{"id":"06294e5a-729c-475a-a9c3-b0a9a7191cab","arxiv_id":"2504.14195","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"River is a tree-based Condorcet voting rule that satisfies independence of Pareto-dominated alternatives, a property not shared by the other Split Cycle refinements according to the paper.","lead":"This paper introduces River, a voting rule that builds a tree from pairwise majority margins and selects the root as winner, and it claims the rule can resist manipulation by adding Pareto-dominated alternatives. The method is a simplified Ranked Pairs, and the paper presents proofs that it satisfies IPDA, a property it says other major rules lack.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 4.4's Ranked Pairs counterexample is invalid: printed profile P2 does not yield the claimed margin graph or winners, so the paper's unique-IPDA claim is unsupported.","rationale":"The reader's weakest-assumption is exactly where the central argument fails. I verified the arithmetic: the printed P2 cannot produce the margin graph the proof describes, and in no parsing does it yield the claimed winner change d vs b. This is not a minor typo: Theorem 4.4 is the only evidence that Ranked Pairs violates IPDA, and the abstract's headline contribution depends on that. The positive results about River itself (Propositions 3.2, 3.3, Theorem 4.2, Theorems 4.6/4.8) may still be correct, but the paper's distinctive claim—unlike the other three methods—is unsupported. The clone-independence assertion in the abstract is also unproven, but the counterexample flaw is the decisive issue. I recommend no change to the reader's REJECT verdict; the paper needs a corrected counterexample or a revised claim.","tokens_in":14397,"tokens_out":15242,"duration_ms":105515,"concrete_test":"Recompute the margin graph from the printed Profile P2 in Appendix B (ballot groups: 7: d>b>a>c; 5: b>a>c>d; 2: c>d>b>a; 2: b>a>d>c; 2: c>d>b>a; 1: c>b>d>a; 1: b>d>a>c; 1: d>c>b>a; 1: b>c>a>d). Then run Ranked Pairs (with any fixed tiebreaker for equal margins) on P2 and on P2−a. For the paper's claim to hold, the margin graph must equal the one in the proof (b→a 22, b→c 10, a→c 8, d→a 4, c→d 2, d→b 2) and the winners must be RP(P2)={d}, RP(P2−a)={b}. If the margins or winners differ, Theorem 4.4's Ranked Pairs counterexample fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central comparative claim—River is the only Split Cycle refinement among Ranked Pairs, Beat Path, and Stable Voting that satisfies IPDA—rests on Theorem 4.4, whose proof uses profile P2 to show Ranked Pairs violates IPDA. The printed P2 does not support this. Using the 22 voters listed in Appendix B (groups of 7,5,2,2,2,1,1,1,1 over d>b>a>c, b>a>c>d, c>d>b>a, b>a>d>c, c>d>b>a, c>b>d>a, b>d>a>c, d>c>b>a, b>c>a>d), direct calculation gives margins b→a 22, b→c 10, a→c 10, d→a 6, d→b 2, and c–d 0. The proof's described sequence assumes a→c 8, d→a 4, c→d 2, d→b 2. With the actual printed margins, Ranked Pairs selects d both on P2 and on P2−a; with the alternative reconstruction the reader used, it selects b both times. Either way, RP(P2)={d} and RP(P2−a)={b} do not both hold. Since no other counterexample is supplied for Ranked Pairs, Theorem 4.4's conclusion for Ranked Pairs is not established, and the abstract's uniqueness claim is unsupported. A secondary issue: the abstract states River shares independence of clones with the other methods, but no proof of clone independence appears in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces River, a resolute Condorcet-consistent voting rule that constructs a directed tree by processing majority-margin edges in decreasing order while avoiding cycles and multiple incoming edges. The authors claim that River is a refinement of Split Cycle, satisfies ISDA, IPDA, and IQDA under appropriate tiebreaker conditions, and is simpler to compute and explain than Ranked Pairs, Beat Path, and Stable Voting. The central comparative claim is that unlike those methods, River satisfies independence of Pareto-dominated alternatives (IPDA), with Theorem 4.4 providing counterexamples for the other rules and Theorems 4.6, 4.8, and 4.10 proving IPDA/IQDA for River.","tokens_in":14789,"tokens_out":18208,"duration_ms":153907,"significance":"If fully established, the contribution would be genuinely useful: River is a resolute rule that is easy to describe, produces an interpretable tree certificate, and appears to avoid the Pareto-dominated-alternative spoiler effect. The proof strategy via covering and quasi-Pareto domination is elegant, and the paper is careful about tiebreaker conditions for general profiles. However, the advertised comparative uniqueness claim is not currently supported because the Ranked Pairs counterexample in Theorem 4.4 is invalid for the printed profile P2, and the paper's abstract claims independence of clones without providing a proof or precise reference. The constructive contribution about River itself is plausible and well motivated, but the paper needs a corrected or replaced counterexample and a complete proof of Theorem 4.8 before the central claims can be accepted.","major_comments":[{"comment":"The Ranked Pairs counterexample is not supported by the printed profile. Direct calculation from the nine voter groups in Profile P2 gives margins b→a 22, b→c 10, a→c 8, d→a 6, d→b 2, and m(c,d)=0. In the full profile, Ranked Pairs thus adds (b,a), (b,c), (a,c), (d,a), and then (d,b), skipping only the zero-margin edge (c,d), with winner RP(P2)={d}. In P2−a the positive margins are b→c 10 and d→b 2, with c–d tied at 0, so Ranked Pairs adds b→c and d→b and yields RP(P2−a)={d}, not {b}. The proof's claim that Ranked Pairs 'adds (b,c) and (c,d), skipping (d,b)' requires m(c,d)>m(d,b), which the printed profile does not provide. Since Theorem 4.4 is the sole support for the abstract and Table 1 claims that Ranked Pairs violates IPDA, the central comparative claim is not established as written.","section":"Theorem 4.4 and Appendix B (Profile P2)"},{"comment":"The abstract and introduction state that River 'shares with those many desirable properties, including independence of clones,' but the manuscript contains no definition of clone independence, no theorem establishing it for River, and no precise citation to a proof. Section 4 and Table 1 do not treat independence of clones at all. This advertised property is therefore unverified in this manuscript; either add a proof or a specific reference to a proof, or remove the claim from the abstract and introduction.","section":"Abstract and Section 1 (no theorem)"},{"comment":"The proof of Theorem 4.8 is not actually supplied: the appendix says it is 'literally the same as the proof of Theorem 4.10, only with all occurrences of the prefix quasi- removed.' This reduction is not literal, because the proof of Theorem 4.10 uses the quasi-Pareto-consistency conditions from Definition D.1—specifically that every edge (x,z) is ranked after both (y,z) and (y,x) when y quasi-Pareto-dominates x—and those conditions are not implied by Pareto-consistency alone. The argument for Theorem 4.8 requires its own proof, or at least an explicit explanation of why the extra quasi-Pareto conditions are not needed in the Pareto-dominated case.","section":"Theorem 4.8 and Appendix D"}],"minor_comments":[{"comment":"The sentence describing the Ranked Pairs sequence contains 'adds (d,a)' twice, and the intended sequence is unclear; this should be corrected regardless of the counterexample's validity.","section":"Theorem 4.4 proof text"},{"comment":"There is a spelling error: 'absense' should be 'absence.'","section":"Section 4.1"},{"comment":"There is a spelling error: 'inpedendence' should be 'independence.'","section":"Section 4.2"},{"comment":"Observation 4.5 is stated with an intuitive explanation rather than a formal proof; since it is used later in Theorem 4.6, a one-line derivation of the covering inequalities would improve the presentation.","section":"Observation 4.5"}],"recommendation":"major_revision","confidential_remarks":"The Ranked Pairs counterexample in Theorem 4.4 is demonstrably wrong for the printed profile P2: direct computation gives RP(P2)={d} and RP(P2-a)={d}, so no IPDA violation is exhibited. Since this is the only counterexample offered for Ranked Pairs, the paper's uniqueness claim is unsupported unless a corrected profile is supplied. I also recommend requiring a genuine proof of Theorem 4.8, since the 'literally the same' reduction to Theorem 4.10 is not valid as stated. The paper may be salvageable, but the comparative claims need substantive repair."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the paper's core contribution is real, but the headline comparison is not. River's proofs for ISDA/IPDA/IQDA are new and look sound; the Ranked Pairs counterexample in Theorem 4.4 is wrong for the printed profile. Directly from P2 (22 voters, as printed) the margins are b→a 22, b→c 10, a→c 8, d→a 6, d→b 2, and c–d 0. Ranked Pairs picks d both on P2 and on P2−a, so there is no IPDA violation. The proof describes a different margin graph than the one P2 actually induces. That is load-bearing: the abstract's claim that River is the only Split Cycle refinement satisfying IPDA relies on this counterexample. The Beat Path and Stable Voting examples I haven't checked carefully, but they don't repair the Ranked Pairs point.\n\nWhat is genuinely good: the definition of River is simple, the tree certificate idea is appealing, and the theorems that River is a resolute Split Cycle refinement satisfying ISDA/IPDA (with appropriate tiebreakers) and IQDA are proven in a way that seems correct. The Observation that Pareto domination implies covering is clean and does real work. For a reader who wants a transparent Condorcet method, the River rule itself is worth knowing.\n\nSoft spots beyond Theorem 4.4: the abstract asserts independence of clones, but the body neither proves nor cites a proof for River; that's a missing argument. The paper also leans on 'by hand' computability without complexity analysis, which is a minor omission, not a flaw.\n\nWho is this for: researchers in computational social choice and anyone using pairwise preference aggregation for LLM alignment or citizen assemblies. The paper deserves a serious referee; the error is local and fixable. I would send it back for major revision—replace or repair the Ranked Pairs counterexample, and either prove clone independence or soften the abstract—rather than reject it outright. If the uniqueness claim cannot be restored, the paper still stands as a contribution on River's own axioms.","headline":"Solid River proofs, broken Ranked Pairs counterexample: the central comparative claim doesn't survive the printed profile, but the core IPDA result for River is real.","tokens_in":15237,"tokens_out":8639,"would_cite":true,"duration_ms":64524,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91B12","91B14"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper introduces River, a resolute refinement of Split Cycle that satisfies independence of Pareto-dominated alternatives, a property it claims no other Split Cycle refinement satisfies.","keywords":["River method","Condorcet-consistent voting rules","Split Cycle","Ranked Pairs","Beat Path","Stable Voting","independence of Pareto-dominated alternatives","majority margins"],"falsifier":"Recompute, from Appendix B's profiles P1–P4, the winners of Split Cycle, Ranked Pairs, Beat Path, and Stable Voting with and without the Pareto-dominated alternative $a$. The paper's comparative claim is settled against it if any of the claimed changes fails to reproduce—for instance if Ranked Pairs on P2 returns $b$ both with and without $a$ rather than $d$ turning into $b$.","tokens_in":14219,"feed_emoji":"🗳️","tokens_out":9126,"duration_ms":81284,"temperature":0.7,"pith_summary":"River is a single-winner voting method that starts from the pairwise majority margins of a preference profile and builds a directed tree by scanning edges from strongest to weakest, adding an edge only if it would create neither a cycle nor a second incoming edge to any alternative. The unique source of that tree is the winner. The paper shows that River is a refinement of Split Cycle, so every River winner is immune to majority complaints, and that River satisfies the standard basic axioms including monotonicity and the Condorcet winner criterion. Its main new claim is that River satisfies independence of Pareto-dominated alternatives (IPDA): removing an alternative that every voter ranks below another alternative never changes the winner. The paper further claims that no other Split Cycle refinement—Ranked Pairs, Beat Path, or Stable Voting—satisfies IPDA, and that a quasi-Pareto version of the property also holds under suitable tiebreakers.","feed_headline":"River voting rule blocks Pareto-dominated spoilers","feed_subtitle":"Tree-based River is the only Split Cycle refinement that satisfies independence of Pareto-dominated alternatives.","key_machinery":"The River diagram $M_{RV}$ is the object that carries the argument. It is built by ordering all majority edges by decreasing margin and adding an edge only when it satisfies both constraints: (Cy) it does not create a cycle, and (Br) it is not a second incoming edge to any alternative. Because of (Br) the final graph is a spanning tree with exactly one source, and that source is the River winner. The same diagram is the certificate for immunity: for any edge into the winner, the tree contains a unique path from the winner back to the source of that edge with strength at least as high. The IPDA proof additionally uses the covering observation that Pareto domination implies $m(y,z) \\ge m(x,z)$ for every $z$ other than $x,y$, with strict inequality in uniquely weighted profiles, which is what forces the dominated alternative's outgoing edges to be rejected.","core_discovery":"The central discovery, stated in the paper's own terms, is that River is a resolute refinement of Split Cycle that satisfies IPDA, a property the other three refinements do not have. The branching rule is the mechanism: if $y$ Pareto-dominates $x$, the edge $(y,x)$ has the maximum possible margin and is added first, while any edge out of $x$, say $(x,z)$, is processed only after the stronger edge $(y,z)$ and is therefore rejected either by branching or by the cycle it would close through $y$ and $x$. Thus $x$ appears in the River tree as a leaf attached to $y$ and contributes nothing else, so deleting $x$ leaves every other edge intact and the same source wins. The proofs cover uniquely weighted profiles, general profiles with Pareto-consistent tiebreakers, and the generalization to quasi-Pareto-dominated alternatives in Theorems 4.6, 4.8, and 4.10.","pith_inferences":["Editorial: If the comparative IPDA claim survives re-checking, River combines resolute output, simple hand computation, and a spoiler-independence guarantee, which could make it attractive for participatory or constrained settings where transparency matters.","Editorial: The tree certificate's uniqueness suggests a natural measure of how far a winner is from each rival, namely the strength of the single rebutting path, which the paper does not develop; this could support explanations of outcomes to voters.","Editorial: The quasi-Pareto generalization indicates the construction is not tied to full unanimity dominance; any dominance relation that guarantees earlier processing of the dominator's outgoing edges would yield a similar independence property.","Editorial: A natural empirical test is to sample random preference profiles, add a Pareto-dominated alternative, and measure how often Ranked Pairs, Beat Path, and Stable Voting actually change winner; the paper establishes possibility, not frequency."],"forward_implications":["River's winner is always an immune alternative, so River is Condorcet-consistent and never leaves the Smith set.","River satisfies ISDA: deleting a Smith-dominated alternative never changes the winner, for all preference profiles.","River satisfies IPDA for uniquely weighted profiles, and for general profiles when equipped with any Pareto-consistent tiebreaker; it also satisfies the stronger IQDA with quasi-Pareto-consistent tiebreakers.","The River diagram has exactly one fewer edge than alternatives, so the winner's rebuttal certificate is a single tree path to every rival, viewable by hand.","Ranked Pairs, Beat Path, Stable Voting, and Split Cycle all fail IPDA, according to the paper's counterexamples, so among Split Cycle refinements River is uniquely positioned against this form of agenda manipulation."],"supporting_citations":[{"why":"defines Split Cycle and the immunity criterion, the baseline that River refines","marker":"[17]"},{"why":"introduces Ranked Pairs, the method River is presented as a simplified variation of","marker":"[18]"},{"why":"introduces Beat Path, one of the Split Cycle refinements compared against River","marker":"[19]"},{"why":"introduces Stable Voting, another compared refinement and the source of the monotonicity contrast","marker":"[20]"},{"why":"introduces the independence-of-Pareto-dominated-alternatives condition under the name reduction condition","marker":"[21]"},{"why":"introduces the immunity concept used to define Split Cycle winners","marker":"[15]"},{"why":"records the original 2004 description of the River construction","marker":"[29]"},{"why":"provides the lemma that margin graphs with even margins can be realized by preference profiles, used for the examples","marker":"[32]"}],"fun_headline_variants":["River voting rule vetoes Pareto-dominated candidates","River tree kills Pareto spoilers, unlike Ranked Pairs","Only Split Cycle refinement blocking Pareto-dominated alternatives","River: simple Condorcet rule that blocks Pareto spoilers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"River's own IPDA proof is internal to the tree construction, but the paper's comparative claim that no other Split Cycle refinement satisfies IPDA depends on the four Appendix B profiles having exactly the margin graphs and winner changes stated in Theorem 4.4; a direct check of the Ranked Pairs example suggests one of those changes may not reproduce.","fun_headline_variants_meta":{"raw":{"variants":["River voting rule vetoes Pareto-dominated candidates","River tree kills Pareto spoilers, unlike Ranked Pairs","Only Split Cycle refinement blocking Pareto-dominated alternatives","River: simple Condorcet rule that blocks Pareto spoilers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000472,"raw_usage":{"total_tokens":2287,"prompt_tokens":829,"completion_tokens":1458,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":445,"completion_tokens_details":{"reasoning_tokens":1395}},"tokens_in":445,"tokens_out":1458,"duration_ms":10562,"temperature":1.0,"reasoning_tokens":1395,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T11:56:41.144303+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute, from Appendix B's profiles P1–P4, the winners of Split Cycle, Ranked Pairs, Beat Path, and Stable Voting with and without the Pareto-dominated alternative $a$. The paper's comparative claim is settled against it if any of the claimed changes fails to reproduce—for instance if Ranked Pairs on P2 returns $b$ both with and without $a$ rather than $d$ turning into $b$.","supporting_citations":[{"cited_title":"Public Choice 197, 1–62 (2023)","cited_arxiv_id":null,"evidence_quote":"defines Split Cycle and the immunity criterion, the baseline that River refines"},{"cited_title":"Social Choice and Welfare 4, 185–206 (1987)","cited_arxiv_id":null,"evidence_quote":"introduces Ranked Pairs, the method River is presented as a simplified variation of"},{"cited_title":"Social choice and Welfare 36, 267–303 (2011)","cited_arxiv_id":null,"evidence_quote":"introduces Beat Path, one of the Split Cycle refinements compared against River"},{"cited_title":"Constitutional Political Economy, 421–433 (2023)","cited_arxiv_id":null,"evidence_quote":"introduces Stable Voting, another compared refinement and the source of the monotonicity contrast"},{"cited_title":"Princeton University Press, Princeton (1973)","cited_arxiv_id":null,"evidence_quote":"introduces the independence-of-Pareto-dominated-alternatives condition under the name reduction condition"},{"cited_title":"Social Choice Under Incomplete, Cyclic Preferences","cited_arxiv_id":"math/0201285","evidence_quote":"introduces the immunity concept used to define Split Cycle winners"},{"cited_title":"http://lists.electorama.com/pipermail/ election-methods-electorama.com/2004-October/014018.html","cited_arxiv_id":null,"evidence_quote":"records the original 2004 description of the River construction"},{"cited_title":"Math´ ematiques et sciences humaines97, 5–17 (1987) 12 Appendix The supplementary material consists of three parts","cited_arxiv_id":null,"evidence_quote":"provides the lemma that margin graphs with even margins can be realized by preference profiles, used for the examples"}],"review_version":1}