.2 The cycle reduction and the quartet estimates” · Show in paper

“6.4 A uniform Mellin estimate” · Show in paper

Joint idealization of prime variables into real log coordinates and Haar-unit residues (retaining any Landau-Page exceptional character), maximal-atom Schwartz-Zippel coincidence removal, and permutation symmetrization over bulk variables.

Source excerpts

“8.2 A progression estimate retaining the exceptional term” · Show in paper

“8.5 Removing arithmetic coincidences in the line conditions” · Show in paper

“9.4 Symmetrization and contradiction” · Show in paper