OctoInformBeta

OpenAI preprint, September 2026 · 3. Cutoffs, spines and tree diagrams

Lemma 3.1 [Cutoff reflector]: The reflection unit preserves cells through dimension qq

Coskeletal models admit a cutoff reflector.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

The cutoff problem concerns universal maps from models into the category of models coskeletal at a fixed level q∈ℕq\in\mathbb N.

The globe category 𝔾\mathbb G has one object gjg_j for every dimension j∈ℕj\in\mathbb N.✦ Its morphisms are identities and, for m<jm<j, the source and target faces σm,j,τm,j:gm→gj\sigma_{m,j},\tau_{m,j}:g_m\to g_j; write σj\sigma_j and τj\tau_j for the faces from dimension jj to dimension j+1j+1. A globular set ZZ is a family of cell sets ZjZ_j with source and target maps sj,tj:Zj+1→Zjs_j,t_j:Z_{j+1}\to Z_j satisfying the globular identities (1): sj∘sj+1=sj∘tj+1,tj∘sj+1=tj∘tj+1.s_j\circ s_{j+1}=s_j\circ t_{j+1},\qquad t_j\circ s_{j+1}=t_j\circ t_{j+1}.(1)✦ A morphism f:Z→Wf:Z\to W of globular sets is a family of maps fj:Zj→Wjf_j:Z_j\to W_j commuting with source and target.✦

A globular presheaf FF is a functor 𝔾op→Set\mathbb G^{\mathrm{op}}\to\mathrm{Set}.✦ Its underlying globular set has cells F(gj)F(g_j) and source and target maps F(σj)F(\sigma_j) and F(τj)F(\tau_j).✦ A natural transformation of globular presheaves induces a globular map whose cell maps are its components.✦ The representable disk DjD_j is the presheaf 𝔾(−,gj)\mathbb G(-,g_j)✦; its face maps are the Yoneda images of the corresponding faces of 𝔾\mathbb G.✦

A table of dimensions records a single globe or a finite succession of globular attachments, with a distinguished last globe.✦ The realization SS of a table whose last globe has dimension ℓ\ell is a finite globular sum equipped with the last globe inclusion λ:Dℓ→S\lambda:D_\ell\to S. An attachment adds a globe DhD_h along a dimension b<ℓ,hb<\ell,h: its source bb-face is glued to the target bb-face of the last globe of SS. Writing ιL\iota_L and ιR\iota_R for the pushout inclusions, the attached realization and its last globe inclusion are given by (2): S′=S∐DbDh,λ′=ιR:Dh→S′,Db→λ∘τb,ℓS,Db→σb,hDh.\begin{gathered} S'=S\amalg_{D_b}D_h,\qquad \lambda'=\iota_R:D_h\longrightarrow S',\\ D_b\xrightarrow{\lambda\circ\tau_{b,\ell}}S,\qquad D_b\xrightarrow{\sigma_{b,h}}D_h. \end{gathered}(2)✦ For a single globe, the last globe inclusion is its identity. The height of a table, or of its shape SS, is the greatest dimension of its constituent globes. Let Θ0\Theta_0 be the category of finite globular sums; the globe inclusion γ:𝔾→Θ0\gamma:\mathbb G\to\Theta_0 sends gjg_j to DjD_j and faces to their representable face maps. For a finite globular sum SS whose last globe has dimension ℓ\ell and a dimension b<ℓb<\ell, the gluing face inclusion eS,b:Db→Se_{S,b}:D_b\to S is the target bb-face of that last globe followed by the last globe inclusion λ\lambda.

A category CC under globular arities is a category with finite globular sums as objects and an identity-on-objects functor ιC:Θ0→C\iota_C:\Theta_0\to C.✦ A globular theory CC is a category under globular arities for which ιC\iota_C sends every elementary gluing (2) to a pushout.✦ The operations from a shape SS to a shape TT in a globular theory CC are the morphisms C(S,T)C(S,T), and the globe inclusion of CC is γC=ιC∘γ\gamma_C=\iota_C\circ\gamma. A morphism F:B→EF:B\to E of globular theories is a functor satisfying F∘ιB=ιEF\circ\iota_B=\iota_E.✦

Two operations a,b:Dj→Sa,b:D_j\to S of a globular theory are parallel if their source restrictions agree and their target restrictions agree; in dimension zero every pair is parallel.✦ An admissible pair p=(a,b)p=(a,b) of operations from a disk of dimension jj to a shape SS in a globular theory is a parallel pair with height⁡(S)≤j+1\operatorname{height}(S)\le j+1.✦ For a theory morphism F:B→EF:B\to E and an admissible pair p=(a,b)p=(a,b) of operations Dj→SD_j\to S in BB, a filler zz of the image of pp is an operation z:Dj+1→Sz:D_{j+1}\to S in EE with the boundary conditions (3): z∘ιE(σj)=F(a),z∘ιE(τj)=F(b).z\circ\iota_E(\sigma_j)=F(a),\qquad z\circ\iota_E(\tau_j)=F(b).(3)✦

For a family of admissible pairs pgp_g in a globular theory BB, a free extension u:B→B′u:B\to B' by fillers is a theory morphism with specified fillers zgz_g for their images, universal with these choices: any theory morphism a:B→Ea:B\to E and fillers wgw_g for the images of pgp_g extend uniquely to a theory morphism b:B′→Eb:B'\to E satisfying b∘u=ab\circ u=a and b(zg)=wgb(z_g)=w_g.✦ A cellular presentation of a globular theory CC is a sequence of theories starting at an initial theory, with each successive stage a free extension by a family of fillers, and compatible maps from the stages exhibiting CC as their colimit.✦ A Grothendieck coherator CC is a globular theory that has a cellular presentation and a filler for every admissible pair in CC, with fillers taken for the identity morphism of CC.✦

A model XX of CC is a presheaf Cop→SetC^{\mathrm{op}}\to\mathrm{Set} sending each elementary globular sum to the pullback (4): X(S∐DbDh)≅X(S)×X(Db)X(Dh).X(S\amalg_{D_b}D_h)\cong X(S)\times_{X(D_b)}X(D_h).(4)✦ Write 𝒜\mathcal A for the category of models of CC and their natural transformations. The underlying globular set of XX is that of its restriction along γC\gamma_C; its cells are denoted XjX_j.✦ A model morphism f:X→Yf:X\to Y induces the globular map with components fj:Xj→Yjf_j:X_j\to Y_j obtained by the same restriction.✦

Two cells of a globular set are parallel when they have the same source and target, with every pair of zero-cells parallel.✦ A jj-boundary of a globular set ZZ is the unique empty boundary when j=0j=0, and a pair of parallel (j−1)(j-1)-cells when j>0j>0.✦ Writing BjZB_jZ for its boundary set, the boundary map is specified in (5): B0Z={*},Bj+1Z={(x,y)∈Zj×Zj∣x and y are parallel},∂0(x)=*,∂j+1(x)=(sjx,tjx).\begin{gathered} B_0Z=\{*\},\qquad B_{j+1}Z=\{(x,y)\in Z_j\times Z_j\mid x\text{ and }y\text{ are parallel}\},\\ \partial_0(x)=*,\qquad \partial_{j+1}(x)=(s_jx,t_jx). \end{gathered}(5) A model XX is qq-coskeletal when the boundary maps of its underlying globular set satisfy (6): ∂j:Xj→BjXis bijective for every j>q.\partial_j:X_j\longrightarrow B_jX \quad\text{is bijective for every }j>q.(6)✦ Let 𝒜q\mathcal A_q be the full subcategory of 𝒜\mathcal A consisting of qq-coskeletal models, and let Gq:𝒜q↪𝒜G_q:\mathcal A_q\hookrightarrow\mathcal A be the inclusion.

The Claim

Theorem 1. Let CC be a Grothendieck coherator and let q∈ℕq\in\mathbb N. For every model X∈𝒜X\in\mathcal A, there are a model RR and a morphism η:X→R\eta:X\to R satisfying (7): R∈𝒜q,for every Y∈𝒜q and f:X→GqY,there is a unique ℓ:R→GqY with ℓ∘η=f.\begin{gathered} R\in\mathcal A_q,\\ \text{for every }Y\in\mathcal A_q\text{ and }f:X\to G_qY,\\ \text{there is a unique }\ell:R\to G_qY \text{ with }\ell\circ\eta=f. \end{gathered}(7)✦ Choose such a pair for each XX, writing R¯qX\overline R_qX for the chosen underlying reflection and ηX:X→R¯qX\eta_X:X\to\overline R_qX for its unit. The reflection object RqXR_qX is R¯qX\overline R_qX regarded as an object of 𝒜q\mathcal A_q.✦ These choices give the reflection functor (8) and the adjunction (9): Rq:𝒜→𝒜q,X↦RqX.R_q:\mathcal A\longrightarrow\mathcal A_q,\qquad X\longmapsto R_qX.(8)✦ Rq⊣Gq.R_q\dashv G_q.(9)✦ For every model X∈𝒜X\in\mathcal A and every dimension jj with 0≤j≤q0\le j\le q, the unit satisfies (10): (ηX)j:Xj→(R¯qX)jis bijective.(\eta_X)_j:X_j\longrightarrow(\overline R_qX)_j\quad\text{is bijective}.(10)✦

In the paper, the result reads: for q≥0q\ge0, the inclusion 𝒜q⊆𝒜\mathcal A_q\subseteq\mathcal A has a left adjoint RqR_q whose unit X→RqXX\to R_qX is bijective on cells in every dimension at most qq, and for 0≤k<q0\le k<q there is a further reflector Rkq:𝒜q→𝒜kR_k^{\,q}:\mathcal A_q\to\mathcal A_k with RkqRq≅RkR_k^{\,q}R_q\cong R_k naturally.

What Is Checked

The statements concern a Grothendieck coherator CC, its presheaf models, and the boundary maps (5) of their restricted globular sets.

For a model XX and a qq-coskeletal model YY, the adjunction bijection ΦX,Y\Phi_{X,Y} sends a morphism from RqXR_qX to YY to its composite with the unit ηX\eta_X after applying the inclusion GqG_q; its inverse gives the unique factorization through that unit.✦

For models Y,Y′∈𝒜qY,Y'\in\mathcal A_q, a morphism g:Y→Y′g:Y\to Y', and a morphism h:RqX→Yh:R_qX\to Y, the adjunction bijection is natural in the coskeletal target: it takes composition with gg to composition with Gq(g)G_q(g), as in (11): ΦX,Y′(g∘h)=Gq(g)∘ΦX,Y(h).\Phi_{X,Y'}(g\circ h)=G_q(g)\circ\Phi_{X,Y}(h).(11)✦