{"id":"5cd26805-e61f-45df-8fb5-60ab84e525b8","arxiv_id":"2608.11139","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For four-player XOR games, commuting-operator value one forces a perfect GHZ-equatorial strategy exactly when each player has at most three active questions; the Klein four-group game shows the threshold is sharp.","lead":"Quantum XOR games with up to three questions per player always have a simple four-qubit entangled realization whenever a perfect general strategy exists. The paper proves the cliff edge: with four questions per player, a small Klein four-group game escapes this simple realization even though a perfect general strategy still exists.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the reader's flagged Proposition A.7 checks out; the remaining caveat is reproducibility of the Lean formalization, not a mathematical gap.","rationale":"The paper's central claim rests on the exact circuit-lifting theorem for four ternary pages and the all-length Magnus obstruction for the Klein game. The reader identified Proposition A.7 as the most fragile sub-claim because it is marked 'Supporting' in the Lean correspondence and combines several nontrivial ingredients: a ternary Hamming-code coset count, a double-point enumeration, and a Gram-matrix connectivity argument. I re-derived the key identities and found them coherent: the feature vectors defined in the text indeed yield G_ij = 3 - d(q_i,q_j), so the nonzero graph joins distances 1, 2, and 4; the coset argument gives 3n1 + 2n2 = 9; and the question-block size lower bound gives 2n1 + n2 - n4 >= 3, forcing the nonzero graph to have at most six edges, contradicting connectivity on ten vertices. I also verified the more arithmetic part of Theorem 4.2: the listed linear combination of equation (4.4) instances, together with the specified multiples of (4.5) and (4.6), cancels all P variables and produces the displayed 8R12 + 8(Q13 - Q23 - Q31 + Q32) + 8(-S31 + S32) = 4v^2, hence v^2 is even, contradicting v odd. No missing case, circular step, or unexplained assumption was found. The remaining weakness is external: the Lean formalization is cited without a commit hash, and several key statements are only 'Supporting' in the correspondence table, so the formal proof cannot be independently audited from the manuscript alone. This affects confidence, not the mathematical soundness as presented, so the ACCEPT verdict remains unchanged.","tokens_in":19680,"tokens_out":52688,"duration_ms":460547,"concrete_test":"Pin the GitHub repository [TZB+es] to a specific commit, build with the tracked Lean 4.30.0 toolchain, and run #print axioms on the Lean declarations QIT.XORGame.tenBallCover_not_fullRationalCircuit and QIT.XORGame.primitiveCircuit_lifts_fourByThree, confirming the axiom set matches the seven disclosed native_decide bridge axioms plus the three standard axioms.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I reviewed the candidate weak point, Proposition A.7, in detail and found it internally consistent. With the stated feature vectors, the Gram matrix is G_ij = 3 - d(q_i,q_j), so the nonzero off-diagonal graph joins distances 1, 2, and 4, exactly as claimed. The double-point count 3n1 + 2n2 = 9 follows from the Hamming-code coset argument, and the agreement lower bound 3n1 + 2n2 + n3 >= 48 together with n1+n2+n3+n4 = 45 gives 2n1 + n2 - n4 >= 3, hence n1+n2+n4 = 9 - (2n1+n2-n4) <= 6. A connected graph on ten vertices needs at least nine edges, so the contradiction is valid. I also spot-checked the Magnus coefficient arithmetic in Theorem 4.2: the weighted equation table cancels all P terms and leaves 8R12 + 8(Q13 - Q23 - Q31 + Q32) + 8(-S31 + S32) = 4v^2, forcing v even, which contradicts the oddness of v derived from the target parity. No internal inconsistency or missing case surfaced. The only respect in which the argument is not fully closed from the manuscript text is the formal verification: the repository URL carries no commit hash, and Proposition A.7 and Theorem 3.1 are marked 'Supporting' rather than 'Direct' in the Lean correspondence. This is a reproducibility caveat, not a demonstrated mathematical error.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper determines the exact threshold in the number of active questions per player at which perfect commuting-operator strategies for four-player XOR games can fail to admit a GHZ-equatorial (MERP) realization. Theorem 1.1 states that every four-player XOR game with at most three active questions per player and commuting-operator value one admits a perfect MERP strategy, while the Klein four-group game with four active questions per player has commuting-operator value one but no perfect MERP strategy. The positive direction is proved by reducing to the identity L_E = R_E for every support E subset [3]^4, obtained through an exact circuit-lifting theorem for primitive support-minimal circuits, with signed-occurrence forest matchings and ternary Hamming-geometry exclusions handling the ordering problem. The sharpness direction is proved through a Cayley-game analysis and a degree-two Magnus-expansion obstruction showing that the Klein game admits a PREF but no true refutation of any length. An appendix provides the extremal Hamming-cube exclusions and a Lean 4 formalization correspondence.","tokens_in":19885,"tokens_out":52774,"duration_ms":428442,"significance":"If correct, the paper resolves a natural boundary problem left open after the three-player collapse theorem of Bene Watts and Helton and the MERP-PREF duality of Bene Watts, Harrow, Kanwar, and Natarajan. The four-player, three-question completeness theorem and the four-question separation are both sharp, and the proof introduces transferable techniques: exact lifting of integral incidence circuits into ordered noncommutative refutations, and all-length obstructions via Magnus coefficients. I checked the two most delicate points—the ten-point Hamming exclusion in Proposition A.7 and the weighted Magnus sum in Theorem 4.2—and found the arguments internally consistent. The accompanying Lean development, with a statement-by-statement correspondence and no sorry or admit declarations, is a substantial verification asset, although the repository lacks a commit hash and some key statements are marked as supporting rather than direct.","major_comments":[],"minor_comments":[{"comment":"The GitHub repository [TZB+es] is cited without a commit hash, and Table 2 marks Proposition A.7 and Theorem 3.1 as 'Supporting' rather than 'Direct'; since these are the main combinatorial and lifting inputs, please pin the repository revision and clarify the exact formal status, or soften the claim that the principal threshold theorem is fully formalized.","section":"Appendix C"},{"comment":"In the final paragraph of Section 5, 'An robust game-algebra methods may provide a useful framework' should read 'Robust game-algebra methods may provide a useful framework'.","section":"Section 5"},{"comment":"The weighted-sum table in Theorem 4.2 is central but presented without derivation; adding a short verification that the listed coefficients annihilate the P family and leave exactly 8R12, 8(Q13 - Q23 - Q31 + Q32), and 8(-S31 + S32) would make the proof substantially easier to audit.","section":"Section 4.2"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is a strong fit for the journal. The only non-mathematical caveats are the reproducibility of the Lean repository and the AI-assisted-tools disclosure, which the editor may wish to handle according to journal policy."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this paper closes the four-player boundary case that was explicitly open after the two-question regime and the three-player collapse. It proves that three active questions per player are still safe—commuting-operator value one forces a perfect GHZ-equatorial strategy—and builds the Klein four-group game at four questions that separates the two. Four is sharp. That is genuinely new, and the main theorem is well supported.\n\nThe proof is the real work. The circuit-lifting argument on four ternary pages is intricate but coherent: primitive circuits localized to at most ten clauses, signed-occurrence forest matchings to order the occurrences, and ternary Hamming geometry to kill the exceptional configurations. I checked the two most fragile spots. Proposition A.7's exclusion of a ten-ball cover: the Gram matrix computation, the ternary Hamming-code coset counting, and the connectivity/edge-count contradiction all line up. And the degree-two Magnus certificate in Theorem 4.2: the weighted equation table cancels correctly and forces v even against v odd. I did not find a gap.\n\nThe soft spots are minor and mostly about verification hygiene. The Lean formalization is a real asset—no sorry/admit, and a statement-by-statement correspondence—but the repository has no commit hash, and several load-bearing statements (Theorem 3.1, Proposition A.7) are marked 'Supporting' rather than 'Direct' in Lean, while the all-length obstruction's general Magnus lemma is 'Alternative.' So the kernel-checked claim is slightly narrower than the manuscript suggests. That is a reproducibility caveat, not a demonstrated mathematical error. The authors are honest about it in the appendix.\n\nAlso worth noting: the discussion is candid about what the result does not settle—whether the Klein game has a finite-dimensional non-MERP perfect strategy, and quantitative robustness below the threshold. The citation pattern to WHKN19/WH23/WHK23 is appropriate, with no red flags. The use of an AI-assisted harness is disclosed; the math stands on its own.\n\nWho is this for? Anyone working on MERP completeness, XOR games, or noncommutative obstruction spaces. The appendix is dense but the main text is readable. I would send it to a serious referee, not desk reject. My only request would be a pinned commit hash and a clearer mapping of which Lean statements are verbatim versus supporting.","headline":"Sharp k=4 threshold for four-player XOR MERP completeness, with a solid proof apparatus and a Lean formalization that deserves a referee despite reproducibility nits.","tokens_in":20483,"tokens_out":2085,"would_cite":true,"duration_ms":19094,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81P40","20F05","05C70"],"pacs":[],"model":"deepseek-v4-flash","headline":"Four-player XOR games with at most three active questions per player have a perfect GHZ-equatorial strategy whenever their commuting-operator value is one; at four questions, the Klein four-group game separates the two properties.","keywords":["four-player XOR games","GHZ-equatorial strategies","MERP strategies","commuting-operator value","primitive circuits","Klein four-group game","Magnus expansion","ternary Hamming cube"],"falsifier":"Because the positive half concerns only supports inside the finite set $[3]^4$, an exhaustive computation of the two obstruction spaces $L_E$ and $R_E$ over all reduced supports with at most three active questions per player would settle Theorem 1.1(i): any support with $L_E \\neq R_E$ would be a counterexample. For the negative half, exhibiting any balanced word over the eight Klein clauses in which an odd number of $O_0$ clauses occur and each player's projected word freely reduces to the identity would refute the all-length obstruction and with it Theorem 1.1(ii).","tokens_in":19412,"feed_emoji":"🎲","tokens_out":17160,"duration_ms":142182,"temperature":0.7,"pith_summary":"What is the smallest number of active questions per player at which a perfectly winnable four-player XOR game can stop being realizable on the four-qubit GHZ state with equatorial measurements? The paper's answer is three. Every four-player XOR game with at most three active questions per player and commuting-operator value one—perfect winnability in the commuting-operator model—admits a perfect GHZ-equatorial strategy, meaning perfect play reduces to consistency of additive phase equations. At four active questions per player, a single example, the Klein four-group game, has commuting-operator value one yet inconsistent phase equations, so no GHZ-equatorial strategy exists. Four is therefore the sharp local-question threshold, and the paper identifies the exact mechanism: abelian phase obstructions lift to ordered noncommutative refutations up to three questions, while the Klein translation pattern produces an abelian obstruction with no ordered noncommutative refutation of any length.","feed_headline":"Three questions per player: perfect XOR games stay GHZ-realizable","feed_subtitle":"At four questions, a Klein four-group game wins perfectly under commuting operators yet has no GHZ-equatorial strategy.","key_machinery":"The argument runs through two obstruction spaces attached to a reduced support $E$: the abelian space $L_E$ of parity images of integer incidence relations among the clause rows, and the noncommutative space $R_E$ generated by parity vectors of balanced words, ordered clause words in which every player's projected word freely reduces to the identity. A parity-permuted refutation, an element of $L_E$ with odd target parity, blocks a perfect GHZ-equatorial strategy; a true refutation, a balanced word with odd target parity, blocks commuting-operator value one. The positive half proves $L_E=R_E$ for every $E\\subseteq[3]^4$ via an exact circuit-lifting theorem: every primitive support-minimal circuit, a minimal integer dependence with coprime coefficients, is realized by a balanced word with the same signed multiplicities, using signed-occurrence forest matchings to build the order and ternary Hamming geometry to eliminate the exceptional ten-clause supports. The negative half shows that in the Klein four-group game a parity-permuted refutation exists while no true refutation of any length does, using an even-subgroup free normal form and degree-one and degree-two Magnus coefficients to produce an integral parity certificate.","core_discovery":"The central claim, stated as Theorem 1.1, is an exact equivalence at three questions and a separation at four. For every finite four-player XOR game in which each player has at most three active questions, the paper proves that commuting-operator value one holds if and only if the game has a perfect MERP strategy, a realization on the four-qubit GHZ state with equatorial single-qubit observables such that clause satisfaction becomes a set of additive angle equations. Conversely, the paper constructs the Klein four-group game, with four active questions per player and eight uniformly weighted clauses, whose commuting-operator value is one but whose phase equations are inconsistent. Because the three-question direction is affirmative and the four-question direction admits a counterexample, the paper concludes that four is the smallest number of active questions per player at which such a separation can occur.","pith_inferences":["The paper leaves open whether the Klein game admits a perfect finite-dimensional tensor-product strategy; settling it either way would sharpen the meaning of the four-question threshold.","A plausible extension of the Magnus argument is that any group whose Cayley translation pattern forces a phase contradiction yields a value-one game without GHZ-equatorial strategies, not just elementary abelian 2-groups.","A natural extension would be a stability theorem: for at most three active questions and value $1-\\varepsilon$, a GHZ-equatorial strategy of value at least $1-f(\\varepsilon)$ with an explicit rate would carry the structural result into the noisy regime.","Because the support class $[3]^4$ is finite, the equality $L_E=R_E$ could be converted into a finite certificate family, making the three-question boundary in principle decidable by exhaustive support-level verification."],"forward_implications":["For games with at most three active questions per player, perfect commuting-operator play and perfect GHZ-equatorial play are the same property: no exotic operator-algebraic construction is needed at value one.","The Klein four-group game is a separation example at four active questions: its phase system is inconsistent, so the GHZ-equatorial model cannot realize it, even though commuting-operator value one persists.","On every ternary four-partite support the abelian and noncommutative obstruction spaces coincide, so every parity-permuted refutation can be ordered into a genuine refutation; in particular any abelian obstruction at three questions pushes the commuting-operator value strictly below one.","The threshold statement is structural: it depends only on the reduced support and target vector, not on the positive clause weights or on whether the distribution over clauses is uniform."],"supporting_citations":[{"why":"It supplies the MERP–PREF duality used to identify the absence of a perfect GHZ-equatorial strategy with the existence of a parity-permuted incidence relation.","marker":"[WHKN19]"},{"why":"It supplies the game-group criterion by which commuting-operator value one is equivalent to the absence of a true refutation, and it proves the three-player collapse that this paper extends to four players.","marker":"[WH23]"},{"why":"It establishes the two-question-per-player completeness regime that anchors the sharp-threshold question at the other end.","marker":"[WW01]"},{"why":"It provides the noncommutative Nullstellensatz formulation that frames a true refutation as an ordered occurrence word in the game group.","marker":"[WHK23]"},{"why":"It contributes the Magnus expansion whose degree-one and degree-two coefficients are assembled into the integral parity certificate for the Klein game.","marker":"[Mag37]"},{"why":"It supplies the free-group and Reidemeister–Schreier background used for the even-subgroup normal form and the Magnus coefficient identities.","marker":"[MKS76]"}],"fun_headline_variants":["Four questions: sharp cutoff for GHZ-equatorial XOR games","Three questions guarantee GHZ strategies; four break them","XOR games: GHZ perfect play possible only up to 3 questions per player","Sharp threshold: 4-player XOR games fail GHZ at 4 questions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the extremal counting claim that no ten-clause support on the full $3^4$ grid, with all four players using all three questions, can be a primitive minimal dependence when the clauses' neighborhoods cover all 81 question tuples; if one exceptional configuration were missed, the proof that every value-one three-question game has a GHZ-equatorial strategy would have a gap.","fun_headline_variants_meta":{"raw":{"variants":["Four questions: sharp cutoff for GHZ-equatorial XOR games","Three questions guarantee GHZ strategies; four break them","XOR games: GHZ perfect play possible only up to 3 questions per player","Sharp threshold: 4-player XOR games fail GHZ at 4 questions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000302,"raw_usage":{"total_tokens":1721,"prompt_tokens":912,"completion_tokens":809,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":732}},"tokens_in":528,"tokens_out":809,"duration_ms":6300,"temperature":1.0,"reasoning_tokens":732,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:25:30.793471+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Because the positive half concerns only supports inside the finite set $[3]^4$, an exhaustive computation of the two obstruction spaces $L_E$ and $R_E$ over all reduced supports with at most three active questions per player would settle Theorem 1.1(i): any support with $L_E \\neq R_E$ would be a counterexample. For the negative half, exhibiting any balanced word over the eight Klein clauses in which an odd number of $O_0$ clauses occur and each player's projected word freely reduces to the identity would refute the all-length obstruction and with it Theorem 1.1(ii).","supporting_citations":[],"review_version":1}