{"id":"60fb4abd-5359-4c4f-8050-d932854ca622","arxiv_id":"2501.11173","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In AG(7,2), there are two affine equivalence classes of 10-caps, one class of 11-caps, one class of 12-caps, and no cap with 13 points.","lead":"This paper classifies the largest quad-free card sets in the 128-card game Quads. It shows there are two kinds of 10-card sets, one kind each of 11- and 12-card sets, and that 13 cards always force a quad.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.3's proof omits union sizes 5–6, leaving a load-bearing gap; a short case analysis patches it. Also, Corollary 1.3(2) is false as stated for low-dimensional caps.","rationale":"The reader's CONDITIONAL verdict is appropriate. The main classification (two 10-cap classes, one 11-cap class, one 12-cap class, maximum cap size 12) is supported by explicit dependent-set templates and affine-basis-exchange arguments, and the internal logic is coherent apart from Lemma 4.3 and several labeling typos (e.g. Theorem 7.1 says '11-cap' where a 12-cap is meant, and the ordering of dependent elements in Theorem 7.4 is inconsistent with the stated extended types). The Lemma 4.3 gap is exactly the reader's weakest assumption: the proof invokes Proposition 3.3 only for union size 7, leaving sizes 5–6 unhandled. This gap is patchable: Lemma 4.2 rules out union size ≤6, and a short complement-set argument rules out union size 7 by producing a quad. The additional issue is that Corollary 1.3(2) is false under the paper's own Definition 3.1, since a 3-cap in a 2-dimensional flat is complete in its affine span; the correct statement is Corollary 8.3 restricted to dimension 7. This does not affect the maximum-cap-size result or the '13 cards guarantee a quad' application. The paper should be accepted only after the lemma is fixed and the corollary is qualified, so the existing conditional verdict stands.","tokens_in":22222,"tokens_out":30591,"duration_ms":264189,"concrete_test":"Close the missing cases of Lemma 4.3 analytically: if |B1∪B2∪B3|≤6, two 5-supports in a 6-set intersect in at least 4, contradicting Lemma 4.2. If the union has size 7, all three supports are 5-sets; write Ci for the 2-element complement in the union. The cap condition |Bi∩Bj|∈{2,3} forces |Ci∩Cj|=0, so the three Ci are disjoint and the unique element outside their union lies in all three Bi; then x1+x2+x3 equals that basis element, a quad. Recompute Theorems 6.5 and 7.4 with this completed lemma and confirm the dependent-set templates still contain no four-point affine dependence.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The classification (Theorems 5.4, 6.5, 7.4, 8.2) depends on Lemma 4.3, which asserts that any three dependent supports have union size 8. The proof only treats |union|=7 via the 6-dimensional cap bound of Proposition 3.3; sizes 5 and 6 are not addressed. They are easy to rule out: if |B1∪B2∪B3|≤6, then two 5-supports in a 6-set intersect in at least 4, contradicting Lemma 4.2; if the union has size 7, all three supports must be 5-sets. Writing Ci for the 2-element complement of each Bi inside the 7-element union, the cap condition |Bi∩Bj|∈{2,3} forces |Ci∩Cj|=0, so the Ci are pairwise disjoint. The unique element outside C1∪C2∪C3 then lies in all three Bi, making x1+x2+x3 equal to a single basis element, i.e. a quad. So the lemma is true, but the written proof has a real gap that propagates into every later classification theorem. Separately, Corollary 1.3(2) ('a cap in AG(7,2) is complete iff it has size 12') is false under Definition 3.1: a 3-cap in a 2-dimensional flat, e.g. {0,e1,e2}, has QC1 equal to its affine span and is therefore complete, yet has size 3. The valid statement is Corollary 8.3, restricted to 7-dimensional caps.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies, up to affine equivalence, all caps (quad-free subsets) of size at least 10 in AG(7,2), continuing the program of [2]. The main results are: two affine equivalence classes of 10-caps, one class of 11-caps, and one class of 12-caps; maximum cap size 12; no 10- or 11-cap is complete; and hence 13 cards guarantee a quad in Quad-128. The method represents a cap as an affine basis plus a dependent set, assigns 'extended types' to bases, constructs explicit dependent-set templates, and uses a basis exchange theorem and inclusion-exclusion to rule out impossible intersection patterns. The proof relies on the published classification of lower-dimensional caps in [2].","tokens_in":22420,"tokens_out":31491,"duration_ms":256703,"significance":"If the classification is correct, it settles the maximum cap size in AG(7,2) and provides the first complete classification above dimension 6 in this setting, with concrete templates for each equivalence class. The explicit dependent-set templates in Tables 1-3 and the inclusion-exclusion contradiction for 13-caps are transparent and checkable; the lower-dimensional input from [2] is independent, so there is no circularity. The main weaknesses are the incomplete proof of Lemma 4.3, an overstatement in Corollary 1.3(2), and notational/typographical errors in the 12-cap arguments; these are fixable but currently block full confidence.","major_comments":[{"comment":"The proof of Lemma 4.3 only considers the case |B1∪B2∪B3|≤7 and immediately applies Proposition 3.3 to conclude |D′|≤2. Proposition 3.3, however, is stated only for 6-dimensional caps; if |B1∪B2∪B3| is 5 or 6, the subcap C′ has dimension 4 or 5 and the cited bound does not apply. Since Lemma 4.3 is used in Theorems 6.1, 6.5, 7.4, and 8.2, this is a load-bearing gap. The lemma is true and can be repaired by a short case analysis: rule out unions of size 5 and 6 using Lemma 4.2, then handle union size 7 by showing the three 5-subsets have pairwise-disjoint 2-element complements inside the 7-set. The written proof, however, is incomplete.","section":"§4, Lemma 4.3"},{"comment":"Corollary 1.3(2) states that a cap in AG(7,2) is complete if and only if it has size 12. Under Definition 3.1, this is false: the 3-cap {0,e1,e2} in a 2-dimensional flat satisfies QC1(C)=aff(C) and is complete although it has size 3. The correct statement is Corollary 8.3(2), which restricts to complete caps of dimension 7. The abstract and introduction should be amended accordingly.","section":"Corollary 1.3(2) and §8 (Corollary 8.3)"},{"comment":"The proof of the 12-cap classification is difficult to follow and contains several errors that need correction. In Theorem 7.1, 'Let C be an 11-cap' should be 'Let C be a 12-cap', and later 'the 7-dimensional 10-cap C\\{x1}' should be '11-cap C\\{x1}'. More substantively, in Theorem 7.4 the proof of Case 1 uses an inconsistent ordering of the dependent elements: the triple C123 = B∪{x1,x3,x2} has type (3,3,2) only when the unique 2-intersection pair is (x2,x3), but the theorem's lexicographic extended type places the 2 at (x1,x2). Consequently the text '|B123| = |B234| = 1' is wrong (the exceptional triples are {1,2,3} and {1,2,4}), and the later use of |B234|=1 is not justified by the stated type. The case can be repaired by a consistent permutation of the dependent elements, but as written the derivation of the template for x4 is not reliable.","section":"§7, Theorems 7.1 and 7.4"}],"minor_comments":[{"comment":"The definition of type writes |B1|-|B2|-...-|Bm|, but the list should be indexed by the dependent set D of size r, not by the basis B of size m; otherwise the notation is inconsistent with Example 2.9 and all later uses.","section":"Definition 2.8"},{"comment":"The proof refers to 'Theorem 7.3', but the relevant statement is Proposition 7.3; there is no Theorem 7.3 in the paper.","section":"Theorem 7.4, proof"},{"comment":"In the 'Other direction' paragraph, the second extended type is printed as '5-5-5-5-(2,3,3,3,3,3,2)' with an extra 3; it should have six entries: (2,3,3,3,3,2).","section":"Theorem 7.4, proof"},{"comment":"The sums x1+x2+x3 are said to involve eight and six elements in the two cases; the correct counts are five and three. The argument only needs the count to differ from four, so this is a harmless miscount.","section":"Theorem 6.5, proof"},{"comment":"The line 'x1 + x2 + x3 + x4 = a2 + a4, which involved six elements' is incorrect: the displayed sum has two elements (or one, depending on the ordering used), and 'involved' should be 'involves'.","section":"Theorem 7.4, Case 1"},{"comment":"The final sentence says 'This shows |B12| ≠ 2', but the argument concerns |B23|; the intended conclusion is |B23|=3.","section":"Theorem 6.1, proof"}],"recommendation":"major_revision","confidential_remarks":"The stress-test concern about Lemma 4.3 lands: the proof gap is real and it propagates into the classification theorems. The overstatement in Corollary 1.3(2) is also real. I do not see these as fatal—the classification appears correct and the patches are local—so I recommend major revision rather than rejection. The many small typos in the 12-cap section suggest the authors should be asked to rewrite that part with a consistent labeling of dependent elements."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version. The paper does something real: it classifies 10-, 11-, and 12-caps in AG(7,2), proves max size 12 and no 13-caps, using the basis-template machinery from [2]. The constructions are explicit and checkable, the equivalence arguments via Lemma 2.12 are sound, and the result does not rest on circular reasoning. I believe the classification is correct.\n\nWhat I'd want fixed before signing off: Lemma 4.3's proof. It says: if three 5-supports have union of size at most 7, then the subcap has dimension at most 6 and therefore at most two dependent elements by Proposition 3.3. But Proposition 3.3 only bounds 6-dimensional caps; it says nothing directly about dimensions below 6. The missing cases are easy: if the union has size at most 6, two 5-subsets intersect in at least four points, contradicting Lemma 4.2. If the union has size 7, the three 2-element complements are pairwise disjoint, leaving one element in all three supports, and x1+x2+x3 is that basis element — a quad. So the lemma is true, but the printed proof doesn't show it. Since Lemma 4.3 is used in Theorems 6.1, 6.5, 7.4, and 8.2, this is load-bearing and needs a real proof in the revision.\n\nSecond, Corollary 1.3(2) is false as stated. \"A cap in AG(7,2) is complete iff it has size 12\" is contradicted by a 3-cap in a 2-flat, like {0,e1,e2}, whose first quad closure is the whole flat. The correct statement is Corollary 8.3: every complete cap of dimension 7 has size 12. The abstract and introduction should be aligned with that. Minor typos: Theorem 7.1 says \"11-cap\" when it means \"12-cap\" at the start of the proof; Theorem 7.4 has a 7-entry extended type \"5-5-5-5-(2,3,3,3,3,3,2)\" that should be the 6-entry version. These are easy to clean up.\n\nOverall: the main classification is believable and useful for anyone working on caps, Sidon sets, or the Quads game. It is incremental within an established program, not a breakthrough, but it is honest progress. It deserves a serious referee and a revision, not rejection.","headline":"A genuine extension of the AG(7,2) classification with a patchable proof gap in Lemma 4.3 and an overstated completeness claim in Corollary 1.3(2).","tokens_in":23099,"tokens_out":7398,"would_cite":true,"duration_ms":64203,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B25","51E10","51E15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Thirteen cards in Quad-128 always contain a quad, because every quad-free set in AG(7,2) has at most 12 points.","keywords":["finite geometry","affine geometry","AG(7,2)","caps","Sidon sets","quad-free sets","complete caps","Quad-128"],"falsifier":"An exhaustive computer search over all 13-subsets of $Z_2^{7}$ for a subset with no four elements summing to zero would settle Theorem 8.2: the existence of a single 13-cap refutes the classification. Short of that, a direct check of Lemma 4.3 — enumerating all triples of 5- or 7-subsets of an 8-set that avoid quads and asking whether any has union size 5 or 6 — would test the weakest link; if such a triple exists, the lemma and its consequences in Theorems 6.1, 6.5, and 7.4 would need repair.","tokens_in":21914,"feed_emoji":"🃏","tokens_out":8558,"duration_ms":82093,"temperature":0.7,"pith_summary":"This paper settles the question of how many cards you can lay out in the game Quad-128 before being forced to find a quad: the answer is 12. It proves that in the affine geometry AG(7,2) representing the deck, every quad-free set (cap) has at most 12 points, and it classifies all caps of size 10, 11, and 12 up to affine equivalence — two types of 10-caps, one type of 11-caps, and one type of 12-caps. Because 12-point caps exist and are maximal, they are exactly the complete caps, while every 10- or 11-point cap can be extended. The result is the first complete classification of caps of size 10 or larger in this seven-dimensional geometry, and it closes the cap problem for Quad-128.","feed_headline":"13 cards force a quad in Quad-128","feed_subtitle":"A full classification of quad-free sets in the 128-card deck pins the maximum at 12 and shows how the largest sets look.","key_machinery":"The classification rests on writing a 7-dimensional cap as an 8-point affine basis B plus a dependent set D, where each dependent point is a sum of five or seven basis elements (Lemma 3.2). The 'type' records how many basis elements each dependent point involves (5 or 7), and the 'extended type' records the sizes of pairwise intersections of these support sets (2, 3, or 4). Each allowed extended type admits a canonical dependent-set template, and Lemma 2.12 shows that two caps with bases fitting the same template are affinely equivalent, so a complete list of possible extended types yields a complete list of equivalence classes. Two tools prune the list: the Affine Basis Exchange Theorem (2.15) rebases a cap to a canonical type, and inclusion-exclusion counting combined with forbidden-triple and forbidden-quadruple lemmas rules out the remaining candidate types. Lemma 4.3 — that three dependent points' supports always cover all eight basis points — is the key structural fact that makes these counts go through.","core_discovery":"The central claim is that the quad-free subsets of AG(7,2) of size at least 10 form exactly four affine equivalence classes, where two caps are equivalent when an invertible affine map of the vector space moves one onto the other. Specifically, there are two classes of 10-point caps, one class of 11-point caps, and one class of 12-point caps. There is no 13-point cap, so the maximum cap size is 12, and a cap in AG(7,2) is complete (maximal quad-free) exactly when it has 12 points. Moreover, no 10- or 11-point cap is complete in any AG(n,2), because each such cap can be enlarged by one point within its own affine span; for dimensions above 7 the same conclusion follows from counting the first quad closure. Stated game-theoretically, 12 cards can avoid a quad in Quad-128, but 13 cards guarantee one.","pith_inferences":["The paper's remark that 7-dimensional caps are classified by extended type while higher-dimensional caps are not (Example 2.14) suggests that a classification in dimension 8 would need additional invariants beyond support-intersection sizes; testing the template method on AG(8,2) is the natural next step.","The exclusion of 13-caps rests on a lengthy inclusion-exclusion count; a machine-generated certificate or a direct SAT/backtracking search for a 13-cap would independently confirm that step, which is otherwise the least mechanically verified part of the proof.","The exact value M(7)=12 sits inside the known asymptotic bounds for M(n); the same support-type machinery could in principle attack the next dimension, where the extended-type invariant is known to be insufficient, so new combinatorial invariants would need to be introduced.","For game design, the 13-card threshold is a clean pigeonhole guarantee; the analogous threshold for larger EvenQuads decks (Z_2^n) is open, and the complete-cap results here show that maximum-size caps in dimension 7 are not the small complete caps constructed earlier."],"forward_implications":["Any 13 cards in Quad-128 are guaranteed to contain a quad, while 12 cards can be chosen quad-free.","There are exactly two affine equivalence classes of 10-caps, one class of 11-caps, and one class of 12-caps in AG(7,2).","The maximum size of a cap in AG(7,2) is 12, and the complete caps are precisely the 12-caps.","Every cap of size 10 or 11 in any AG(n,2) is incomplete — it can be enlarged by one point — so a complete cap in dimension 7 must have size 12.","Combined with earlier classifications of caps of size up to 9, this gives a full classification of all caps in dimensions n ≤ 7."],"supporting_citations":[{"why":"Supplies the base framework: encoding Quads cards as vectors in Z_2^n, the definition of caps, the classification of caps of size up to 9 and dimension up to 6, and the propositions (3.3 and 3.5) used as starting points.","marker":"[2]"},{"why":"Provides the earlier construction of small complete caps in Z_2^n, which the paper's completeness results for 10- and 11-caps extend and contrast with.","marker":"[8]"},{"why":"Gives the previous asymptotic bounds on the maximum cap size in Z_2^n, which the exact value M(7)=12 sharpens at n=7.","marker":"[10]"}],"fun_headline_variants":["Quad-free caps peak at 12 in AG(7,2)","No 13-cap: quad-free sets fully classified","12-point caps are complete and maximum in Quad-128","Exactly four cap classes for size 10 and up","Quad-128: 12 cards safe, 13 guarantee a quad"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof leans on the claim that the supports of any three dependent points in an 8-point basis together use all 8 basis points; the argument given only eliminates union size 7 and leaves union sizes 5 and 6 unaddressed.","fun_headline_variants_meta":{"raw":{"variants":["Quad-free caps peak at 12 in AG(7,2)","No 13-cap: quad-free sets fully classified","12-point caps are complete and maximum in Quad-128","Exactly four cap classes for size 10 and up","Quad-128: 12 cards safe, 13 guarantee a quad"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000638,"raw_usage":{"total_tokens":2882,"prompt_tokens":834,"completion_tokens":2048,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":450,"completion_tokens_details":{"reasoning_tokens":1964}},"tokens_in":450,"tokens_out":2048,"duration_ms":16772,"temperature":1.0,"reasoning_tokens":1964,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T18:34:34.641590+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"An exhaustive computer search over all 13-subsets of $Z_2^{7}$ for a subset with no four elements summing to zero would settle Theorem 8.2: the existence of a single 13-cap refutes the classification. Short of that, a direct check of Lemma 4.3 — enumerating all triples of 5- or 7-subsets of an 8-set that avoid quads and asking whether any has union size 5 or 6 — would test the weakest link; if such a triple exists, the lemma and its consequences in Theorems 6.1, 6.5, and 7.4 would need repair.","supporting_citations":[],"review_version":1}