{"id":"c07388d9-69e2-4235-a678-bc72151cba1e","arxiv_id":"2412.15767","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves modularity of new rank-three Nahm sums from a lift-dual construction and gives two new rank-three counterexamples to Zagier's duality conjecture.","lead":"This paper constructs new modular infinite sums, known as Nahm sums, by lifting known rank-two examples to rank three and applying a duality operation. It also finds two new cases where a conjecture about these sums fails.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1 depends entirely on the external Vlasenko–Zwegers identities (5.1)–(5.2), which are not reproved here; a single coefficient or index error there would invalidate the modular product formulas (1.16)–(1.17).","rationale":"The reader’s verdict was CONDITIONAL, and the reader’s weakest assumption is exactly the external dependence on (5.1)–(5.2). My stress-test agrees: this is the single most load-bearing point because Theorem 1.1 is the only claim whose proof relies on material absent from the paper. The other main theorems (3.8)–(3.10) and (4.4)–(4.6) are proved in full using the stated q-series machinery, and the counterexample section is a byproduct rather than the core assertion. The external identities are published in a peer-reviewed paper, so reliance on them is not itself an error, but the lack of any re-proof or numerical sanity check leaves the central new identity less self-contained than the rest of the paper. The concrete coefficient-check I propose would settle whether the dependence actually matters. I found no internal inconsistency in the constant-term or Bailey-pair arguments, and the substitution q → q^{4/3} is a valid formal operation on these series. Hence I do not propose changing the reader’s verdict: the paper is likely correct, with this caveat.","tokens_in":32542,"tokens_out":43204,"duration_ms":329502,"concrete_test":"Independently verify (5.1) and (5.2) by direct q-series expansion to, say, 60 terms: compute the left sides by summing over 0 ≤ i,j ≤ 40 and expand the J-products on the right to the same order; any mismatch invalidates Lemma 5.1. Then, assuming those pass, substitute q → q^{4/3}, perform the dissection in (5.13)–(5.16) automatically, and check that the resulting expression equals the right-hand sides of (1.16)–(1.17) through q^{50}. If both checks pass, the dependency on the external result is harmless; if either fails, Theorem 1.1 must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.1 is the only route to modularity for the dual of the lift of Zagier’s Example 11. Its key input is Lemma 5.1, which is derived solely from the two identities (5.1)–(5.2), attributed to Vlasenko–Zwegers and proved in the authors’ earlier paper [5]. The present manuscript neither reproves these identities nor provides any independent check. The subsequent step replaces q by q^{4/3} in those identities and then dissects the resulting q-series by residues modulo 3 (equations (5.13)–(5.16)). This substitution is formally legitimate, but it amplifies the risk: any typographical or computational error in the J-indices or in the constant coefficients of (5.1)–(5.2) would propagate directly into the claimed eta-quotient expressions (1.16)–(1.17). In particular, the factors 2 and the specific indices 21, 6, 9, 12, 18, 3 in the J_{a,45} products are exactly what determines the modular transformation behavior; a wrong coefficient could turn a modular form into a sum of forms of different weights, breaking the central conclusion. The manuscript itself flags the dependence as 'totally unexpected', and the external identities are substantial enough that the reader cannot verify them from the text alone. Thus, if (5.1) or (5.2) is even slightly wrong, the main new modularity theorem for the Example-11 dual sums fails.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a 'lift-dual' operation for Nahm sums: starting from a rank-two Nahm sum with matrix A and vector B, it lifts to a rank-three sum via the matrix and vector in (1.7), then takes the dual via the standard duality operator D in (1.6). Applying this to Zagier's rank-two Examples 1, 9, and 11, the authors state explicit Rogers–Ramanujan type identities (Theorems 3.1, 4.1, 1.1) that express the resulting rank-three Nahm sums as products of generalized Dedekind eta functions, thereby proving modularity. Section 6 contains a specialization a=3/2 with additional modular triples, and two new counterexamples to Zagier's duality conjecture are claimed. The proofs use constant-term extraction, contour integrals, and Bailey-pair techniques.","tokens_in":32827,"tokens_out":25930,"duration_ms":197533,"significance":"If the results are correct, the paper provides several new explicit modular rank-three Nahm sums, a useful contribution to the ongoing classification in Nahm's problem. The proofs are detailed and rely on standard q-series methods, and the identities are explicit enough to be checked numerically. The claimed counterexamples to Zagier's duality conjecture are potentially significant for the broader understanding of that conjecture. However, the paper's motivating identity (1.8), which underlies the lift-dual construction, appears to be false as stated; this does not necessarily invalidate the modularity theorems (which are proved independently), but it undermines the paper's central narrative and requires substantial revision.","major_comments":[{"comment":"The identity f_{A,B,0}(q) = f_{\\tilde A,\\tilde B,0}(q) is stated for all A,B, but it is false for the lift defined in (1.7) even in the basic case a=2 of Example 1. Taking A = [[2,-1],[-1,2]], B=(0,0), the left side has q^2-coefficient 4 (from the triples (1,0), (0,1), (1,1)), while the right side, with \\tilde A = [[2,0,1],[0,2,1],[1,1,2]], has q^2-coefficient 3 (from (1,0,0), (0,1,0), (0,0,1)). Thus the claim that every rank-two modular triple lifts to a rank-three modular triple 'for free' is not correct as stated, and the presentation of the sums in Sections 3–5 as 'lift-dual' of Zagier's examples is not justified for all parameters. The authors must correct the identity, impose the conditions under which it actually holds, or reframe the construction without claiming a lift relation.","section":"§1, Eq. (1.8)"},{"comment":"The proof of Theorem 1.1 relies entirely on the Vlasenko–Zwegers identities (5.1)–(5.2), which are neither proved nor derived in this paper. These identities are the sole input to Lemma 5.1, and a single index or coefficient error in them would propagate directly into the modular product formulas (1.16)–(1.17). Because this is the only route to the modularity claim for the Example-11 dual sums, the authors should state precisely where in [5] these identities are proved, and ideally include an independent verification (for example, a numerical check of (1.16)–(1.17) to high order) so that readers can rule out transcription errors.","section":"§5, Lemma 5.1 and Theorem 1.1"},{"comment":"The nonmodularity conclusions in Section 6.3 depend on the assertion that the expressions in (6.9) and (6.10) are sums of modular forms of weights 0 and 1, but no weight computation or modular transformation law is provided. The authors should show explicitly that the first term in each expression transforms as a weight-0 modular form and the second as a weight-1 modular form (after multiplying by the appropriate power of q), and that the weight-1 piece does not vanish identically. Without this, the counterexamples to Zagier's duality conjecture are not fully substantiated.","section":"§6.2 and §6.3, Theorem 6.2"},{"comment":"The parameters m and ν in Theorem 3.1 are rational in the intended application (for instance, m = 1/(4(a-1))), but the right-hand sides involve J_{a,m} with indices such as 4(4m+1) and 4(4m+ν+1) that are not necessarily integers. Since the modularity interpretation of J_{a,m} in (1.10) generally requires integer indices, the authors should specify the domain of m and ν and, where necessary, the level of the modular group for which the products are modular forms. This is a technical point, but it is load-bearing for the modularity claim.","section":"§3, Theorem 3.1"}],"minor_comments":[{"comment":"The word 'sated' should be 'stated' in the sentence 'Nahm’s conjecture, sated explicitly by Zagier'.","section":"§1, first paragraph"},{"comment":"The word 'counterexmaples' is a typo for 'counterexamples'.","section":"§6.3, heading"},{"comment":"The notation in (2.2) introduces a new function but does not clearly distinguish it from J_{a,m} defined in (1.10); please use a different symbol (for example, \\overline{J}_{a,m} or \\mathcal{J}_{a,m}) to avoid ambiguity.","section":"§2, Eq. (2.2)"},{"comment":"The application of the contour integral formula (2.26) does not state the conditions on the contour (poles inside/outside) or the convergence of the interchanges; a remark that all manipulations are valid as formal power series or for |q| sufficiently small would make the proof more rigorous.","section":"§4, proof of Theorem 4.1"},{"comment":"The substitution q replaced by q^{4/3} is essential but may confuse readers because it introduces fractional powers; a brief note that the identities are formal and that the substitution is performed on the q-series coefficients would be helpful.","section":"§5, Eq. (5.13)–(5.16)"}],"recommendation":"major_revision","confidential_remarks":"The most serious issue is Eq. (1.8), which appears to be a misstatement of a known result rather than a deliberate error. The authors should be asked to re-examine the source (Zwegers's unpublished work and [5]) and either prove the identity in the intended parameter range or remove the lift-dual framing from the paper. The modularity identities in Sections 3–5 are explicit and likely correct; with a corrected narrative the paper could still be a valuable contribution. The dependence on the authors' own previous paper [5] for Theorem 1.1 is a self-citation, but the identities are stated explicitly and the authors should be expected to provide a reference to the exact theorem in [5]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, honest contribution to the Nahm sums program. It produces new modular rank-three Nahm sums by applying Zagier's lift-dual idea to his rank-two Examples 1, 9, 11, and it gives two new counterexamples to Zagier's duality conjecture. The main structural caveat is that the proof of one key theorem leans on an external identity from the authors' earlier paper; that is a real soft spot but not a fatal one.\n\nWhat's genuinely new: the Rogers–Ramanujan type identities (3.8)–(3.10), (4.4)–(4.6), (1.16)–(1.17), and the modular product formulas in Section 6.3. These are explicit, checkable, and proven in detail with standard q-series tools: Bailey pairs, constant term extractions, contour integrals, and Slater-type identities. The two counterexamples to Zagier's duality conjecture are new and complement Wang's earlier rank-four examples. The paper is careful about which lifts are positive definite and which give essentially new sums.\n\nWhere I'd want more: Theorem 1.1 depends entirely on the Vlasenko–Zwegers identities (5.1)–(5.2), which are not reproved here. The stress-test concern about an index or coefficient error there is legitimate: the q^{4/3} substitution and mod-3 dissection would propagate any typo into the eta-quotient expressions (1.16)–(1.17). That said, (5.1)–(5.2) are published and peer-reviewed in the authors' earlier paper [5]; the reliance is transparent and the paper flags it. A referee should verify the indexing against [5] and do a low-order numerical check of (1.16)–(1.17). Also, Theorem 6.2 states the weights of the modular forms involved without showing the computation; that's a minor gap, easily filled.\n\nOverall, the math looks coherent, the citation pattern is honest, and the new identities are a real extension of the classification of modular triples. This is for specialists in q-series and modular forms, and for people using Nahm sums in CFT. It deserves a serious referee; I'd send it out with a request to check the external-identity dependence and the weight computation. My own verdict would be accept after minor revision.","headline":"A solid, workmanlike extension of the Nahm sums program that produces new modular rank-three sums and two counterexamples to Zagier's duality conjecture; the main caveat is a heavy reliance on one external identity from the authors' own earlier paper.","tokens_in":33416,"tokens_out":2682,"would_cite":true,"duration_ms":21857,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11P84","33D15","33D60","11F03"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper establishes that the lift-dual operation applied to three of the known rank-two Nahm sums yields new rank-three modular Nahm sums, with explicit Rogers-Ramanujan type identities, and exhibits two new rank-three counterexamples…","keywords":["Nahm sums","modular triples","Rogers-Ramanujan type identities","Bailey pairs","lift-dual operation","duality conjecture","q-series","modular forms"],"falsifier":"Expand both sides of identities (1.16), (1.17), (3.8)-(3.10), and (4.4)-(4.6) as power series in $q$ and compare coefficients through $q^{50}$; any mismatch at a single order would falsify the corresponding identity and hence the modularity claim for that Nahm sum. For the claimed counterexamples, the obstruction is visible directly from Theorem 6.2: the two summands have modular weights 0 and 1, so one can check by a modular-symbol computation that no shift $q^C$ kills the weight-1 part.","tokens_in":32313,"feed_emoji":"","tokens_out":10451,"duration_ms":83124,"temperature":0.7,"pith_summary":"Nahm sums are q-series built from a symmetric rational matrix and two rational vectors, indexed by triples of nonnegative integers; a sum is modular when a suitable q-power of it is a modular form. The paper tries to produce new modular rank-three Nahm sums by a lift-dual operation: lift a known rank-two sum to rank three with an embedding that leaves the sum unchanged, then apply the duality map $(A,B,C)\\mapsto(A^{-1},A^{-1}B,\\tfrac12 B^TA^{-1}B-\\tfrac{r}{24}-C)$. For three of the known rank-two examples the lifted matrix is positive definite, and the paper proves that the dual sums are modular by writing them as finite combinations of the q-product functions $J_m$ and $J_{a,m}$. A byproduct is two new rank-three counterexamples to the duality conjecture, cases where a Nahm sum is not modular for any $C$ while its dual is modular. If the identities are correct, the known list of modular triples grows systematically and the boundary of the duality conjecture is drawn more sharply.","feed_headline":"Lift-dual trick yields three new modular rank-three Nahm sums","feed_subtitle":"Explicit product identities prove the duals modular, and two cases break the duality conjecture.","key_machinery":"The carrying mechanism is the two-step operation: the lifting operator sends a rank-two datum $(A,B,C)$ to a rank-three datum $(\\tilde A,\\tilde B,C)$ with $\\tilde A$ and $\\tilde B$ as displayed in (1.7), preserving the value of the Nahm sum with $C=0$, and the dual operator sends $(A,B,C)$ to $(A^{-1},A^{-1}B,\\frac12 B^TA^{-1}B-\\frac r{24}-C)$. The proof machinery is a set of q-series techniques: constant-term extraction and contour integration to reduce the triple sums, Bailey pairs and their change-of-base transformations to evaluate the reduced sums, and a catalogue of single-sum Rogers-Ramanujan type identities. A crucial auxiliary input is the 3-dissection of the Example 10 identities (5.1)-(5.2), which supplies the dissection of a double sum used in the proof of Theorem 1.1.","core_discovery":"The central claim is that the dual Nahm sums obtained from the lift-dual operation applied to Zagier's rank-two Examples 1, 9, and 11 are modular. The proof is the explicit Rogers-Ramanujan type identities (3.8)-(3.10), (4.4)-(4.6), and (1.16)-(1.17), which express the sums as finite linear combinations of products of $J_m=(q^m;q^m)_\\infty$ and $J_{a,m}=(q^a,q^{m-a},q^m;q^m)_\\infty$; since these are eta-type products, modularity follows. The paper further claims that the same search produces two new rank-three counterexamples to the duality conjecture: for the matrix (6.1) with vectors $(1/2,1/2,0)^T$ and $(1,1,1)^T$, the Nahm sums are not modular for any $C$, while their dual sums with matrix (6.31) are modular (Theorem 6.3); Theorem 6.2 exposes the obstruction by writing the nonmodular sums as sums of a weight-zero and a weight-one modular form.","pith_inferences":["The lift-dual recipe is not limited to the three examples; any modular rank-two triple with positive definite lift is a candidate, and the paper's methods should extend, although the required dissections may not always exist.","The weight-0-plus-weight-1 obstruction suggests a sharper form of the duality conjecture: the dual of a modular Nahm sum is modular exactly when all summands in its natural decomposition have the same weight.","The new counterexamples lower the rank at which the duality conjecture fails from four to three, placing the failure at the lowest rank currently known."],"forward_implications":["The dual Nahm sums of the lifted Examples 1, 9, and 11 are modular, with explicit product formulas (3.8)-(3.10), (4.4)-(4.6), and (1.16)-(1.17); modularity follows because each side is a finite combination of the functions $J_m$ and $J_{a,m}$.","Because the lifting identity preserves the sum, every modular rank-two triple whose lifted matrix is positive definite yields a modular rank-three triple; for the three treated examples the duals give genuinely new rank-three modular triples.","Two pairs (matrix, vector) found in the search are counterexamples to the duality conjecture: the lifted Nahm sums for $B=(1/2,1/2,0)^T$ and $B=(1,1,1)^T$ are not modular for any $C$, while their duals are modular.","Theorem 6.2 shows the failure is structural: the nonmodular sums split into a modular form of weight 0 plus one of weight 1, so no choice of q-power can make them modular."],"supporting_citations":[{"why":"Supplies the rank-two modular triples (Examples 1, 9, 11) and the duality observation that motivates the lift-dual operation.","marker":"[20]"},{"why":"Provides the identities for Example 1 used in Section 3 and conjectures the Example 10 identities (5.1)-(5.2) on which Theorem 1.1 depends.","marker":"[15]"},{"why":"Proves the Example 10 identities (5.1)-(5.2) and gives a proof of the lifting identity (1.8).","marker":"[5]"},{"why":"First source of the lifting identity $f_{A,B,0}=f_{\\tilde A,\\tilde B,0}$ that makes the lift-dual construction possible.","marker":"[21]"},{"why":"Contains a proof of the lifting identity (1.8).","marker":"[11]"},{"why":"Supplies the contour-integral and constant-term method for rank-three sums used in (4.8) and (5.12).","marker":"[18]"},{"why":"Provides the earlier rank-four counterexamples to the duality conjecture and the identity (6.29) used in Theorem 6.2.","marker":"[19]"},{"why":"Slater's list supplies the single-sum identities (2.10)-(2.16) used to evaluate the reduced sums in Section 4.","marker":"[14]"},{"why":"Supplies several modular-product evaluations (Entries 3.5.4-3.5.6, 4.2.8-5.3.1, and (2.23)) used throughout the proofs.","marker":"[2]"},{"why":"Provides identities (6.52)-(6.53) needed in the proof of Theorem 6.3.","marker":"[16]"}],"fun_headline_variants":["Three new modular Nahm sums from a lift-dual trick","Two new rank-three counterexamples to Nahm duality","Explicit identities prove three lift-dual Nahm sums modular","Lift Zagier's rank-two sums to get modular rank-three duals","Lift-dual yields three modular sums and two counterexamples"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of Theorem 1.1 assumes the two identities (5.1) and (5.2) for Zagier's Example 10, which are taken from earlier work and not reproved here; if either identity is wrong, the product formulas (1.16)-(1.17) and the modularity of those two dual sums collapse.","fun_headline_variants_meta":{"raw":{"variants":["Three new modular Nahm sums from a lift-dual trick","Two new rank-three counterexamples to Nahm duality","Explicit identities prove three lift-dual Nahm sums modular","Lift Zagier's rank-two sums to get modular rank-three duals","Lift-dual yields three modular sums and two counterexamples"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001213,"raw_usage":{"total_tokens":4968,"prompt_tokens":892,"completion_tokens":4076,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":508,"completion_tokens_details":{"reasoning_tokens":3989}},"tokens_in":508,"tokens_out":4076,"duration_ms":28050,"temperature":1.0,"reasoning_tokens":3989,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T11:07:21.256118+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Expand both sides of identities (1.16), (1.17), (3.8)-(3.10), and (4.4)-(4.6) as power series in $q$ and compare coefficients through $q^{50}$; any mismatch at a single order would falsify the corresponding identity and hence the modularity claim for that Nahm sum. For the claimed counterexamples, the obstruction is visible directly from Theorem 6.2: the two summands have modular weights 0 and 1, so one can check by a modular-symbol computation that no shift $q^C$ kills the weight-1 part.","supporting_citations":[{"cited_title":"Zagier, The dilogarithm function, in Frontiers in Number Theor y, Physics and Geometry, II, Springer, 2007, 3–65","cited_arxiv_id":null,"evidence_quote":"Supplies the rank-two modular triples (Examples 1, 9, 11) and the duality observation that motivates the lift-dual operation."},{"cited_title":"Vlasenko and S","cited_arxiv_id":null,"evidence_quote":"Provides the identities for Example 1 used in Section 3 and conjectures the Example 10 identities (5.1)-(5.2) on which Theorem 1.1 depends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the Example 10 identities (5.1)-(5.2) and gives a proof of the lifting identity (1.8)."},{"cited_title":"Zwegers, presentation","cited_arxiv_id":null,"evidence_quote":"First source of the lifting identity $f_{A,B,0}=f_{\\tilde A,\\tilde B,0}$ that makes the lift-dual construction possible."},{"cited_title":"Lee, Algebraic structures in modular q-hypergeometric series, PhD Thesis, University of California, Berkeley, 2012","cited_arxiv_id":null,"evidence_quote":"Contains a proof of the lifting identity (1.8)."},{"cited_title":"Wang, Explicit forms and proofs of Zagier’s rank three examp les for Nahm’s problem, Adv","cited_arxiv_id":null,"evidence_quote":"Supplies the contour-integral and constant-term method for rank-three sums used in (4.8) and (5.12)."},{"cited_title":"Counterexamples to Zagier's Duality Conjecture on Nahm Sums","cited_arxiv_id":"2411.09701","evidence_quote":"Provides the earlier rank-four counterexamples to the duality conjecture and the identity (6.29) used in Theorem 6.2."},{"cited_title":"Slater, Further identities of the Rogers–Ramanujan type , Proc","cited_arxiv_id":null,"evidence_quote":"Slater's list supplies the single-sum identities (2.10)-(2.16) used to evaluate the reduced sums in Section 4."},{"cited_title":"Andrews and B.C","cited_arxiv_id":null,"evidence_quote":"Supplies several modular-product evaluations (Entries 3.5.4-3.5.6, 4.2.8-5.3.1, and (2.23)) used throughout the proofs."},{"cited_title":"Xia and O.X.M","cited_arxiv_id":null,"evidence_quote":"Provides identities (6.52)-(6.53) needed in the proof of Theorem 6.3."}],"review_version":1}