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

Eventual categoricity for abstract elementary classes

For every infinite cardinal λ\lambda there is a cardinal μ(λ)≥λ\mu(\lambda) \ge \lambda such that for every abstract elementary class KK with LS(K)≤λ\mathrm{LS}(K) \le \lambda, if KK is categorical in some cardinal κ≥μ(λ)\kappa \ge \mu(\lambda), then KK is categorical in every cardinal κ′≥μ(λ)\kappa' \ge \mu(\lambda), proved in ZFC without assuming amalgamation, joint embedding, tameness, no maximal models, or arbitrarily large models.

An abstract elementary class KK that is categorical in unboundedly many cardinals is categorical on a tail of the cardinals, including the existence of a model at every cardinal in that tail.

For each infinite cardinal θ\theta there is a cardinal bK(θ)b_K(\theta) such that whenever κ≥bK(θ)\kappa \ge b_K(\theta) is a categoricity cardinal, M∈KM \in K has size κ\kappa, f:M→Mf: M \to M is strong, and aˉ\bar{a} is a tuple of length at most θ\theta in MM, there is a strong g:M→Mg: M \to M with g(f(ai))=aig(f(a_i)) = a_i for every coordinate ii, without assuming κθ=κ\kappa^\theta = \kappa.

There is an unbounded class of strong-limit cardinals μ\mu of countable cofinality at which KK is categorical, any two μ\mu-sized strong extensions of a μ\mu-model have a strong amalgam of size μ\mu, amalgamation of bounded tuples is characterized by equality of bounded diagrams on parameter sets of size at most g(σ)g(\sigma), and there are at most μ\mu singleton types over each μ\mu-model.

Given a working scheme C\mathcal{C} with componentwise arrows and a stationary finite prescription, the marked-increment construction Cb\mathcal{C}^b is again a working scheme on an unbounded class of sufficiently closed eligible levels, and its nn-fold iteration Cn\mathcal{C}_n admits amalgamated boundary completion and exact boundary lifting.

There is a threshold λ>λ0\lambda > \lambda_0 such that in arbitrary cardinalities, if A≤KBA \le_K B are pointed models with ∥A∥≥λ\|A\| \ge \lambda and P(A)=P(B)P(A) = P(B), then A=BA = B; moreover, for every pointed model MM with ∥M∥>λ\|M\| > \lambda, ∣P(M)∣=rk(P(M))=∥M∥|P(M)| = \mathrm{rk}(P(M)) = \|M\|.

For every finite allowed shape LL and every monotone assignment (κs:s∈L)(\kappa_s : s \in L) of cardinals at least γ\gamma (a common geometric level from Theorem 13.5 above λ\lambda), any prescribed boundary isomorphism between two geometric systems with that assignment extends to an isomorphism of their tops.

With the fixed data of either the marked or geometric application, there is an unbounded class of countable-cofinality strong limit cardinals in ZFC at which all finite-depth conclusions hold simultaneously.

Methods

Set-coding reduction of AECs with LS(K)≤λ\mathrm{LS}(K) \le \lambda to a set of small-model and small-pair codes, with first-order set-theoretic recovery of membership and strong substructure via directed unions.

Source excerpts

“Lemma 2.2 (Recovery from small models and pairs). Fix an infinite cardinal λ and a finitary language τ of cardinality at most λ. An AEC in τ with LS(K) ≤ λ is determined by its models of cardinality at most λ and its strong inclusions between such models, both taken up to isomorphism. These data have a set code, and the recovered class and strong-submodel relation are uniformly first-order definable in set theory from that code.” · Show in paper

Finite-support Ehrenfeucht-Mostowski order presentations E(I)E(I) constructed via finite-arity partition calculus (Erdős-Rado) and indexed by H(δ)=Q(Γδ)H(\delta) = \mathbb{Q}^{(\Gamma_\delta)} to control short-tuple orbits and reverse strong maps.

Source excerpts

“Theorem 2.6 (Order presentation). Suppose K has arbitrarily large models and LS(K) = ℓ. There are at most ℓ labels, each of positive finite arity, and a functor E from nonempty linear orders and order embeddings to K and strong maps, with the following properties.” · Show in paper

“Definition 3.1. For an infinite limit ordinal δ, set Γδ = Z^(δ), H(δ) = Q^(Γδ).” · Show in paper

Disjoint block transpositions, ordinal-vector witnesses, and compact placement to obtain uniform testing bounds and amalgamation at countably cofinal strong-limit closure levels.

Source excerpts

“Lemma 4.7 (Ordinal-vector witnesses). Let C be a set, and let U be a definable class of functions C → Ord, with fixed set parameters allowed in its definition. There is a set U0 ⊆ U such that each u ∈ U dominates some v ∈ U0 coordinatewise.” · Show in paper

“Proposition 4.9 (Compact placement). Let A = E(I), where I is an infinite well-order, and let χ > ℓ + σ satisfy χ^σ = χ. Suppose that for every X ⊆ A of size at most χ a possible joint diagram qX(d, X) is prescribed, where d has length at most σ. Assume these prescriptions restrict coherently and that their parameter diagrams are the natural ones of X. Then one placement of d over I realizes all the prescriptions.” · Show in paper

Working-scheme calculus using stationary free extensions via majority approximation sequences, universal fresh strips, and approximate retractions for descent.

Source excerpts

“Theorem 6.10 (Stationary free extension). Let A → B be an arrow at a sufficiently closed level and let p = tp(v/A), where |v| ≤ σ. There is a specified actual type p^B over B, called its free lift, extending p.” · Show in paper

“Corollary 6.16 (Approximate retractions). For same-level A ≤ B there are arrows (ri : B → A : i < µ) such that, for every a ∈ A, |{i < µ : ri(a) ≠ a}| < ℵ0.” · Show in paper

Relative isolation and all-dominated extensions constructed via Fodor's pressing-down lemma on stationary sets and transferred downward through approximate retractions.

Source excerpts

“Fodor’s pressing-down argument [Fod56] makes j(i) constant on a stationary subset.” · Show in paper

“Theorem 7.9. At sufficiently closed eligible levels, every finite-tuple type p = tp(b/X) has an all-dominated realization: there is Y ≥ X all-dominated over X + b realizing p.” · Show in paper

Return-tree binary splitting argument keeping full maps coherent along individual branches while retaining only boundary maps in a common accumulator to prove joint boundary diagram preservation.

Source excerpts

“We construct branch worlds Ws ∈ Cm for s ∈ ≤ϵ2, carrying tuples fs, with coherent full arrows along each branch. Their proper boundaries have compatible arrows qs : ∂Ws −→ ∂A into an increasing common accumulator A ∈ Cm.” · Show in paper

Uniform finite-depth set-theoretic coding, a fixed register of bound requests, and a common majorant Φ\Phi with limit-point hierarchy GjG_j to obtain common closure levels in ZFC.

Source excerpts

“We may therefore choose a fixed, monotone definable ordinal function Φ such that Φ(α) > ϑ, Φ(α) > α, Bq(d, β⃗) < Φ(α) (max β⃗ ≤ α).” · Show in paper