{"id":"18bfe183-81d0-496d-95f1-f6cccae3ddcc","arxiv_id":"2607.28857","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Pairs of multiplicatively dependent sums in almost disjoint multiplicative groups are, up to finitely many exceptions, root-of-unity twists; triples are classified assuming the abc conjecture.","lead":"This paper proves when sums of two numbers from two almost disjoint multiplicative groups of algebraic numbers can be multiplicatively dependent. It shows that, excluding finitely many exceptional cases, such dependent sums are just root-of-unity twists of each other, and it gives a conditional classification for three sums.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Gap in §7.4.2: non-primitive sub-sum analysis of the six-term unit equation (7.32) omits sub-sums involving the last two entries, leaving a possible hole in the proof of Theorem 7.3 and hence Theorem 1.5.","rationale":"The reader's weakest assumption centered on the necessary almost-disjointness hypothesis and the non-effectiveness/abc-dependence. Those are honest limitations, not correctness gaps. The most load-bearing threat to the paper's central claims is internal: the proof of the key technical Theorem 7.3 appears to omit a class of configurations in the six-term unit-equation argument. The assertion that any non-primitive zero sub-sum must lie among the first four coordinates is not self-evident and is false for at least some algebraic specializations (e.g., γ=-1). Because the m=3 classification (Theorem 1.5) rests directly on Theorem 7.3, this gap, if real, would invalidate the paper's main conditional result as written. It may be repairable by a more exhaustive sub-sum analysis using Corollary 3.7, so a conditional verdict remains appropriate pending such a repair. The reader's own flags about §7.4.2 typos point to the same section, but their stated weakest assumption is different; hence partial/disagree. The concrete test above would settle whether the omitted configurations are harmless or fatal.","tokens_in":29871,"tokens_out":20428,"duration_ms":200881,"concrete_test":"Enumerate all 62 proper nonempty subsets of the six entries in (7.32): {xy, −x, −y, 1, −γy^{|r|}, γ}. For each subset that can sum to zero under x,y∈O_S^×, y≠1, and (7.16), derive the resulting relation among x,y,γ. Then verify that each relation either forces y=−x^{±1} (and then z=y^{±2}), or is an equation F(y)∈O_S^× with F∈\\bar{Q}(T) having a root in \\bar{Q}^× and not of the form A T^n, so Corollary 3.7 bounds the exceptional y. If any relation admits an infinite family of S-unit solutions satisfying x−1,y−1∉O_S^× but not the theorem's conclusion, the proof of Theorem 7.3 fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"In the proof of Theorem 7.3, case b+c=0, the paper reduces to (x−1)(y−1)=γ(y^{|r|}−1) and forms the six-term unit equation (7.32): xy−x−y+1−γy^{|r|}+γ=0. It then asserts: 'Since y^{|r|}≠1, a proper subset of the first four coordinates of (7.32) must have sum 0.' This assertion is not justified. A proper zero sub-sum may include the last two coordinates, e.g. {1,γ} if γ=−1, or {xy,−γy^{|r|}} if γ=xy^{1−|r|}. The case analysis that follows checks only two- and three-term sums among the first four coordinates (xy−x, xy−y, −x+1, −y+1, and the three-term sums); sub-sums involving −γy^{|r|} or γ are never considered. Such sub-sums could conceivably yield additional infinite families satisfying (7.16) but not the conclusion y=−x^{±1}, z=y^{±2}. Some of these configurations are likely excludable by Corollary 3.7 (rational functions with roots in \\bar{Q}^×), but the proof does not say so. Since Theorem 1.5 is deduced directly from Theorem 7.3, this unhandled case is load-bearing for the conditional m=3 classification.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies multiplicative dependence of non-zero sums x_i + y_i with x_i in a finitely generated multiplicative group Gamma and y_i in an almost disjoint finitely generated group Delta of algebraic numbers. Theorem 1.2 gives an unconditional classification for m=2: apart from finitely many exceptions, x_1/x_2 = y_1/y_2 is a root of unity. The proof reduces to perfect powers in sumsets (Theorem 4.1) and to S-unit equations, with the unavoidable non-effectiveness coming from the subspace theorem. For m=3, Theorem 1.5 gives, conditionally on the weak abc-conjecture, a classification up to finitely many exceptions: the only infinite families are the two described in (1.1) and (1.2), up to permutation and root-of-unity twists. The technical core is Theorem 7.3, a conditional statement about three elements x,y,z of a finitely generated group Lambda for which x-1, y-1, z-1 are minimally multiplicatively dependent modulo Lambda. The final section derives Theorem 1.5 from Theorem 7.3. A general conjecture, Conjecture 1.6, is also proposed.","tokens_in":30210,"tokens_out":17698,"duration_ms":182477,"significance":"If correct, Theorem 1.2 is a clean and natural contribution: it identifies the only infinite family of multiplicatively dependent pairs in the sumset of almost disjoint multiplicative groups, and it is unconditional apart from the inherent non-effectiveness. The conditional m=3 classification is strong and well-motivated, and the paper is honest about its dependence on the weak abc-conjecture. The structure is good, and the paper gives explicit examples showing that the hypotheses are sharp. However, as demonstrated below, the proof of Theorem 7.3 contains a genuine gap in the analysis of a six-term unit equation, and since Theorem 1.5 is deduced from Theorem 7.3, this gap is load-bearing for the conditional classification. The unconditional Theorem 1.2 appears sound.","major_comments":[{"comment":"The assertion that, because y^{|r|} != 1, a proper zero sub-sum must involve only the first four coordinates of (7.32) is unjustified and in fact false. A zero sub-sum may include the fifth or sixth coordinate. For example, with x=2, y=3, gamma=1, r=1, the sub-sum xy - y - gamma y^r = 6-3-3 = 0 vanishes, while no proper subset of {xy, -x, -y, 1} sums to zero. The case analysis that follows checks only two- and three-term sums among the first four coordinates; all zero sub-sums involving -gamma y^{|r|} or gamma are omitted. Some of these omitted cases may be excluded by Corollary 3.7 or Theorem 3.6, but the manuscript does not perform that exclusion. Since Theorem 1.5 is derived from Theorem 7.3, this gap must be repaired.","section":"§7.4.2, equation (7.32), case b+c=0"},{"comment":"In the same paragraph, the list of three-term sums contains a sign/justification error. With coordinates (xy, -x, -y, 1), the relevant three-term sums include xy-x-y, xy-x+1, xy-y+1, and 1-x-y; the paper's '-x-y-1' is not the correct expression. More importantly, the vanishing of 1-x-y (i.e., x+y=1) does not imply that x-1 and y-1 are S-units; for instance, x=2, y=-1. The finiteness of S-unit solutions to x+y=1 follows from Theorem 3.6, but the printed argument uses a false implication. This is local and repairable, but it must be corrected.","section":"§7.4.2, three-term sums after (7.32)"}],"minor_comments":[{"comment":"The sentence 'Since groups Gamma and Delta are almost disjoint, gamma and delta must be algebraically independent' should read 'multiplicatively independent'. Two algebraic numbers cannot be algebraically independent over Q in the usual sense. The intended argument works once 'multiplicatively' is substituted.","section":"§5.3.6"},{"comment":"In (7.31), gamma should be an S-unit, not merely an element of O_S; the derivation from (7.29) gives gamma in O_S^times. The subsequent use of (7.32) as a six-term S-unit equation requires this.","section":"§7.4.2, equation (7.31)"},{"comment":"The final line of the paper says 'Theorem 7.3 is proved'; this should be 'Theorem 1.5 is proved'.","section":"§8.2"},{"comment":"The conductor terms appear to be mis-transcribed: the standard weak abc inequality should involve cond(x) + cond(x^{-1}) + cond(x-1). The displayed text repeats cond(x-1) twice. If the literal statement is intended, it is weaker than the form used in Proposition 7.20, where the full sum is bounded via Proposition 7.19. Please correct the typo or clarify.","section":"§3.4, Conjectures 3.8 and 3.9"}],"recommendation":"major_revision","confidential_remarks":"The unconditional Theorem 1.2 is a solid contribution and the paper is well worth publishing once the conditional part is fixed. The gap in §7.4.2 is significant because it affects Theorem 7.3 and therefore Theorem 1.5. I would be willing to review a revision that supplies a complete analysis of all zero sub-sums of (7.32)."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: the unconditional m=2 result (Theorem 1.2) is good, and the m=3 classification under weak abc is the kind of structural theorem people in this area will want to use. But I don't think Theorem 7.3 is proved as written.\n\nWhat's genuinely new: Theorem 1.2 gives a clean finite-exception description — multiplicative dependence of two sums forces x1/x2 = y1/y2 to be a root of unity. The proof is careful: the reduction to perfect powers (Theorem 4.1), the Evertse–Schlickewei–van der Poorten input, and the unit equation argument all hang together. The non-effectiveness is honestly disclosed, and Examples 1.3 and 1.4 show the hypotheses are doing real work. The m=3 classification is a substantial step beyond the m=2 case, and Theorem 7.3 is a reusable structural result. The conjectural framework in Section 1.2 is natural and the paper is up-front about what depends on abc.\n\nSoft spot, and it's not minor: in §7.4.2, after reducing to (7.31) and the six-term unit equation (7.32), the paper says that since y^{|r|} ≠ 1, a proper zero sub-sum must come from the first four coordinates. That does not follow. Sub-sums involving −γy^{|r|} or γ can vanish — e.g. 1+γ = 0 if γ = −1, or xy−γy^{|r|} = 0 if γ = xy^{1−|r|} — and these cases are not analyzed. The argument only checks sub-sums among the first four coordinates. Since Theorem 1.5 is deduced directly from Theorem 7.3, this gap is load-bearing for the conditional m=3 classification. It might be patchable: several of the unhandled configurations should be excludable by Corollary 3.7, treating the equations as rational functions with roots in \\bar{Q}^×, but the paper does not do that. The manuscript needs a proofread anyway — there are typos in the abc variants (§3.4) and the author's own preface warns that the text was not checked.\n\nFor the record: none of this affects Theorem 1.2, which I believe is sound. The paper is for people working on unit equations, perfect powers, and multiplicative dependence; it deserves a serious referee, not a desk reject. My recommendation: send to a competent referee with a request to examine §7.4.2 closely, and expect a revision.","headline":"The m=2 theorem is solid and citable; the abc-conditional m=3 classification is a real advance, but the proof of Theorem 7.3 has a gap in the b+c=0 sub-case that needs fixing.","tokens_in":30718,"tokens_out":4489,"would_cite":true,"duration_ms":44798,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["11D61","11J86","11R27"],"pacs":[],"model":"deepseek-v4-flash","headline":"The only infinite family of multiplicatively dependent sums from two almost disjoint multiplicative groups comes from multiplying both summands of one sum by a common root of unity.","keywords":["multiplicative dependence","sumset","finitely generated multiplicative groups","algebraic numbers","root of unity","abc conjecture","exponential Diophantine equations","heights"],"falsifier":"Take two almost disjoint rank-1 groups, for instance Γ=⟨2⟩ and Δ=⟨3⟩, and determine whether the equation 2^a+3^b = z^n has infinitely many solutions with n≥2; the paper's Theorem 4.1 predicts only finitely many, so one infinite parametric family would refute it. For Theorem 1.2 itself, one would need an infinite sequence of dependent pairs (x1,y1),(x2,y2) from almost disjoint finitely generated groups with x1/x2 not a root of unity; a single example cannot refute the theorem because it allows finitely many exceptions.","tokens_in":29727,"feed_emoji":"🔁","tokens_out":7619,"duration_ms":72928,"temperature":0.7,"pith_summary":"This paper studies when two sums from the sumset of two finitely generated multiplicative groups of algebraic numbers can be multiplicatively dependent. The main theorem states that, if the two groups intersect only in a finite group, then with finitely many exceptions the only way this happens is that the two sums are related by multiplying both summands by the same root of unity: x1/x2 = y1/y2 is a root of unity. For three sums, the paper shows — assuming the weak abc conjecture — that only two infinite families occur, up to permutation and root-of-unity twists: the difference-of-squares family and a reciprocal family. These results matter because they pin down when multiplicative dependence can occur in the sumset, a question that arises naturally in exponential Diophantine equations. The paper also proposes a conjectural description of m sums for arbitrary m.","feed_headline":"Root-of-unity twists exhaust dependent-sum pairs","feed_subtitle":"Except for finitely many cases, dependent sums force identical ratios up to roots of unity.","key_machinery":"Almost disjointness yields a height comparison: for x∈Γ and y∈Δ, h(xy) ≫ h(x)+h(y). This prevents height cancellation in products and forces a multiplicative dependence between two sums to mean either a perfect power in Γ+Δ or equality of the sums up to a root of unity. For three sums, the proof reduces to a structural theorem classifying triples (x,y,z) in one group with x−1, y−1, z−1 multiplicatively dependent modulo the group; that theorem uses linear-forms-in-logarithms bounds, a logarithmic-gcd estimate, the primitive-divisor theorem, and the weak abc conjecture.","core_discovery":"The central claim is Theorem 1.2: if Γ and Δ are almost disjoint finitely generated multiplicative groups of algebraic numbers, and x1+y1 and x2+y2 are non-zero and multiplicatively dependent with x1,x2∈Γ, y1,y2∈Δ, then, with finitely many exceptions, x1/x2 = y1/y2 is a root of unity. The proof splits by the exponent of the dependence: if the common power has exponent at least 2, the sum is a perfect power and the finiteness of perfect powers in Γ+Δ bounds heights; if the exponent is 1, the sums are either equal up to a root of unity or their product is a root of unity, each handled by classical unit-equation finiteness. The theorem is non-effective in general, but effective when both groups","pith_inferences":["The non-effectiveness of the general m=2 statement seems tied to the non-effective finiteness theorem used to rule out unequal ratios; any effective generalization would likely need a new height-matching argument or a restriction to low rank, as the rank-1 case illustrates.","If the conjectural description for arbitrary m (Conjecture 1.6) is correct, then all minimally dependent m-tuples are built by concatenating canonical pairs (from differences of powers) and twisted pairs (from reciprocal/inverse pairs), yielding a complete structural description of multiplicative dependence in sumsets of almost disjoint groups.","One could test the conditional three-sum classification in the rank-1 setting, where the proof's non-effective ingredients are replaced by Baker-type explicit bounds, giving a fully effective version of Theorem 1.5 for rank-1 groups.","The weak abc assumption enters only in bounding conductors; for a specific number field K for which Conjecture 3.9 is verified, Theorem 1.5 would become unconditional for that field."],"forward_implications":["If Γ and Δ are almost disjoint, any two multiplicatively dependent non-zero sums from Γ+Δ must, with finitely many exceptions, have x1/x2 = y1/y2 a root of unity.","The sumset Γ+Δ contains only finitely many perfect powers in any given number field, with an effective height bound.","For three sums, assuming weak abc, any minimally dependent triple is, up to permutation and torsion, either of the form (x−y, x+y, x²−y²) or of the reciprocal form (x1+y1, x2+y2, x2⁻¹+y2⁻¹) with x1+y1 multiplicatively dependent on x2y2.","The rank-1 case of Theorem 1.2 is effective, giving explicit height bounds for the exceptional pairs."],"fun_headline_variants":["Dependent sums force matching ratios up to roots of unity","Except finitely many, dependent sums have ratio roots of unity","Multiplicative dependence in sums pins ratios to roots of unity","For almost all dependent sums, ratios agree up to roots of unity","Sums from two groups: dependence implies ratio root of unity"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The main theorem collapses if Γ and Δ meet in an infinite group: Example 1.3 shows that Γ=⟨−1,3⟩ and Δ=⟨2,3⟩ give every power of 3 in Γ+Δ, so multiplicative dependence is abundant. The three-sum classification additionally rests on the unproved weak abc conjecture for the number field K.","fun_headline_variants_meta":{"raw":{"variants":["Dependent sums force matching ratios up to roots of unity","Except finitely many, dependent sums have ratio roots of unity","Multiplicative dependence in sums pins ratios to roots of unity","For almost all dependent sums, ratios agree up to roots of unity","Sums from two groups: dependence implies ratio root of unity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000645,"raw_usage":{"total_tokens":2793,"prompt_tokens":729,"completion_tokens":2064,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":473,"completion_tokens_details":{"reasoning_tokens":1988}},"tokens_in":473,"tokens_out":2064,"duration_ms":15707,"temperature":1.0,"reasoning_tokens":1988,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-03T01:21:49.162379+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take two almost disjoint rank-1 groups, for instance Γ=⟨2⟩ and Δ=⟨3⟩, and determine whether the equation 2^a+3^b = z^n has infinitely many solutions with n≥2; the paper's Theorem 4.1 predicts only finitely many, so one infinite parametric family would refute it. For Theorem 1.2 itself, one would need an infinite sequence of dependent pairs (x1,y1),(x2,y2) from almost disjoint finitely generated groups with x1/x2 not a root of unity; a single example cannot refute the theorem because it allows finitely many exceptions.","supporting_citations":[],"review_version":1}