{"id":"f9517e57-0a1e-4751-b269-0070e9a670ee","arxiv_id":"2602.21687","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Balanced permutation sequences guarantee ordinal PROP1 fairness after every day where they exist, but they exist only for n≤11 (none beyond 61), and full PROP2 existence remains open.","lead":"This paper studies schedules for repeatedly assigning n ranked items to n players, and shows that certain \"balanced\" schedules keep every player proportionally satisfied after every single day, not just after n days. It proves that the strongest such schedules exist only for finitely many n, and leaves open whether a weaker version works for all n.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Balanced sequences for n=7–9 are asserted but not shown; the existence range in the headline claim is unverified.","rationale":"The central implication (balanced ⇒ ordinal-PROP1) appears sound: one can formalize the non-integer case of Lemma 6.2 by checking the Hall-type condition n·|C_i∩[r]| ≥ t·r, which follows from Bal(t,j) as argued (for each r, count of original items ≤r is ≥ floor(tr/n), and adding item 1 covers the remainder). The common-ranking restriction is explicit in §6 and the abstract, so heterogeneous preferences are outside scope, not an internal gap. The reader's weakest_assumption therefore identifies a scope limitation rather than the most load-bearing risk. The real risk is evidential: the existence of balanced sequences for n=7,8,9 is asserted without printed sequences or code, and the weak-balance lemma (7.1) has no proof. These are the places where the positive results could fail. Because neither is a demonstrated error, the conditional verdict is appropriate; the concrete check above would settle them.","tokens_in":18007,"tokens_out":29261,"duration_ms":213178,"concrete_test":"Run an independent backtracking/search (e.g., SAT or CP) for balanced sequences of length n=7,8,9 under Definition 5.1 (Z^t_i[j] ≤ ceil(jn/t) for all t∈[n], j∈[t]). If all three exist, print them; if any does not, the §5 existence claim fails and the paper's positive range must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest existence claim — that balanced sequences exist for all n≤11 — is not independently verifiable from the manuscript. Tables 2–4 print balanced sequences for n=3,4,5,6,10,11 only. The text in §5 says n=7,8 were 'hand-crafted' and larger n were found by a C++ backtracking search, but no sequence is printed for n=7,8,9 and no code is released. Since Theorem 5.4/Table 5 only rule out n>61, the positive range n≤11 is the other half of the dichotomy; if any of n=7,8,9 lacks a balanced sequence, the claimed boundary is wrong. This is a gap in evidence, not a demonstrated contradiction. A related omission is Lemma 7.1: the proof is only 'similar to Lemma 6.2', so the weak-balance ⇒ PROP1 bridge that makes the n=12 example useful is unproved in the text.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies repeated assignment of n indivisible items to n players over n days, where each daily assignment is a permutation. It introduces 'top-balance', 'balance', and 'weak-balance' conditions on the sequence of permutations. The main positive result is Lemma 6.2: a balanced sequence yields an allocation that is ordinally-PROP1 after every day for any valuation consistent with the common item ranking. Theorem 5.4 proves that no balanced sequence exists for n>61, based on counting arguments over the first four days, and the text claims existence for n≤11. A weak-balance condition is claimed to also imply ordinal-PROP1 (Lemma 7.1), with an example for n=12. Section 8 then proposes a sufficient condition for ordinal PROP2 and leaves the existence question open.","tokens_in":18208,"tokens_out":16332,"duration_ms":134429,"significance":"If fully proved, the paper offers a novel and elegant bridge between combinatorial design (balanced Latin squares) and ordinal fair division. The dominance/bijection method in Lemma 6.2 is a clean, parameter-free sufficient condition, and the impossibility results are nontrivial and falsifiable. The paper is also candid about the open question for PROP2. However, several load-bearing proof details are currently missing, and the claimed positive examples for n=7–9 are not documented, so the existing version is not fully verified.","major_comments":[{"comment":"The bijection f: D_i → Z_t^N is only sketched. For example, the n clones of Z_t^i[1] are mapped to 'some lower-valued items' without specifying which items or counting them. The proof needs an explicit interval decomposition of Z_t^N (e.g., blocks [⌊(j−1)n/t⌋+1, ⌊jn/t⌋] with boundary adjustments) and a verification that Bal(t,j) puts each cloned item in the correct block, or a Hall-type dominance argument. As written, the central implication balance ⇒ PROP1 is not fully established.","section":"Lemma 6.2, non-integer n/t case"},{"comment":"The proof is omitted entirely, with only 'similar to Lemma 6.2' and an illustration in Table 7. Since weak balance is the property that makes the n=12 sequence in Table 8 an ordinal-PROP1 sequence, this missing proof is load-bearing. Provide a complete proof, or at least a formal reduction to Lemma 6.2 with the precise adjusted thresholds.","section":"Lemma 7.1"},{"comment":"The existence claim 'There are balanced sequences for natural numbers n≤11' is not fully supported. Tables 2–4 give sequences for n=3,4,5,6,10,11 but not for n=7,8,9. The text says these were hand-crafted, but no sequences are printed and no code is released. The subsequent statement that n=12 is the first value with no balanced sequence depends on these missing examples. Please include the three sequences in an appendix or supplementary material, or provide a verifiable construction.","section":"Section 5"},{"comment":"Conditions (3) and (4) are stated as sufficient for ordinal PROP2 without proof. The abstract promises a condition that 'guarantees' ordinal PROP2, so the reader needs a proof. A derivation analogous to Lemma 6.2 with two added items should be supplied, or a counterexample to sufficiency if the claim is false. In addition, clarify how 'adding two items' works when fewer than two items are available in the allocation.","section":"Section 8"}],"minor_comments":[{"comment":"There is a typo in the block description: 'items n/t+1,2,...,2n/t' should read 'items n/t+1, n/t+2, ..., 2n/t'; similarly for the j-block.","section":"Lemma 6.2, integer case"},{"comment":"The assertion that extension to heterogeneous preferences 'should be easy' is not proved. Either provide a proof sketch or state explicitly that the results are for a common ranking.","section":"Footnote 2"},{"comment":"The phrase 'some lower-valued items' is used informally. It would be clearer to define the exact indices or use a multiset notation.","section":"Section 8"}],"recommendation":"major_revision","confidential_remarks":"The key issues are fixable: the proof of Lemma 7.1 and the non-integer case of Lemma 6.2 need to be written out, and the n=7,8,9 balanced sequences should be listed or the search code released. Without these, the main claims are not fully verifiable. I would also suggest making the C++ backtracking code available for reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Erel and Adams have a solid combinatorial idea here. The balance condition (1) is a new object — a strengthening of Latin squares that controls the order statistics of each player's bundle at every prefix. The main positive result, that balanced sequences imply perpetual ordinal PROP1, is plausible and the proof for integer n/t is actually clean. The impossibility results (no balanced sequence for n>61) are detailed and look correct. This is a real step for repeated assignment, where most prior work only guarantees fairness at the end or under cardinal valuations.\n\nThe paper deserves a careful referee, but it is not ready as-is. The soft spots are:\n\n— The non-integer case of Lemma 6.2 is a sketch, not a formal bijection. Phrases like 'some lower-valued items' need to be replaced with an actual counting argument and an explicit mapping. This is fixable and probably minor.\n\n— Lemma 7.1, the weak-balance analogue, is simply asserted — 'the proof follows a similar structure' — with no proof. Since the weak condition differs exactly at the integers, the dominance bijection doesn't follow by analogy; it needs its own argument. This is the largest mathematical gap.\n\n— The existence of balanced sequences for n=7,8,9 is claimed but no sequences are printed, no code is released. Tables cover only 3,4,5,6,10,11. The n≤11 range is half of the main dichotomy, so this evidence gap matters.\n\n— The arXiv metadata abstract contradicts the body: one says n≤12 and n>108, the other n≤11 and n>61. That's sloppy and should be fixed.\n\n— The sufficiency of conditions (3)/(4) for PROP2 is asserted without proof, and the footnote about heterogeneous preferences 'should be easy' is not a proof.\n\nNone of these is a refutation. The core idea is sound and the open question at the end is attractive. A referee could reasonably ask for the missing pieces and accept after revision. I'd send it to peer review.","headline":"A new combinatorial condition — balanced permutation sequences — implies per-day ordinal PROP1, and the paper deserves refereeing despite several unproven pieces.","tokens_in":18666,"tokens_out":4991,"would_cite":true,"duration_ms":41459,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["05B15","91B32"],"pacs":[],"model":"deepseek-v4-flash","headline":"A 'balanced' permutation sequence — every player's j-th best item among the top ceil(jn/t) after day t — makes the cumulative allocation proportionally fair up to one item after every day, for any valuation consistent with the common rankin","keywords":["balanced permutation sequence","repeated assignment","ordinal proportionality","PROP1","Latin square","fair division","indivisible goods","perpetual fairness"],"falsifier":"Run an exhaustive backtracking search for a balanced permutation sequence of length n=12; the paper asserts none exists (Proposition 5.3), so any valid sequence found would disprove the impossibility claim. Alternatively, check n=6k for k=19 (n=114) for a weakly balanced sequence, which the paper claims does not exist.","tokens_in":17898,"feed_emoji":"⚖️","tokens_out":9098,"duration_ms":76051,"temperature":0.7,"pith_summary":"The paper studies a repeated assignment problem: n players, n items, one item per player per day. It asks for a schedule that is fair after every single day, not only after n days. The main claim is that a 'balanced' sequence of permutations — one in which after t days every player's j-th best item is among the top ceil(jn/t) items — guarantees that the cumulative allocation is proportional up to one item (PROP1) for all valuations consistent with a common item ranking. The paper proves this via a dominance bijection between each player's bundle and the set of all allocated items. It also shows that such balanced sequences are scarce: they exist for n up to 11 but none exist for n>61. Weaker balance conditions still yield PROP1 for some larger n, and a still weaker condition that would yield PROP2 is left as an open question.","feed_headline":"Balanced permutations guarantee daily fairness for n up to 11","feed_subtitle":"The same ordering condition fails for n>61, and a weaker version still leaves the two-item case open.","key_machinery":"The key object is a 'balanced' sequence of permutations, a Latin square reinforced with the condition that after t days, for every player, the j-th best item in their bundle is no worse than the item ranked ceil(jn/t). This ordering condition carries the whole argument: it enables a bijection between n clones of the player's augmented bundle and the n·t allocated items, with each clone matched to an item of equal or lower value, thereby proving ordinal PROP1. The balance condition is exactly a quantitative statement about how evenly the top items are distributed across players and days.","core_discovery":"The central discovery is the balance condition: after t days, for each player i, the j-th best item in i's bundle satisfies Z^i_t[j] ≤ ceil(jn/t). This condition is both sufficient and (in a sense) necessary for a strong ordinal fairness guarantee: a balanced repeated assignment is perpetually PROP1, meaning that after every day, each player's cumulative bundle, when augmented by one item, is worth at least 1/n of the total value of all items allocated so far, for every valuation consistent with the common ranking. The proof constructs a bijection between n clones of the player's bundle and the multiset of all n×t allocated items, showing each clone is at least as good as the item it is matc","pith_inferences":["Because the balance condition is purely ordinal and valuation-free, the same fixed schedule works for any positive valuations sharing the same item ranking; one could also try running the construction with chores by reversing the order, an extension the paper does not develop.","The impossibility proofs examine only the first four days of the schedule, so the nonexistence of balanced sequences for large n is driven by a very short initial segment; it may be possible to characterize the exact set of n that admit balanced sequences by extending the counting argument or with a larger computer search.","The paper asserts in a footnote that extending the positive result to players with different rankings 'should be easy' but gives no proof; a rigorous heterogeneous-preferences version of the main lemma would be a natural follow-up and would make the result applicable beyond common ordinal preferences.","If a sequence satisfying the PROP2 condition (4) exists for every n, it would give a universal guarantee of proportionality up to two items for all n; conversely, a further impossibility would show that no fixed c can be guaranteed for all n, answering the paper's open question in the negative."],"forward_implications":["For every n up to 11, there is a concrete schedule where after each day t, each player's j-th best item lies within the top ceil(jn/t) items; the paper gives tables for n=3,4,5,6,10,11.","This balance condition guarantees ordinal PROP1 after every day, independent of the exact numerical values of items, for any valuation consistent with the common ranking.","No balanced schedule exists for n=12 or any n>61; the impossibility is proved by a counting argument using only the first four days.","A relaxed condition (replacing ceil(jn/t) by floor(jn/t)+1) also guarantees ordinal PROP1, and a weakly balanced schedule exists for n=12; however, for n=6k with k≥19 even this relaxed condition fails.","The paper leaves open whether an even weaker condition, Z^i_t[j] ≤ floor((j+1)n/t+1), which would guarantee ordinal PROP2, can be satisfied for every n."],"fun_headline_variants":["Balanced permutations guarantee daily fairness for n≤12","Strong balance gives PROP1 daily but only for n up to 12","Balanced sequences: PROP1 fairness for n≤12, but not for n>108","Perpetual PROP1 from balanced permutations: only for n≤12"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The entire fairness guarantee assumes all players share a single item ranking (1 > 2 > ... > n); if players value items in different orders, the balance condition built on the global ranking does not by itself ensure each player's individual PROP1, and the paper only sketches — without proving — an extension to that case.","fun_headline_variants_meta":{"raw":{"variants":["Balanced permutations guarantee daily fairness for n≤12","Strong balance gives PROP1 daily but only for n up to 12","Balanced sequences: PROP1 fairness for n≤12, but not for n>108","Perpetual PROP1 from balanced permutations: only for n≤12"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00208,"raw_usage":{"total_tokens":7941,"prompt_tokens":772,"completion_tokens":7169,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":516,"completion_tokens_details":{"reasoning_tokens":7090}},"tokens_in":516,"tokens_out":7169,"duration_ms":47074,"temperature":1.0,"reasoning_tokens":7090,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-02T20:57:57.203393+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run an exhaustive backtracking search for a balanced permutation sequence of length n=12; the paper asserts none exists (Proposition 5.3), so any valid sequence found would disprove the impossibility claim. Alternatively, check n=6k for k=19 (n=114) for a weakly balanced sequence, which the paper claims does not exist.","supporting_citations":[],"review_version":1}