Pushforwards in Inverse Homotopical Diagrams

Chris Kapulkin; Yufeng Li

REVIEW 3 major objections 4 minor 4 references

Pushforwards in Inverse Homotopical Diagrams

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proves that in a category of fibrant objects satisfying natural pullback-pushforward preservation conditions, homotopical Reedy fibrations are closed under pushforward along homotopical Reedy fibrations whenever the indexing…

desk verdict Genuinely new sufficient condition for pushforward closure in homotopical inverse diagrams, but the advertised scope is wider than the proved theorem. read the letter →

arxiv 2506.04472 v1 pith:PZT4ZOUC submitted 2025-06-04 math.CT math.AT

Chris Kapulkin , Yufeng Li This is my paper

classification math.CTmath.AT MSC 18G5555U35

keywords categoriesoffibrantobjectsinverseReedyfibrationshomotopicaldiagramspushforwardsdependenttypetheorylocallycartesianclosedalgebra

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Working with categories of fibrant objects, this paper asks when the homotopical inverse diagrams in $C$ — diagrams that send weak equivalences of the indexing category to weak equivalences of $C$ — are closed under pushforward inside the category of all inverse diagrams. It proves a sufficient condition on the indexing inverse category $I$: for every object $i$, any nonidentity weak equivalence out of $i$ must be the initial object among the nonidentity maps out of $i$, or all maps out of $i$ must be weak equivalences. Under that condition, pushforwards of homotopical Reedy fibrations along homotopical Reedy fibrations are again homotopical Reedy fibrations. This simultaneously covers the two previously studied extreme cases, where $I$ has no weak equivalences or all maps are weak equivalences, and it is motivated by applications to dependent type theory and locally cartesian closed structure.

What carries the argument

The machinery is the inductive pushforward formula for inverse diagrams (Lemma 2.2, rephrased from [FKL24]): the component $(p_*C)_i$ of the pushforward along $p:B\to A$ is the pullback over $A_i$ of $(p_i)_*C_i$ against a limit over all nonidentity maps $f:i\to j$ of the pullbacks $A_f^*(p_*C)_j$, with comparison maps $(B_f^*\mathrm{ev})^\dagger$ connecting to $(p_i)_*B_f^*C_j$. The other load-bearing pieces are the distributive law of Lemma 2.3, which lets the proof compute the relative matching map of the pushforward as a pushforward of the original relative matching map, and Proposition 2.4, which says matching-object functors preserve fibrations. The new combinatorial condition on $I$ — every nonidentity weak equivalence out of $i$ is initial in the slice of nonidentity maps out of $i$, or all maps out of $i$ are weak equivalences — is precisely the condition that makes every comparison map in the inductive step a weak equivalence, so the homotopical property survives pushforward.

What would settle it

Take $I$ to be the span shape $0 \leftarrow 01 \to 1$ with only the two arrows declared weak equivalences, and take $C=\mathbf{Set}$ with bijections as weak equivalences and all maps as fibrations. This shape violates the theorem's condition, and Example 3.7 shows that the pushforward (the exponential span) of two homotopical spans can fail to be homotopical; this failure is a direct counterexample to closure, so a correct necessary-and-sufficient version of the theorem would have to detect it.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 3.5: if $C$ has fibrations and weak equivalences satisfying the 'homotopically logically behaved' Assumption 3.2, and if pushforwards of fibrations along fibrations exist in $C$, then for any finite inverse category $I$ whose weak equivalences satisfy the initial-or-all-weak-equivalences condition, the pushforward of any homotopical Reedy fibration along any homotopical Reedy fibration is a homotopical Reedy fibration. Here homotopical means the diagram sends weak equivalences of $I$ to weak equivalences of $C$, and Reedy fibrancy is measured by the usual matching-object maps. The proof constructs the pushforward objectwise by an inductive pullback formula and shows, by induction on the degree of $I$, that every comparison map appearing in that formula is a weak equivalence; the shape condition is exactly what supplies the needed 2-out-of-3 steps.

Load-bearing premise

The load-bearing premise is that the ambient category $C$ satisfies Assumption 3.2: weak equivalences are preserved by pullback along fibrations, pullbacks and pushforwards preserve weak equivalences between fibrations, and certain precomposition maps are weak equivalences, together with the existence of pushforwards of fibrations along fibrations. The paper asserts but does not verify that these conditions hold for its intended examples.

Editorial extensions

If this is right

If $C$ satisfies Assumption 3.2 and pushforwards along fibrations exist, then homotopical Reedy fibrant diagrams are closed under pushforward along homotopical Reedy fibrations for every finite inverse category $I$ meeting the initial-or-all-weak-equivalences condition.
The theorem covers the two extreme cases previously known, namely $I$ with no weak equivalences and $I$ with all maps weak equivalences, as instances of the same shape condition.
In dependent type theory, this supplies a construction principle for made-to-order models of type theory, because closure under pushforwards lets one impose conditions on propositions while staying in a fibration category.
The result advances the program of showing that suitable locally cartesian closed categories of fibrant objects present the same homotopy theory as locally cartesian closed $(\infty,1)$-categories.
The counterexamples in Example 3.7 show that without the condition, homotopical diagrams are not closed under pushforwards, so the $(\infty,1)$-category of $(\infty,1)$-functors need not be a left exact localization of the $(\infty,1)$-category of ordinary functors.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The condition can be read as a combinatorial description of the branch points in the indexing shape: at any object with a weak equivalence leaving it, either that weak map is the canonical first exit or every exit is weak; this suggests the theorem is close to optimal among shape conditions stated only in terms of $I$.
Because Assumption 3.2 packages several preservation properties, a natural next step is to weaken it and see whether a smaller set of preservation properties still suffices for the same shape condition; the 2-out-of-3 steps in the proof indicate which property is load-bearing at each stage.
A direct way to test how sharp the theorem is would be to classify all finite inverse categories $I$ for which homotopical Reedy fibrant diagrams are closed under pushforwards in $\mathbf{Set}$ with bijections, showing whether the condition is also necessary or merely sufficient.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The new thing here is Theorem 3.5: a shape condition on the indexing category I under which pushforwards of homotopical Reedy fibrations remain homotopical Reedy fibrations. It sits between Shulman's case (no maps in I are weak equivalences) and Kapulkin-Lumsdaine's case (all maps are weak equivalences), and the condition itself is new. The proof is a careful inductive extension of KL21, using the pushforward construction from Fiore-Kapulkin-Li. The paper is well organized, and the examples and counterexamples in 3.6 and 3.7 are genuinely illuminating, especially the span counterexample showing why the obvious failure occurs.

What the paper does well: it isolates the precise obstruction and gives a sufficient condition that is permissive enough to cover many shapes beyond the two known extremes. The lemmas leading up to the theorem are clean, and the distributive law (Lemma 2.3) is a useful packaging of a standard argument. The counterexamples are concrete and easy to check.

The soft spots are real but not fatal. First, the abstract says the result holds "in a fibration category," but the actual theorem requires Assumption 3.2: pullbacks and pushforwards must be homotopically logically behaved. That is substantially more than the axioms of a Brown category of fibrant objects. The introduction asserts these conditions hold for type-theoretic fibration categories and, where appropriate, general categories of fibrant objects, but no proof or reference is supplied for the second and fourth bullets of Assumption 3.2. This matters because the paper's advertised applications depend on those instances. The theorem is explicitly conditional, so the logic is fine, but the scope is overstated. A referee should ask the authors to either prove the verification or point to a precise statement in the literature.

Second, Theorem 3.5 as stated omits the finiteness hypothesis on I that Proposition 2.5 and Lemma 2.2 rely on. Presumably the standing assumption from Section 1 is still in force, but the theorem and abstract should say so. This is a minor presentational gap.

Who is this for? People working on homotopical diagram categories, dependent type theory, and locally cartesian closed fibration categories. The result is worth having even in its conditional form. I would send it to a serious referee: the argument is plausible, the new condition is meaningful, and the gap about Assumption 3.2 is fixable. My own verdict is conditional, leaning positive. Recommend major revision with attention to the logical-behaviour verification and a precise statement of hypotheses.

Referee Report

3 major / 4 minor

Summary. The paper studies pushforwards in inverse diagram categories C^I. It recalls the inductive construction of pushforwards from Fiore-Kapulkin-Li (Lemma 2.2), proves a distributive law (Lemma 2.3), and shows that in a category with fibrations closed under pushforward along fibrations, Reedy fibrations are stable under pushforwards along Reedy fibrations (Proposition 2.5). The main result, Theorem 3.5, gives a sufficient condition on the indexing category I: if every non-identity weak equivalence out of an object i is either initial in the boundary of i or coexists only with weak equivalences out of i, then pushforwards of homotopical Reedy fibrations along homotopical Reedy fibrations are again homotopical Reedy fibrations, assuming the homotopical logical behaviour conditions of Assumption 3.2. Examples and counterexamples of indexing shapes are discussed in Examples 3.6 and 3.7.

Significance. The result is a genuine contribution: it interpolates between the known extremal cases where I has no weak equivalences (Shulman) and where all maps are weak equivalences (Kapulkin-Lumsdaine), giving a new condition on I. The proof is carefully built on the published FKL24 construction and Shulman's Lemma 11.8, and the conditional statement appears plausible. The paper also provides instructive counterexamples showing necessity of the hypotheses. Its main weakness is that the key Assumption 3.2 is not verified for the advertised examples, so the paper currently establishes a conditional theorem without demonstrated instances.

major comments (3)

[Abstract; Introduction, p. 2; Assumption 3.2] The abstract states the result for "a fibration category", but Theorem 3.5 relies on Assumption 3.2, which imposes four nonstandard homotopical-logical conditions beyond the axioms of a fibration category (e.g., pullbacks preserve weak equivalences between fibrations, and the precomposition map (g^*ev)^† is a weak equivalence). The introduction asserts without proof or reference that these conditions hold for type-theoretic fibration categories and, when appropriate, categories of fibrant objects. As a result, the paper currently establishes a conditional theorem with no verified instances. Please add proofs or precise references for these examples, or qualify the abstract and introduction so that the theorem's hypotheses match its stated scope.
[Theorem 3.5, proof, second case] In the second case of the proof, the map (p_i)_*(k_i,C_w) is asserted to be a trivial fibration without explanation. For this it must be shown that (k_i,C_w) is both a fibration and a weak equivalence: the former follows from the fact that k: C ↠ B is a Reedy fibration (the map (k_i,C_w) is a pullback of a relative matching map), and the latter from homotopicalness of C and B together with 2-out-of-3. The argument should state these facts explicitly, since the conclusion that (p_*C)_w is a weak equivalence depends on this step.
[Lemma 3.4] The proof of Lemma 3.4 invokes "right properness" without defining it or giving a reference. If this means weak equivalences are stable under pullback along fibrations (a standard axiom for categories of fibrant objects), the authors should say so; if it is an additional assumption, it should be included in Assumption 3.2 or stated as a hypothesis of the lemma. This is important because the proof of the "all maps are weak equivalences" case of Theorem 3.5 depends on this property.

minor comments (4)

[Throughout] There are several typos and formatting issues: "digrams" on p. 1, "pushfowards" in the Section 2 heading, "finitely completely category" in Lemma 2.2, and the stray "C I/ /B" in Lemma 2.2. These should be corrected.
[Introduction, p. 1] The phrase "in the proof of homotopy canonicity by the first-named author and Sattler" is missing a citation; a reference should be supplied.
[Example 3.7] The counterexample is described too tersely: the notation Set^→ (or Set with arrows) is not defined, and the diagram would benefit from a more explicit description of the exponential span [A,B] and why it fails to be homotopical.
[Proposition 2.4 and Theorem 3.5] The finiteness hypotheses are not consistently stated: Proposition 2.4 assumes each underslice of I is finite, Proposition 2.5 assumes I is finite, and Theorem 3.5 does not mention finiteness at all. The theorem should explicitly list all standing hypotheses (finiteness, existence of pushforwards along fibrations, Assumption 3.2) to be self-contained.

Circularity Check

0 steps flagged · score 2.0 of 10

Theorem 3.5 is proven from stated base-category hypotheses, not from its conclusion; the new shape condition has genuinely independent content, and the self-citations to [FKL24]/[KL21] function as tools with hypotheses that do not contain the target result. The unverified scope of Assumption 3.2 is an applicability gap, not a circularity.

full rationale

The paper's derivation chain is a genuine conditional proof: Theorem 3.5 derives closure of homotopical Reedy fibrations under pushforwards from Assumption 3.2 (homotopically logically behaved pullbacks/pushforwards in the base category C) and a shape condition on the indexing category I, via Lemmas 2.2, 2.3, 3.3, 3.4 and Propositions 2.4, 2.5. No hypothesis is equivalent by construction to the conclusion: Assumption 3.2 is stated at the base-category level, the theorem's conclusion is a diagram-category closure property, and the proof derives, rather than assumes, that property. The sufficient condition on I is new and non-vacuous: it strictly generalizes the two previously known extreme cases (Shulman's no-weak-equivalence case and the all-maps-weak-equivalence case of [KL21]), and Example 3.7 exhibits shapes where the condition fails together with a concrete counterexample in Set showing the conclusion genuinely fails, so the theorem is not a restatement of its inputs. The self-citations are used as tools: Lemma 2.2 (the inductive pushforward construction) is cited from [FKL24], whose authors overlap with the present paper, but it is a construction with its own stated hypotheses (existence of base-level pushforwards) that do not include the target result, and the paper reproduces a proof sketch of it; [KL21] is explicitly used for its method ('revisiting the proof'), not imported as the theorem. Per the hard rules, these are real evidence and do not constitute circularity. One flagged gap, weighed here: the introduction asserts without proof that the assumptions are 'satisfied by type-theoretic fibration categories [Shu15] and, when appropriate, general categories of fibrant objects [Bro73]', but Assumption 3.2's two nonstandard conditions (pullbacks preserve weak equivalences between fibrations, and the 'precomposition' map (g*ev)^dagger is a weak equivalence) are not verified or referenced for those examples. This is a correctness/applicability risk for the advertised motivating instances, not a circularity: the theorem is honestly conditional, and the failure of an assumption would limit scope rather than expose a self-referential derivation.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper's central claim rests on the ambient fibration category axioms, the existence and fibrancy of pushforwards, the 'homotopically logically behaved' conditions (Assumption 3.2), and the shape condition on I in Theorem 3.5. No numerical free parameters or invented entities appear.

assumptions (5)

domain assumption C is a category of fibrant objects in the sense of Brown, with finite limits, a wide subcategory of fibrations, and all objects fibrant.
The paper works throughout in a fibration category C, citing Brown's theory. This is the ambient framework for all definitions and theorems.
domain assumption I is a finite inverse category and each underslice of I is finite.
Definitions 1.1 through 1.3 and Proposition 2.4 require finiteness of inverse categories and underslices to define coskeletons and matching objects.
domain assumption Pushforwards of fibrations along fibrations exist in C and are fibrations.
Assumed before Proposition 2.5 and used in Lemma 2.2 and Theorem 3.5 to guarantee the pushforward of a Reedy fibration remains a Reedy fibration.
ad hoc to paper Assumption 3.2: pullbacks and pushforwards in C are homotopically logically behaved, meaning weak equivalences are preserved by pullback along fibrations, pullbacks preserve weak equivalences between fibrations, pushforwards along fibrations preserve weak equivalences between fibrations, and the…
Introduced in Section 3 specifically to make Theorem 3.5 go through. The paper asserts but does not prove that these conditions hold for type-theoretic fibration categories.
ad hoc to paper Condition on I in Theorem 3.5: for each object i, if a nonidentity weak equivalence w:i to j exists, then either all nonidentity maps out of i are weak equivalences or w is initial in the category of nonidentity maps out of i.
This is the sufficient condition on the indexing category stated in Theorem 3.5. It is the main new hypothesis of the paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pushforwards in Inverse Homotopical Diagrams." pith.science (2026). https://pith.science/paper/PZT4ZOUC

@misc{pith2026250604472,
  author       = {Pith},
  title        = {Pith review of: Pushforwards in Inverse Homotopical Diagrams},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PZT4ZOUC}},
  note         = {Machine review of arXiv:2506.04472}
}

read the original abstract

We establish a sufficient condition for the category of homotopical inverse diagrams to be closed under pushforward inside the category of inverse diagrams in a fibration category.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

4 extracted references · 3 canonical work pages

[1]

Homotopy limits in type theory

[AKL15] Jeremy Avigad, Krzysztof Kapulkin, and Peter Lefanu Lumsdaine. “Homotopy limits in type theory”. In: Math. Structures Comput. Sci. 25.5 (2015), pp. 1040–1070. [Bro73] Kenneth S. Brown. “Abstract homotopy theory and generalized sheaf cohomology”. In: Trans. Amer. Math. Soc. 186 (1973), pp. 419–458. [Cis19] Denis-Charles Cisinski. Higher categories ...

work page 2015
[180]

Cubical setting for discrete homotopy theory, revisited

Cam- bridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2019, pp. xviii+430. [CK24] Daniel Carranza and Krzysztof Kapulkin. “Cubical setting for discrete homotopy theory, revisited”. In: Compos. Math. 160.12 (2024), pp. 2856–2903. 12 [FKL24] Marcelo Fiore, Krzysztof Kapulkin, and Yufeng Li. Logical Structure on Inverse Functor ...

work page 2024
[2009]

Univalence for inverse diagrams and homotopy canonicity

eprint: math/0610009. [Shu15] Michael Shulman. “Univalence for inverse diagrams and homotopy canonicity”. In: Mathematical Structures in Computer Science 25.5 (2015). [Szu16] Karol Szumiło. “Homotopy theory of cofibration categories”. In: Homology Homo- topy Appl. 18.2 (2016), pp. 345–357. [Szu17] Karol Szumiło. “Homotopy theory of cocomplete quasicategor...

arXiv 2015
[2024]

Logical Structure on Inverse Functor Categories

arXiv: 2410.11728 [math.CT]. [GK13] Nicola Gambino and Joachim Kock. “Polynomial functors and polynomial monads”. In: Mathematical Proceedings of the Cambridge Philosophical Society 154.1 (2013). [Kap17] Krzysztof Kapulkin. “Locally cartesian closed quasi-categories from type theory”. In: J. Topol. 10.4 (2017), pp. 1029–1049. [KL18] Krzysztof Kapulkin and...

work page Pith review arXiv 2013

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Homotopy limits in type theory

[AKL15] Jeremy Avigad, Krzysztof Kapulkin, and Peter Lefanu Lumsdaine. “Homotopy limits in type theory”. In: Math. Structures Comput. Sci. 25.5 (2015), pp. 1040–1070. [Bro73] Kenneth S. Brown. “Abstract homotopy theory and generalized sheaf cohomology”. In: Trans. Amer. Math. Soc. 186 (1973), pp. 419–458. [Cis19] Denis-Charles Cisinski. Higher categories ...

work page 2015

[2] [180]

Cubical setting for discrete homotopy theory, revisited

Cam- bridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2019, pp. xviii+430. [CK24] Daniel Carranza and Krzysztof Kapulkin. “Cubical setting for discrete homotopy theory, revisited”. In: Compos. Math. 160.12 (2024), pp. 2856–2903. 12 [FKL24] Marcelo Fiore, Krzysztof Kapulkin, and Yufeng Li. Logical Structure on Inverse Functor ...

work page 2024

[3] [2009]

Univalence for inverse diagrams and homotopy canonicity

eprint: math/0610009. [Shu15] Michael Shulman. “Univalence for inverse diagrams and homotopy canonicity”. In: Mathematical Structures in Computer Science 25.5 (2015). [Szu16] Karol Szumiło. “Homotopy theory of cofibration categories”. In: Homology Homo- topy Appl. 18.2 (2016), pp. 345–357. [Szu17] Karol Szumiło. “Homotopy theory of cocomplete quasicategor...

arXiv 2015

[4] [2024]

Logical Structure on Inverse Functor Categories

arXiv: 2410.11728 [math.CT]. [GK13] Nicola Gambino and Joachim Kock. “Polynomial functors and polynomial monads”. In: Mathematical Proceedings of the Cambridge Philosophical Society 154.1 (2013). [Kap17] Krzysztof Kapulkin. “Locally cartesian closed quasi-categories from type theory”. In: J. Topol. 10.4 (2017), pp. 1029–1049. [KL18] Krzysztof Kapulkin and...

work page Pith review arXiv 2013