OctoInformBeta

OpenAI preprint, September 2026 · §8. Relational coding of the diagonal argument

Lemma 45: Unrounded channel evaluation with a structural admissibility predicate in the specialization

Class evaluations and congruence for constructed pairs.

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

An encoded predicate pp over an element type is a data term whose observation at each element of that type is a formula. Class predicates, unions of two classes, and pair predicates are formed through observation channels. The relevant comparison of formulas is by uniform typed proof-term transformers.

Fix a pure type specification P=(S,𝒜,ℛ)P=(S,\mathcal A,\mathcal R), with arbitrary axiom and product-rule relations, a context Γ\Gamma, and an assignment γ\gamma of data and proof modes to variables. Terms have annotated abstractions λx:A.M\lambda x:A.M and dependent products Πx:A.B\Pi x:A.B. Write Γ;γ⊩F\Gamma;\gamma\Vdash F when some proof-mode term has type FF in Γ\Gamma, and F≃Γ,γHF\simeq_{\Gamma,\gamma}H when there are typed open proof transformers from FF to HH and from HH to FF. In the latter judgment a temporary proof-mode input is available to each transformer. The labelled terms have an argument mode and a result mode on each application, annotated abstraction and dependent product; the allowed labels are data to data, data to proof and proof to proof. The mode discipline assigns variables their prescribed modes and sorts data mode, and checks applications and abstraction bodies against their argument and result labels. Abstraction annotations and both parts of a dependent product have data mode, with the bound variable assigned the argument mode of the label. A labelled term is data-only when every application, annotated abstraction and dependent product in it has data as both argument and result mode.

Full beta reduction on the underlying pure terms contracts an applied annotated abstraction by capture-avoiding substitution and permits reduction in function and argument positions, annotations and both binder bodies.✦ Labelled beta reduction has the same closure under term components, and its head contraction requires the application and abstraction to have the same label.✦ Beta convertibility is the equivalence relation generated by full beta reduction✦; labelled beta convertibility is the equivalence relation generated by labelled beta reduction.✦ The labelled typing judgment Γ⊢M:A\Gamma\vdash M:A uses the axiom, variable, weakening, product, abstraction, application and conversion rules, with matching function and application labels and conversion to a sorted expected type along labelled beta convertibility.✦

Fix a classical logical interpretation LL, with its feasible witness family and proof paths. Its connectives are written →,¬,∧,∨,↔\to,\neg,\land,\lor,\leftrightarrow; the last denotes the conjunction of the two implications. Quantifier routes KYK_Y and KPK_P supply universal quantification over elements and predicates, respectively. For logical base profile bb and quantifier domain profile II, a quantifier route is a choice of a feasible product profile triple (I,J,K)(I,J,K) of nonempty sort profiles, each consisting of all sorts typing a normal term in a valid context, with output profile specified by (1), an even incoming proof path from bb to JJ, and an even outgoing proof path from KK to bb, relative to the fixed witness family and collection of profiles. Existential quantification is the negation of a universally quantified negation.✦ The logical falsity ⊥\bot of LL is the witness type chosen at its base profile, and logical truth ⊤\top is ⊥→⊥\bot\to\bot. Logical disjunction F∨HF\lor H is the negation of the conjunction of ¬F\neg F and ¬H\neg H. Proof wrapping of a type FF leaves FF unchanged at the identity path; a positive extension by DD forms a nondependent arrow from DD to the preceding wrapped type with data input and proof output, and a negative extension by HH forms an arrow from the preceding wrapped type to HH with proof input and proof output.

A feasible product profile triple (I,J,K)(I,J,K) consists of feasible profiles satisfying (1): K=Out⁡(I,J):={c∣∃a∈I∃b∈J(a,b,c)∈ℛ}.K=\operatorname{Out}(I,J) :=\{c\mid \exists a\in I\;\exists b\in J\;(a,b,c)\in\mathcal R\}.(1)✦ For a term TT with weak normalization, write T#T^\# for its chosen full beta normal form.✦ The normal form sort profile πΓ(T)\pi_\Gamma(T) of a weakly normalizing term TT in a context Γ\Gamma is the set of all sorts ss for which Γ⊢T#:s\Gamma\vdash T^\#:s.✦ The axiom profile Ax⁡(s)\operatorname{Ax}(s) is the set of axiom successors tt with (s,t)∈𝒜(s,t)\in\mathcal A.✦

The primary signed graph has a positive edge from a feasible profile II to each feasible proper subprofile JJ, a negative edge from II to KK for every feasible product profile triple (I,J,K)(I,J,K), and a positive edge from JJ to KK for that triple.✦ Positive reachability is a finite, possibly empty path all of whose edges are positive.✦ For the logical collection 𝒞\mathcal C of sort profiles, an observation layer is a profile positively reachable in the primary graph from a nonempty axiom profile Ax⁡(s)\operatorname{Ax}(s) with {s}∈𝒞\{s\}\in\mathcal C.✦ A direct observation layer is an observation layer II for which some J,K∈𝒞J,K\in\mathcal C form a feasible product profile triple (I,J,K)(I,J,K).✦ A logical terminal profile is a profile JJ for which some I,K∈𝒞I,K\in\mathcal C form a feasible product profile triple (I,J,K)(I,J,K).✦

An observation channel R=(Rℓ,Rr,Rp)R=(R_\ell,R_r,R_p) consists of logical, raw, and predicate retained data paths, each of odd parity.✦ A retained data path is generated by the identity path, positive extensions retaining a domain and dummy argument, and negative extensions with a data tail; parity counts the negative extensions modulo two.✦ A retained data path RR from II to JJ is formed relative to LL when it is generated from the identity at II, permitting restriction of any target profile to a subset without changing the path or its parity. From a path ending at JJ, a feasible triple (A,J,K)(A,J,K) permits a positive extension to KK using the chosen witness type and value at AA, whereas a feasible triple (J,B,K)(J,B,K) and a formed data tail ending at BB permit a negative extension to KK. A formed observation channel R=(Rℓ,Rr,Rp)R=(R_\ell,R_r,R_p) has logical path RℓR_\ell from II to JJ, raw path RrR_r from II to KK, and predicate path RpR_p from KK to HH.✦ For a base type TT and a path DD, write D[T]D[T] for the type wrapped along DD, build⁡D(T,F)\operatorname{build}_D(T,F) for its builder, and obs⁡D(T,v,a)\operatorname{obs}_D(T,v,a) for its observation. The channel types are defined in (2): Y:=Rℓ[T],C:=Rr[T],𝖯𝗋𝖾𝖽:=Rp[C].Y:=R_\ell[T],\qquad C:=R_r[T],\qquad \mathsf{Pred}:=R_p[C].(2) Reencoding between two odd paths builds along the target path using the source observation as its body.✦ For y:Yy:Y, c:Cc:C, p:𝖯𝗋𝖾𝖽p:\mathsf{Pred}, and a formula family FF on YY, the raw and logical reencodings, roundtrip, evaluation, and predicate constructor are given by raw⁡(y):=reenc⁡Rℓ,Rr(T,y),log⁡(c):=reenc⁡Rr,Rℓ(T,c),ρ(y):=log⁡(raw⁡(y)),p@y:=obs⁡Rp(C,p,raw⁡(y)),enc⁡(F):=build⁡Rp(C,c↦F(log⁡(c))).\begin{aligned} \operatorname{raw}(y)&:=\operatorname{reenc}_{R_\ell,R_r}(T,y),& \operatorname{log}(c)&:=\operatorname{reenc}_{R_r,R_\ell}(T,c),\\ \rho(y)&:=\operatorname{log}(\operatorname{raw}(y)),& p@y&:=\operatorname{obs}_{R_p}(C,p,\operatorname{raw}(y)),\\ \operatorname{enc}(F)&:= \operatorname{build}_{R_p} \bigl(C,c\mapsto F(\operatorname{log}(c))\bigr). \end{aligned}(3) The functions in (3) are written with named bound variables; the underlying terms use de Bruijn indices and newest-first contexts, with capture-avoiding renaming and substitution.

A positive data path is a finite path generated by the identity and positive extensions, retaining a domain annotation and its chosen dummy argument at each extension.✦ A formed positive data path starts with the identity at its source profile and permits restriction of its target to any subset. A feasible triple (A,J,K)(A,J,K) extends a path ending at JJ to KK using the fixed witness type and value at AA. The positive wrapper type of TT along a positive data path is TT at the identity; an extension by a domain DD forms the nondependent data arrow from DD to the preceding wrapper type.✦ Positive wrapper injection leaves a term unchanged at the identity path and, at an extension by a domain DD, forms the constant annotated data abstraction over DD with the preceding injection as its body.✦ Positive wrapper projection leaves a term unchanged at the identity path and, at each extension, applies it to the retained dummy argument before continuing along the preceding path.✦

The carrier of a data tail is the positive wrapper type of its terminal sort along its positive path.✦ Data tail encoding of a formula first wraps it along the tail’s encoding proof path and then takes its positive wrapper injection.✦ Data tail decoding first takes positive wrapper projection and then wraps the result along the tail’s decoding proof path.✦

Admissibility and Comparison

A stable formula family QQ over 𝖯𝗋𝖾𝖽\mathsf{Pred} assigns an interpretable formula to every typed data predicate and retains this property after every mode-preserving context embedding.✦ Such an embedding renames variables into a valid context, preserving their types and modes.✦ Interpretability requires sorting at the logical base profile, data mode, and stable witness extraction.✦ Stable witness extraction turns witness continuations into proof-mode inhabitants after every such embedding.✦

For data predicates p,q:𝖯𝗋𝖾𝖽p,q:\mathsf{Pred} and elements y,z:Yy,z:Y, the roundtrip invariance condition Resp⁡(p)\operatorname{Resp}(p), admissibility guard G(p)G(p), observational relation y∼zy\sim z, and predicate comparison p≈qp\approx q are the formulas in (4): Resp⁡(p):=∀KYy:Y((p@y)↔(p@ρ(y))),G(p):=Resp⁡(p)∧Q(p),y∼z:=∀KPp:𝖯𝗋𝖾𝖽(G(p)→((p@y)↔(p@z))),p≈q:=∀KYy:Y((p@y)↔(q@y)).\begin{aligned} \operatorname{Resp}(p) &:=\forall_{K_Y}y:Y\, \bigl((p@y)\leftrightarrow(p@\rho(y))\bigr),\\ G(p)&:=\operatorname{Resp}(p)\land Q(p),\\ y\sim z &:=\forall_{K_P}p:\mathsf{Pred}\, \bigl(G(p)\to((p@y)\leftrightarrow(p@z))\bigr),\\ p\approx q &:=\forall_{K_Y}y:Y\, \bigl((p@y)\leftrightarrow(q@y)\bigr). \end{aligned}(4) For elements y,z:Yy,z:Y, the class predicate cyc_y and class union predicate uyzu_{yz} are encoded predicates over YY with the bodies specified in (5): cy:=enc⁡(w↦w∼y),uyz:=enc⁡(w↦(w∼y)∨(w∼z)).c_y:=\operatorname{enc}(w\mapsto w\sim y),\qquad u_{yz}:=\operatorname{enc} \bigl(w\mapsto (w\sim y)\lor(w\sim z)\bigr).(5)

For a family of seven predicate queries Q0,…,Q6Q_0,\ldots,Q_6, take Q=Q0Q=Q_0 in the admissibility guard. The class test Is⁡t\operatorname{Is}_t on YY, for t=0,1,2t=0,1,2, is the formula obtained by applying query t+1t+1 to the class predicate of its argument.✦ For t=1,2t=1,2, the edge test TtT_t on two elements of YY is the formula combining their class tests with query t+3t+3 of the union of their classes.✦ The pair membership formula form⁡(p,a,z)\operatorname{form}(p,a,z) and the pair predicate Pair⁡(p,a)\operatorname{Pair}(p,a) over YY, for p:𝖯𝗋𝖾𝖽p:\mathsf{Pred} and a,z:Ya,z:Y, are defined from the class tests Is⁡t\operatorname{Is}_t and edge tests TtT_t in (6): Is⁡t(y):=Qt+1(cy),Tt(y,z):=Is⁡0(y)∧(Is⁡t(z)∧Qt+3(uyz)),form⁡(p,a,z):=(∃KYy:Y(T1(y,z)∧p@y))∨T2(a,z),Pair⁡(p,a):=enc⁡(z↦form⁡(p,a,z)).\begin{aligned} \operatorname{Is}_t(y)&:=Q_{t+1}(c_y),\\ T_t(y,z)&:=\operatorname{Is}_0(y)\land \bigl(\operatorname{Is}_t(z)\land Q_{t+3}(u_{yz})\bigr),\\ \operatorname{form}(p,a,z)&:= \bigl(\exists_{K_Y}y:Y\, (T_1(y,z)\land p@y)\bigr) \lor T_2(a,z),\\ \operatorname{Pair}(p,a)&:= \operatorname{enc}\bigl(z\mapsto\operatorname{form}(p,a,z)\bigr). \end{aligned}(6)

The Constructed Specialization

A specialized relational coding BB is a relational coding with a three-copy observation system and a chosen data type A0A_0.✦ The underlying relational coding of BB has a uniform probe, which supplies a predicate probe for each predicate over the common probe type, and shares the probe’s raw path with the observation channel. Write RR for this shared channel. The base type T0T_0 of a specialized relational coding BB is the type A0A_0 wrapped along the probe’s initial path.✦ Write ZZ for the common probe type and X=Rℓ[Z]X=R_\ell[Z] for its logical element type. The specialized predicate type is Rp[Rr[T0]]R_p[R_r[T_0]]✦; evaluation uses (3) at T0T_0.✦

For a probe and a type AA, the probe raw type is obtained by wrapping AA first along the initial path and then along the raw path.✦ A double channel consists of two odd retained data paths.✦ The probe double channel has the probe’s back path as its first path and its raw path as its second path.✦ The probe parameter type at AA is the double-channel parameter type at the probe raw type of AA, using the probe double channel.✦ The single bundle type at AA is the double wrapper carrier of the probe parameter type at AA, using the probe double channel.✦ The probe type parameter domain is the positive wrapper type of the seed sort along the probe’s lifting path.✦ The probe function type at AA is the nondependent data arrow from the single bundle type at AA to the carrier of the probe’s data tail.✦ The probe function abstraction with formula body FF is an annotated data abstraction over the single bundle type whose body is the data tail encoding of FF.✦ The wrapped probe abstraction at AA is obtained by applying the data path builder along the probe’s function path, at the probe function type of AA, to the probe function abstraction.✦ The uniform probe abstraction is an annotated data abstraction over the probe type parameter domain; at its bound parameter it takes positive wrapper projection along the lifting path and uses the resulting type in the wrapped probe abstraction.✦ Probe specialization of vv to AA applies vv in data mode to the positive wrapper injection of AA along the lifting path.✦ The probe application observation at AA specializes its probe to AA and observes the result along the function path, at the probe function type. Its test applies the newly supplied function to the argument in data mode and decodes the result through the data tail.✦

The tagged observation type YBY_B is formed by nested inner and outer double channels over ZZ. The embedding τt:X→YB\tau_t:X\to Y_B, for t=0,1,2t=0,1,2, reencodes into the first copy path and then embeds through the inner and outer channels, with the inner marker set precisely at t=1t=1 and the outer marker precisely at t=2t=2.✦ For tagged elements y,z:YBy,z:Y_B, tagged observational equivalence y≍zy\asymp z is the formula requiring equivalence of their inner markers, equivalence of their outer markers, and equivalence of all their observations by uniform predicate probes.✦ The double wrapper carrier of a type along a double channel is obtained by wrapping the type first along its first path and then along its second path.✦ Double-channel evaluation of a value against a formula body observes the value along the second path, using the builder along the first path applied to that body as its test.✦ The encoded bit observation of a double-channel value is its double channel evaluation against constant logical truth.✦ The signed payload embedding along a double channel builds along its second path from the observation of the newly bound first-path value against the payload. That observation is used unchanged when the sign is true and negated when it is false.✦ Signed payload recovery of a value against a formula body is its double-channel evaluation against that body when the encoded bit observation holds, and the negation of that evaluation otherwise; this choice is the logical disjunction of the two guarded branches.✦ The observation copy structure has the shared logical path as source, the probe’s raw path as first copy path, and the observation system’s inner and outer double channels as its two copy channels.✦

The iterated carrier of TT at depth zero is TT, and each successor depth applies the double wrapper carrier once more✦; the fourfold carrier is the iterated carrier at depth four.✦ The predicate code type at TT wraps the fourfold carrier of TT along the double channel’s first path.✦ Its bottom predicate is built along that path with constant body ⊥\bot.✦ The double-channel parameter type at TT wraps its predicate code type along the second path.✦

The query tagged embedding of a payload uses the four binary digits of a query index from zero through eight, listed from the lowest digit upward, as its signs for iterated signed embedding.✦ Iterated embedding leaves the payload unchanged for the empty sign list and, for a leading sign, first embeds along the remaining signs and then applies that leading sign at the corresponding iterated carrier.✦ The tag predicate for an empty bit string is ⊤\top. For a leading bit it conjoins the encoded bit observation, negated precisely when that bit is false, with signed recovery of the tag predicate for the remaining bits, at the carrier whose depth is the length of those remaining bits. The tagged payload encoder for a tag tt from zero through seven builds along the first path over the fourfold carrier. Its body is the disjunction of the zero-query tag conjoined with fourfold signed recovery of the observation of the payload along its odd payload path, and the query tag whose index is t+1t+1.✦ For an odd path UU, the payload decoder builds along UU from the observation of the encoded payload along the double channel’s first path, tested against the zero-query tagged embedding of the newly supplied argument.✦

The dummy observation at TT is built along the second path over the predicate code type with constant body ⊤\top.✦ The dummy parameter test observes a parameter along that second path against the bottom predicate at TT.✦ The tagged observation query at tag tt observes its observation term along the second path, over the predicate code type, against the tagged payload encoding of its predicate argument.✦

The encoded data point for marker c0c_0 and payload cc is built along the second path over the predicate code type. Its body conjoins the observation along the first path at the fourfold carrier against the tagged embedding of c0c_0 at query index eight with the observation of the decoded predicate argument against cc along the payload path.✦ The source data bundle of d,m,yd,m,y is built along the second path over the first wrapper type. For a first-path input, its body takes the observation against mm when the observation against dd holds, and the observation against yy otherwise.✦ The inner observation probe for a predicate pp is built along the first path over the parameter type. Its body is ⊥\bot when the dummy parameter test holds and otherwise is the tagged observation query at tag seven of the bound parameter against pp.✦ Bundle access observes its bundle along the second path at the first wrapper type, against a test built along the first path. At the bound input that test is ⊤\top when its guard holds and otherwise observes the bundle against the supplied inner data argument.✦ Recovery access at TT uses bundle access over the parameter type, with the dummy parameter test as guard and the inner observation probe applied to the supplied predicate as its inner data argument.✦

The three-copy observation of yy against aa first uses signed recovery through the outer channel at the middle copy carrier. Its test recovers through the inner channel at the first copy wrapper type and then observes against aa along the first copy path.

A formed specialization BB is a specialized relational coding with formed probe and observation channels, a chosen type A0A_0 of the probe’s seed sort in data mode, and matching channel shapes: YB=Rℓ[T0]Y_B=R_\ell[T_0], the outer second path is RℓR_\ell, and the outer first path sends the middle copy carrier over ZZ to T0T_0.

For a predicate ii over ZZ and x:Xx:X, the probe order i≼xi\preccurlyeq x is the observation along the logical path of xx against the uniform predicate probe for ii.✦ Probe element equivalence on XX is the relation that holds between two elements when their probe order observations are equivalent at every predicate over ZZ.✦ For a predicate ii over ZZ, the condition Ext⁡(i)\operatorname{Ext}(i) requires evaluation by ii to respect probe element equivalence.✦ The condition Valid⁡(x)\operatorname{Valid}(x) for an element x:Xx:X requires the probe order observation i≼xi\preccurlyeq x to respect comparison of predicates ii over ZZ.✦

The relational point encoding of yy at AA is the encoded data point in the probe double channel at the probe raw type of AA, using the predicate path and constant truth marker, with the raw channel encoding of yy as payload.✦ The relational argument bundle of mm and yy at AA is the source data bundle in the probe double channel at the probe parameter type of AA, with entries the dummy observation at the probe raw type of AA, mm, and the relational point encoding of yy.✦ The diamond observation of a probe vv at AA, parameter mm, and element yy is its probe application observation against the relational argument bundle of mm and yy.✦ The probe section predicate of vv at AA and mm is the encoded predicate whose observation body sends yy to the diamond observation of vv at A,m,yA,m,y.✦ The section pair query for A,m,v,aA,m,v,a is the tagged observation query at tag six of mm against the pair predicate formed from the probe section predicate of vv and aa, using the probe double channel and predicate path at the probe raw type of AA.✦

The relational element lift of aa at A,mA,m is built along the logical path over the common probe type, with body assigning to each probe its section pair query for A,m,aA,m,a.✦ The predicate pullback of ii along lifting at A,mA,m is the encoded predicate whose body evaluates ii on the relational element lift of its argument.✦ Predicate pullback recovery of ii from a single-bundle argument at AA uses recovery access in the probe double channel at the probe raw type of AA, with the predicate path as payload path. Its supplied predicate sends a parameter mm to the pullback of ii along lifting at A,mA,m.✦

The structural parameter m0m_0 is the structural query realization in the probe double channel at the probe raw type of A0A_0, using the predicate path, the constant truth marker, and the seven structural properties. For a specialized predicate vv, these queries are determined by the observation invariance property H(v)H(v), class representation properties Ct(v)C_t(v), edge representation properties Et(v)E_t(v), and relational closure property D(v)D(v) in (7). The left copy predicate vℓv^\ell of a specialized predicate vv is obtained with the predicate constructor enc⁡Z\operatorname{enc}_Z over ZZ in (7); the indices are t=0,1,2t=0,1,2 for CtC_t and t=1,2t=1,2 for EtE_t: H(v):=∀KYy,z:YB(y≍z→((v@y)↔(v@z))),Ct(v):=∃KYx:X∀KYw:YB((v@w)↔(w≍τt(x))),Et(v):=∃KYx:X∀KYw:YB((v@w)↔((w≍τ0(x))∨(w≍τt(x)))),vℓ:=enc⁡Z(x↦v@τ1(x)),D(v):=Ext⁡(vℓ)→∀KYx:X(Valid⁡(x)→((v@τ2(x))→(vℓ≼x))).\begin{aligned} H(v)&:=\forall_{K_Y}y,z:Y_B\, \bigl(y\asymp z\to((v@y)\leftrightarrow(v@z))\bigr),\\ C_t(v)&:=\exists_{K_Y}x:X\,\forall_{K_Y}w:Y_B\, \bigl((v@w)\leftrightarrow(w\asymp\tau_t(x))\bigr),\\ E_t(v)&:=\exists_{K_Y}x:X\,\forall_{K_Y}w:Y_B\, \bigl((v@w)\leftrightarrow ((w\asymp\tau_0(x))\lor(w\asymp\tau_t(x)))\bigr),\\ v^\ell&:=\operatorname{enc}_Z(x\mapsto v@\tau_1(x)),\\ D(v)&:=\operatorname{Ext}(v^\ell)\to \forall_{K_Y}x:X\, \bigl(\operatorname{Valid}(x)\to ((v@\tau_2(x))\to(v^\ell\preccurlyeq x))\bigr). \end{aligned}(7) The seven structural properties, in order, are H,C0,C1,C2,E1,E2,DH,C_0,C_1,C_2,E_1,E_2,D. The specialized queries QtQ_t are the structural queries of m0m_0; the specialized guard is Resp⁡(v)∧Q0(v)\operatorname{Resp}(v)\land Q_0(v). Specialized observational comparison and pairing use these queries and this guard in (4) and (6). The structural query realization is built along the second path over the predicate code type from a finite marked disjunction of the seven structural properties. In the ordered branch for tag tt, the marker tests the tagged embedding of the chosen marker at query index t+1t+1, and the property is applied to the decoded predicate; the disjunction ends with ⊥\bot.

The Claim

Theorem 1. Suppose the observation channels are formed, QQ is a stable formula family over 𝖯𝗋𝖾𝖽\mathsf{Pred}, and the element and predicate quantifier routes are fixed. Suppose also that Γ⊢C:s\Gamma\vdash C:s for some sort ss. Let y,z,wy,z,w have type YY in Γ\Gamma and have data mode. Class evaluation satisfies (8): (cy@w)≃Γ,γ(w∼y).(c_y@w)\simeq_{\Gamma,\gamma}(w\sim y).(8)✦ Union evaluation satisfies (9): (uyz@w)≃Γ,γ((w∼y)∨(w∼z)).(u_{yz}@w)\simeq_{\Gamma,\gamma} \bigl((w\sim y)\lor(w\sim z)\bigr).(9)✦ If Γ;γ⊩y∼z\Gamma;\gamma\Vdash y\sim z, class comparison gives (10): Γ;γ⊩cy≈cz.\Gamma;\gamma\Vdash c_y\approx c_z.(10)✦

Let BB be a formed specialized relational coding, and use its specialized types, guard, and queries. Let p,p′p,p' have its predicate type and a,a′,za,a',z its tagged observation type, all in data mode in Γ\Gamma. Suppose Γ;γ⊩p≈p′\Gamma;\gamma\Vdash p\approx p' and Γ;γ⊩a∼a′\Gamma;\gamma\Vdash a\sim a'. Pair input evaluation satisfies (11): (Pair⁡(p,a)@z)≃Γ,γ(Pair⁡(p′,a′)@z).\bigl(\operatorname{Pair}(p,a)@z\bigr) \simeq_{\Gamma,\gamma} \bigl(\operatorname{Pair}(p',a')@z\bigr).(11)✦ Pair comparison gives (12): Γ;γ⊩Pair⁡(p,a)≈Pair⁡(p′,a′).\Gamma;\gamma\Vdash \operatorname{Pair}(p,a)\approx\operatorname{Pair}(p',a').(12)✦ For any data-mode p,a,zp,a,z of the indicated types, pair evaluation satisfies (13): (Pair⁡(p,a)@z)≃Γ,γform⁡(p,a,z).\bigl(\operatorname{Pair}(p,a)@z\bigr) \simeq_{\Gamma,\gamma}\operatorname{form}(p,a,z).(13)✦

In the paper, the result reads: cy@w≃(w∼y)c_y@w\simeq(w\sim y) and uyz@w≃((w∼y)∨(w∼z))u_{yz}@w\simeq((w\sim y)\lor(w\sim z)) are unrounded evaluation equivalences; parameters y∼y′y\sim y' and z∼z′z\sim z' give cy≈cy′c_y\approx c_{y'} and uyz≈uy′z′u_{yz}\approx u_{y'z'}; under the admissibility hypothesis ad:Admad:\operatorname{Adm}, the tests Is⁡t\operatorname{Is}_t and TtT_t respect ∼\sim, and assumptions p≈p′p\approx p', a∼a′a\sim a', z∼z′z\sim z' give form⁡(p,a,z)≃form⁡(p′,a′,z′)\operatorname{form}(p,a,z)\simeq\operatorname{form}(p',a',z'), while the first two give Pair⁡(p,a)≈Pair⁡(p′,a′)\operatorname{Pair}(p,a)\approx\operatorname{Pair}(p',a'), with pv⁡(p,a,z):Pair⁡(p,a)@z≃form⁡(p,a,z)\operatorname{pv}(p,a,z): \operatorname{Pair}(p,a)@z\simeq\operatorname{form}(p,a,z) under the same admissibility hypothesis.

What Is Checked

The channel statements concern formed logical, raw, and predicate paths from profiles II to JJ, II to KK, and KK to HH, respectively, with the stable family and data-mode typing hypotheses in the theorem. Specialized formation requires formed underlying probe and copy channels, Γ⊢A0:s0\Gamma\vdash A_0:s_0 in data mode for the probe’s seed sort s0s_0, and the equality YB=Rℓ[T0]Y_B=R_\ell[T_0]; it also requires the outer second path to equal RℓR_\ell and the outer first path, applied to the middle copy carrier over ZZ, to give T0T_0.✦ The pair statements are over the formed specialized relational coding, its constructed admissibility guard and seven property queries, the fixed logical interpretation and element and predicate quantifier routes, and typed data-mode predicates and tagged observations. The congruence statements also assume proof-mode inhabitants of the indicated predicate and observation comparisons.