{"id":"73187f7b-2ae9-469b-98bd-69ccfe5ff8e7","arxiv_id":"2412.11283","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Cyclic polytopes have infinitely many algebraically independent volume invariants coming from iterated-integral signatures.","lead":"This paper shows that cyclic polytopes carry infinitely many algebraically independent invariants built from iterated-integral signatures of paths along their edges. The result gives a new algebraic handle on cyclic polytopes and connects polytope combinatorics to shuffle-algebra invariant theory.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Definition 4.3's antipode sign is dimension-dependent, not length-dependent; as printed it makes vol3 fail to be time-reversal invariant and contradicts Proposition 4.5.","rationale":"Read in good faith, the paper's central claim is that Inv^d contains infinitely many algebraically independent shuffle elements. The proof splits by parity: the odd-d half relies on the identification Inv^d_{>=d+3} = TimeRevInv^d (or the whole algebra), and the even-d half on an intersection with TimeRevInv^d and on LoopClosureInv^d. The weakest spot is the definition of TimeRevInv^d: the printed antipode sign cannot be right, because time reversal of a word of length k multiplies by (-1)^k, not by (-1)^{d+1}. Using the printed formula yields an immediate contradiction in d=3 between Propositions 3.12 and 4.5. Since TimeRevInv is used in the proof of Theorem 4.9, this is a genuine load-bearing issue. It is likely a typographical slip rather than a fatal mathematical error: the standard length-dependent antipode makes vol3 time-reversal invariant exactly as Proposition 4.5 would require. I agree with the Reader's CONDITIONAL verdict: the paper is plausible and the machinery is coherent, but as written it needs correction and independent verification of the companion paper [19] for the even-dimensional case. The concern does not change the verdict, but it does sharpen the list of required revisions.","tokens_in":14980,"tokens_out":23553,"duration_ms":219201,"concrete_test":"Take d=3 and w=vol3 from Equation (5). Evaluate A(w) under the printed Definition 4.3: since d+1=4, A(vol3) = reverse(vol3) = -vol3, so w is not fixed, contradicting vol3 in Inv^3_{>=6} = TimeRevInv^3 from Propositions 3.12 and 4.5. The decisive check is to recompute Definition 4.3 with the standard length-dependent antipode A(w) = (-1)^{|w|} w^rev and re-run Propositions 4.5 and 4.8; if the corrected sign restores vol3 in TimeRevInv^3 and the [19] citations check out, the main theorem stands.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Definition 4.3 defines TimeRevInv^d = {w | w = Aw} with A(w) = (-1)^{d+1} w^rev, a sign that depends on the dimension d rather than on the length of w. Time reversal of a word of length k acts by A(w) = (-1)^k w^rev. For d=3 the printed formula gives A(vol3) = +vol3^rev = -vol3, so vol3 is not in TimeRevInv^3. But Proposition 3.12 shows vol3 is in Inv^3_n for all n >= 4, hence vol3 is in Inv^3 and therefore in Inv^3_{>=6}; Proposition 4.5 states Inv^3_{>=6} = TimeRevInv^3. Thus the written definitions make the paper internally inconsistent at exactly the point where the proof of Theorem 4.9 uses TimeRevInv for the odd-d case and in the even-d intersection. This is a load-bearing defect, although it can likely be repaired by replacing (-1)^{d+1} with the standard length-dependent sign (-1)^{|w|}. Separately, the even-d part of Theorem 4.9 still relies on [19, Prop 4.3, Lemma 4.7] and on an algebraic-independence statement imported from [19], so a complete verification requires checking the companion results as well.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies linear combinations of iterated-integral signatures of piecewise linear paths through the vertices of a cyclic polytope, focusing on those combinations invariant under the stabilizer C^d_n of positive matrices. The authors describe C^d_n explicitly (Propositions 3.8 and 3.10), prove that the signed volume is the basic invariant (Proposition 3.12), and then characterize the intersection Inv^d_{\\ge d+3} in terms of time-reversal invariants for odd d (Proposition 4.5) and loop-closure invariants for even d (Proposition 4.8). The main result, Theorem 4.9, claims that Inv^d_{\\ge d+3} \\cap I(PL^d_{d+2}), and in particular the ring Inv^d of volume invariants, contains infinitely many algebraically independent elements with respect to the shuffle product, hence is infinitely generated.","tokens_in":15312,"tokens_out":14116,"duration_ms":123793,"significance":"If the main theorem is correct, it establishes that the ring of volume invariants of cyclic polytopes is infinite-dimensional in every dimension, a genuinely new structural result that connects rough-path theory, convex geometry, and invariant theory. The paper carefully computes the stabilizer groups C^d_n, gives a clean formulation of the volume as an iterated integral, and provides explicit invariant polynomials in low-degree cases; these computations are valuable and no numerical fitting is involved. The heavy reliance on the companion paper [19] and a compressed transcendence-degree argument, however, mean that the main theorem is not fully self-contained as written.","major_comments":[{"comment":"Definition 4.3 defines the antipode A by A(w)=(-1)^{d+1} w^rev, with the sign depending on the ambient dimension d rather than on the length |w| of the word. The standard antipode of the concatenation Hopf algebra is A(w)=(-1)^{|w|} w^rev, and it is this length-dependent map that is adjoint to time reversal of paths. With the printed sign, for d=3 one has A(vol3)=-vol3, so vol3 is not in TimeRevInv^3. But Proposition 3.12 implies vol3 lies in Inv^3_n for all n\\ge 4, hence in Inv^3_{\\ge 6}, while Proposition 4.5 asserts Inv^3_{\\ge 6}=TimeRevInv^3. The written definitions are therefore internally inconsistent at exactly the point used in the proof of Theorem 4.9, both in the odd case and in the even-dimensional intersection. This is repairable by replacing (-1)^{d+1} with (-1)^{|w|} throughout Definition 4.3 and its subsequent uses, but as it stands it is a load-bearing defect.","section":"Definition 4.3"},{"comment":"The even-dimensional part of the main theorem depends on three results imported from the companion paper [19]: the identification of rotation-invariant signature elements with LoopClosureInv^d (cited as [19, Prop. 4.3]), the adjoint property of left and right loop closures (cited as [19, Lemma 4.7]), and the statement that LoopClosureInv^d contains an infinite algebraically independent subset. None of these is proved in the present manuscript. Since these statements carry half of the proof of Theorem 4.9, the main claim is conditional on [19]. The authors should either prove these results in an appendix or quote them in full with explicit statements and hypotheses so that the even-dimensional case can be verified independently.","section":"Theorem 4.9 and Proposition 4.8"},{"comment":"The transcendence-degree argument in the last paragraph of the proof is too compressed to be checked. In particular, the claim that, if the kernel of the composition R[s_1,s_2,\\ldots]\\to R\\langle 1,\\ldots,d\\rangle/I(PL^d_{d+2}) is algebraic over R[s_1,\\ldots,s_N], then one obtains an injection R[s_{N+1},s_{N+2},\\ldots]\\to R[x_1,\\ldots,x_{d+2}] does not follow: the composition may have a nonzero kernel on the tail variables, and the image need not contain a polynomial subring on infinitely many variables. A rigorous argument using the Krull dimension or transcendence degree of the image is needed to conclude that the kernel contains infinitely many algebraically independent elements. This step is load-bearing for the theorem's main claim, so it must be repaired or expanded.","section":"Proof of Theorem 4.9, final paragraph"}],"minor_comments":[{"comment":"The example states that a Macaulay2 computation shows the invariants of degree at most 6 in Inv^3_4 are spanned by 18 elements, but no code or reproducible script is provided. Since the example is used to motivate Conjecture 4.1, please supply the computation or a precise description of the algorithm used.","section":"Example 4.2"},{"comment":"The displayed direct-sum decomposition of Inv^d is garbled as typeset: both summands appear to lie inside I(PL^d_{d+2}), so the displayed equality cannot hold. Please replace it with a correct statement, for instance by explicitly separating the part inside the kernel I(PL^d_{d+2}) from a complementary summand.","section":"Section 4, before Theorem 4.9"},{"comment":"The sentence 'Inv4_{\\ge d+3} \\cap I(PL^4_6)' uses a parameter d that is already fixed as 4; it should read 'Inv^4_{\\ge 7} \\cap I(PL^4_6)'.","section":"Outlook, last paragraph of the even-d subsection"}],"recommendation":"major_revision","confidential_remarks":"The manuscript's even-dimensional main theorem is explicitly conditional on [19], an arXiv preprint co-authored by one of the present authors. I recommend that the editor require the authors to state the borrowed results verbatim and either prove them in an appendix or confirm the status of [19]. The paper otherwise fits the special volume on positive geometry and, after the antipode-sign correction and a rigorous transcendence-degree argument, would make a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: this paper puts a new object on the table—the ring Inv^d of volume invariants of cyclic polytopes, built from iterated-integral signatures—and shows it is infinitely generated in every dimension. That is a real contribution, and the paper is mostly careful. But as printed, Definition 4.3 contains a sign error that makes the paper internally inconsistent, and it needs to be fixed before I can trust the main theorem.\n\nThe good parts first. The stabilizer computation in Section 3 is solid and well-organized: Propositions 3.8 and 3.10 give a full description of C^d_n, and Proposition 3.12 cleanly characterizes when the signed volume is invariant under a permutation. The structural identifications in Section 4—Inv^d_{≥d+3} = TimeRevInv^d or LoopClosureInv^d ∩ TimeRevInv^d depending on parity—are sensible and well-motivated. The overall strategy for Theorem 4.9, using infinite algebraic independence in the shuffle algebra and then pushing it into the quotient by I(PL^d_{d+2}), is sound in spirit.\n\nThe soft spots are real, though. The antipode definition in Definition 4.3 sets A(w) = (-1)^{d+1} w^rev. That sign depends on the ambient dimension, not on the length of w. For d=3, A(vol3) = +vol3^rev = -vol3, so vol3 is not in TimeRevInv^3. But Proposition 3.12 puts vol3 in Inv^3, hence in Inv^3_{≥6}, and Proposition 4.5 identifies Inv^3_{≥6} with TimeRevInv^3. So the written definitions contradict each other exactly at the point where the proof of Theorem 4.9 relies on TimeRevInv. This is a load-bearing defect, though almost certainly a typo: replacing (-1)^{d+1} with (-1)^{|w|}, the standard antipode sign, repairs it.\n\nSecond, the even-dimensional half of Theorem 4.9 leans on two external results from [19], a companion preprint by one of the present authors, and from [11] with overlapping authors. I could not verify those from the text here. The transcendence-degree argument in the last paragraph of the proof is also compressed: the injection R[s_{N+1},...] → R[x_1,...,x_{d+2}] is asserted, not proved, but it is plausible. Third, Example 4.2 reports Macaulay2 output without the code, which is a reproducibility gap, though minor.\n\nBottom line: this is a paper a serious referee should engage with, not desk-reject. The construction is new, the computations are mostly careful, and the defects are repairable. I'd send it back for revision, asking for the sign fix and for a fuller statement of exactly which parts of Theorem 4.9 rest on [19]. I'd cite the corrected version.","headline":"Genuinely new ring of cyclic-polytope volume invariants, but the printed antipode sign is wrong in a load-bearing way and the even-d case is conditional on a companion paper.","tokens_in":15774,"tokens_out":3118,"would_cite":true,"duration_ms":27525,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60L10","13A50","52B05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Cyclic polytopes carry infinitely many independent volume invariants in every dimension.","keywords":["iterated integrals","piecewise linear paths","shuffle algebra","permutation group action","signed volume","invariants","positive matrices","cyclic polytopes"],"falsifier":"For a fixed $d$, exhibit a finite generating set for $Inv^d_{\\ge d+3}\\cap I(PL^d_{d+2})$ as a shuffle algebra; Theorem 4.9 asserts that no such finite set exists, so a finite generation proof or a finite degree bound on its algebraically independent elements would refute the main claim. Alternatively, for even $d$, find a counterexample to either of the two imported statements from the companion paper, since the even case depends entirely on them.","tokens_in":14809,"feed_emoji":"📐","tokens_out":11995,"duration_ms":99670,"temperature":0.7,"pith_summary":"This paper shows that cyclic polytopes have not just a volume but a whole infinite family of independent geometric attributes, all computed from iterated integrals along paths that run through the polytope's vertices. The volume of a cyclic $d$-polytope equals the signed iterated integral of a piecewise linear path through its vertices, and different paths are related by a subgroup of combinatorial automorphisms. The paper characterizes the linear combinations of iterated integrals invariant under this subgroup and proves that in every dimension there are infinitely many algebraically independent invariants with respect to the shuffle product. Consequently the ring of volume invariants is infinitely generated, so cyclic polytopes carry a large supply of canonical features rather than a single numerical invariant.","feed_headline":"Cyclic polytopes carry infinitely many independent volume invariants","feed_subtitle":"Path-signature geometry exposes an endless supply of canonical polytope features beyond plain volume.","key_machinery":"The machinery is the signature of a piecewise linear path, a linear form $S(X)$ on words in $d$ letters whose value on a word $i_1\\cdots i_k$ is the iterated integral of the corresponding coordinate differentials over the simplex $0\\le t_1\\le\\cdots\\le t_k\\le 1$. The concatenation identity for signatures makes the signature multiplicative under path concatenation, and the shuffle identity makes the map $H^d_n$ from the shuffle algebra to polynomials an algebra homomorphism. The signed volume is the value on the word $vol_d=\\sum_{\\sigma\\in S_d} sgn(\\sigma)\\sigma(1)\\cdots\\sigma(d)$. The invariance subgroup $C^d_n$ is the stabilizer of positive matrices, hence the group of combinatorial automorphisms of the cyclic polytope that preserve the sign of all maximal minors. The proof describes these groups with the evenness criterion for cyclic polytopes, identifies the $n\\ge d+3$ invariants as time-reversal and loop-closure invariants via the antipode and cyclic rotations, and uses the weak Chen–Chow theorem to convert equalities of signature polynomials into equalities of words.","core_discovery":"The central claim is Theorem 4.9: for every dimension $d$, the intersection $Inv^d_{\\ge d+3} \\cap I(PL^d_{d+2})$, and in particular the full ring of volume invariants $Inv^d$, contains infinitely many algebraically independent elements with respect to the shuffle product. Concretely, no finite list of iterated-integral expressions generates all invariant polynomial functions on cyclic $d$-polytopes. For odd $d$ the paper gives the large-$n$ invariant ring explicitly: it is either the whole word algebra when $(d+1)/2$ is odd or the subalgebra of time-reversal invariants when $(d+1)/2$ is even. For even $d$ it identifies the invariant ring with loop-closure invariants, possibly intersecting time-reversal invariants, and imports from a companion paper the fact that this loop-closure ring contains infinitely many algebraically independent elements. Because the independent family lies inside the vanishing ideal of paths with $d+2$ control points, it belongs to $Inv^d$ itself while being invisible on the minimal nontrivial number of vertices.","pith_inferences":["If the even-dimensional dependence on the companion paper could be removed by a direct construction of infinitely many algebraically independent loop-closure invariants, Theorem 4.9 would become self-contained for all $d$ rather than conditional on an external result.","Restricting the volume invariants to $SL_d$ orbits yields functions on the positive Grassmannian; Theorem 4.9 then suggests the positive Grassmannian carries infinitely many algebraically independent functions of this signature type, beyond the volume itself.","The induced equivalence relation on cyclic polytopes is probably much finer than the three explicit moves (permutations, translations, deleting collinear vertices); if Conjecture 4.1 holds, all $d$-polytopes with $d+2$ vertices of equal volume would be equivalent under the full invariant ring, a rigidity statement one could test numerically for small $d$."],"forward_implications":["In every dimension $d$, the shuffle subalgebra $Inv^d$ is not finitely generated, so any finite collection of iterated-integral signatures is insufficient to capture all invariant features of cyclic polytopes.","For odd $d$ with $(d+1)/2$ even, the large-$n$ invariant ring is exactly the ring of time-reversal invariants; for odd $d$ with $(d+1)/2$ odd, every signature word is invariant for paths with at least $d+3$ control points.","For even $d$, invariance under all cyclic rotations of control points for all $n$ is equivalent to loop-closure invariance, which ties cyclic-polytope invariants to signatures of loops and to the equivalence relations studied in the companion paper.","Since the infinite independent family lies in $I(PL^d_{d+2})$, all these new invariants vanish on paths with $d+2$ control points and first become visible for polytopes with at least $d+3$ vertices."],"supporting_citations":[{"why":"Supplies the concatenation identity for signatures, which underlies the recursive description of $H^d_n$ and the Hopf-algebra structure.","marker":"[5]"},{"why":"Gives the shuffle identity, making $H^d_n$ an algebra homomorphism so algebraic independence can be studied in the shuffle algebra.","marker":"[20]"},{"why":"Provides the signed-volume formula for polytope volume and the spanning argument used in the proof of weak Chen–Chow; it motivates why $vol_d$ is a volume invariant.","marker":"[12]"},{"why":"Provides the weak Chen–Chow lemma that converts equality of all signature values into equality of words.","marker":"[11]"},{"why":"Supplies the evenness criterion identifying facets of cyclic polytopes from positive maximal minors, used to describe the groups $C^d_n$.","marker":"[14]"},{"why":"Classifies the combinatorial automorphism groups of cyclic polytopes, which the paper refines to obtain $C^d_n$ in Propositions 3.8 and 3.10.","marker":"[15]"},{"why":"Imports the identification of even-dimensional invariant rings with loop-closure invariants and the infinite algebraic independence of the loop-closure ring; the even case rests on these.","marker":"[19]"},{"why":"Establishes that Lyndon words freely generate the shuffle algebra, used to prove infinite algebraic independence in the full algebra and for time-reversal invariants.","marker":"[21]"}],"fun_headline_variants":["Infinitely many independent invariants for cyclic polytopes","Cyclic polytopes have an infinite independent invariant ring","Path-signature approach yields infinite polytope invariants","Theorem: cyclic polytopes admit infinite independent invariants","Infinite canonical invariants for cyclic polytopes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"For even dimensions, the proof imports two facts from the companion paper: that the large-$n$ invariant ring consists exactly of signature quantities unchanged by closing a path into a loop (intersected with time-reversal invariants in one parity case), and that this loop-closure ring contains infinitely many algebraically independent elements. If either imported fact is false, the even-dimensional half of the main theorem is not derived in this paper.","fun_headline_variants_meta":{"raw":{"variants":["Infinitely many independent invariants for cyclic polytopes","Cyclic polytopes have an infinite independent invariant ring","Path-signature approach yields infinite polytope invariants","Theorem: cyclic polytopes admit infinite independent invariants","Infinite canonical invariants for cyclic polytopes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000511,"raw_usage":{"total_tokens":2443,"prompt_tokens":859,"completion_tokens":1584,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":475,"completion_tokens_details":{"reasoning_tokens":1504}},"tokens_in":475,"tokens_out":1584,"duration_ms":12042,"temperature":1.0,"reasoning_tokens":1504,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T15:06:52.741538+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a fixed $d$, exhibit a finite generating set for $Inv^d_{\\ge d+3}\\cap I(PL^d_{d+2})$ as a shuffle algebra; Theorem 4.9 asserts that no such finite set exists, so a finite generation proof or a finite degree bound on its algebraically independent elements would refute the main claim. Alternatively, for even $d$, find a counterexample to either of the two imported statements from the companion paper, since the even case depends entirely on them.","supporting_citations":[{"cited_title":"Iterated Integrals and Exponential Homomorphisms","cited_arxiv_id":null,"evidence_quote":"Supplies the concatenation identity for signatures, which underlies the recursive description of $H^d_n$ and the Hopf-algebra structure."},{"cited_title":"Lie Elements and an Algebra Associated With Shuffles","cited_arxiv_id":null,"evidence_quote":"Gives the shuffle identity, making $H^d_n$ an algebra homomorphism so algebraic independence can be studied in the shuffle algebra."},{"cited_title":"Invariants of multidimensional time se- ries based on their iterated-integral signature","cited_arxiv_id":null,"evidence_quote":"Provides the signed-volume formula for polytope volume and the spanning argument used in the proof of weak Chen–Chow; it motivates why $vol_d$ is a volume invariant."},{"cited_title":"Areas of areas generate the shuffle algebra","cited_arxiv_id":"2002.02338","evidence_quote":"Provides the weak Chen–Chow lemma that converts equality of all signature values into equality of words."},{"cited_title":"Neighborly and cyclic polytopes","cited_arxiv_id":null,"evidence_quote":"Supplies the evenness criterion identifying facets of cyclic polytopes from positive maximal minors, used to describe the groups $C^d_n$."},{"cited_title":"Automorphism groups of cyclic polytopes,","cited_arxiv_id":null,"evidence_quote":"Classifies the combinatorial automorphism groups of cyclic polytopes, which the paper refines to obtain $C^d_n$ in Propositions 3.8 and 3.10."},{"cited_title":"Conjugation, loop and closure invariants of the iterated-integrals signature","cited_arxiv_id":"2412.19670","evidence_quote":"Imports the identification of even-dimensional invariant rings with loop-closure invariants and the infinite algebraic independence of the loop-closure ring; the even case rests on these."},{"cited_title":"Free Lie Algebras, volume 7 of London Mathematical Society Monographs, New Series","cited_arxiv_id":null,"evidence_quote":"Establishes that Lyndon words freely generate the shuffle algebra, used to prove infinite algebraic independence in the full algebra and for time-reversal invariants."}],"review_version":1}