{"id":"0757e047-77e0-428a-b6f5-e6ee47ecd880","arxiv_id":"2501.13364","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"A customer-led Stackelberg game with group buying and company team formation is claimed to reduce payments and boost revenues in satellite constellation services, but the proofs and experiments have significant gaps.","lead":"This paper proposes a market model where customers, not companies, lead: customers team up to request satellite imaging tasks and companies form teams to supply them. The authors claim this cuts customer payments by 23% and raises company revenues several-fold, based on simulations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The ordinal-potential proof of Theorem 1 fails: a unilateral deviation can increase a company's payoff while leaving Φ unchanged, so pure-strategy NE existence is unproven.","rationale":"The reader's weakest assumption correctly identifies the invalid ordinal-potential step. My independent check of the model's definitions confirms the concern: Φ is a function of the set of tasks covered by at least one team, not of which teams win those tasks, so it cannot track the per-company payoff changes that an ordinal potential must track. The three-company counterexample is a legitimate instance under the paper's own Definition 1 and payoff definitions, and it gives ΔΦ=0 with Δr_y>0. This is internally inconsistent with the proof, not merely a difference from the existing literature. Because Theorem 1 is the foundation for Theorem 2 and for the uniqueness corollaries, the theoretical guarantee at the center of the paper is unproven. The experimental section does not rescue this: the simulations assume task allocation and team formation algorithms that are not shown to compute the claimed equilibria, and the reported 23%/6.7x headline numbers are not consistently stated across abstract, introduction, and tables. Since the reader already rejected the paper for essentially this reason, no verdict adjustment is needed.","tokens_in":23618,"tokens_out":6514,"duration_ms":705355,"concrete_test":"Run a brute-force verification of the ordinal-potential condition on small instances of the follower game: 3 companies, 1 task, discrete i.i.d. offer distributions, enumerating all strategy profiles and all unilateral deviations. The counterexample above (y1,y2,y3 all offering {a,b}; initial A_y1=A_y2={y1,y2}, A_y3={y3}; y1 deviating to {y1}) must produce Δr_y1>0 and ΔΦ=0. If this sign mismatch is reproduced, the claimed equivalence in Appendix B.1 is false and the proof of Theorem 1 collapses.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1 (Appendix B.1) claims that Φ(spf)=Σ_m E_m[spf] is an ordinal potential for the follower game, requiring Δr_y>0 ⇔ ΔΦ>0 for every unilateral deviation. This is false. Because Pr(m,k) sums to 1 over all teams that can serve task k, Φ equals the sum of payments of tasks that are covered by at least one team; it does not record how expected profit is split among teams. Under Definition 1, a team forms only when all its members announce exactly the same coalition, so a unilateral deviation can make a company leave a team without changing the set of covered tasks, while changing the competitive split. Concrete counterexample: Y={y1,y2,y3}, each with Sy={a,b}, one task k requiring {a,b}, payment 100, identical offer distributions. Initial spf: A_y1=A_y2={y1,y2}, A_y3={y3}, so teams {y1,y2} and {y3} both cover k and each wins with probability 1/2; y1's payoff is (100/2)/2=25 and Φ=100. If y1 deviates to A_y1={y1}, all three companies become singleton teams, each winning with probability 1/3; y1's payoff becomes 100/3>25, while Φ stays 100 because k is still covered by every team. Thus Δr_y1>0 but ΔΦ=0, violating the ordinal-potential equivalence. The hand-written 'Direction 1' and 'Direction 2' simply assert the missing monotonicity. Since Theorem 2's Stackelberg existence invokes Theorem 1, it inherits the gap; Corollary 1 additionally assumes strict monotonicity in a way that essentially posits a unique global maximizer rather than deriving uniqueness. The central theoretical guarantee is therefore unsupported as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a customer-led Stackelberg game model for task allocation in two-sided markets. Customers act as leaders and may use a group-buying discount to form tasks; companies act as followers and may form teams to cover the requested services. The main theoretical claims are that a pure-strategy Nash equilibrium exists in the follower game for any feasible task set (Theorem 1), that a pure-strategy Stackelberg equilibrium exists (Theorem 2), and that uniqueness holds under a 'strict monotonicity' condition. The paper also reports satellite-constellation simulation experiments claiming a 23% reduction in customer payments and a 6.7-fold increase in company revenues. The theoretical proof in Appendix B.1 is an ordinal-potential argument, and the experimental comparison is between a customer-led treatment with group-buying discounts and a company-led baseline without them.","tokens_in":23981,"tokens_out":10201,"duration_ms":95310,"significance":"The setting is relevant: customer-led mechanisms with team formation are natural for markets such as Earth-observation services, and the paper makes a concrete modeling attempt with a realistic simulation domain. If the existence theorems were correct, the paper would contribute a useful formal framework. However, the central theoretical result is not established: the proposed potential function is not ordinal, and the proof contains a concrete monotonicity failure. The uniqueness results are essentially tautological assumptions rather than derived conditions. The headline experimental numbers are also artifacts of the chosen discount function and of comparing treatments with and without group buying. The paper therefore does not currently provide the claimed theoretical guarantees or empirical evidence. I see no machine-checked proofs, reproducible code, or parameter-free derivations that would offset these gaps.","major_comments":[{"comment":"The proposed ordinal potential function is invalid. Since for each task k the sum of Pr(m,k) over all teams able to serve k is 1, the quantity Phi(spf)=sum_m E_m[spf] depends only on the set of tasks that at least one team can cover; it does not record how the expected profit is allocated among teams. Consequently, a unilateral deviation that changes the partition but preserves the covered-task set leaves Phi unchanged even when the deviating company's revenue changes. Concretely, take three companies y1,y2,y3 with Sy={a,b} for each, one task k with Rk={a,b} and payment 100, identical offer distributions, lambda=1, and zero costs. Under A_y1=A_y2={y1,y2}, A_y3={y3}, teams {y1,y2} and {y3} each win with probability 1/2, so y1's revenue is 25 and Phi=100. If y1 deviates to A_y1={y1}, all three companies are singletons, each wins with probability 1/3; y1's revenue becomes 100/3>25, while Phi remains 100. This violates the claimed equivalence Delta r_y>0 iff Delta Phi>0 in 'Direction 1' and 'Direction 2' of the proof. Thus the proof does not establish the existence of a pure Nash equilibrium, and the theorem is unsupported.","section":"Appendix B.1, Theorem 1"},{"comment":"The proof of Theorem 2 relies on the assumption that for every leader strategy spl there is a pure-strategy Nash equilibrium BR(spl) of the follower game. Since Theorem 1 is not established, this premise is unavailable. In addition, the argument that any mixed strategy of a customer assigning positive probability to a task-failing action is strictly dominated by a pure strategy that guarantees success is false: whether a task is covered depends on the companies' team formation, which a customer cannot control, and it is possible that every pure customer action leads to failure for some follower best response. Therefore the finite-normal-form argument does not establish a pure-strategy leader equilibrium. Corollary 2 inherits these problems.","section":"Appendix B.2, Theorem 2"},{"comment":"The 'strict monotonicity' assumptions restate the desired uniqueness conclusion rather than provide a sufficient condition that can be checked from the market primitives. In an ordinal potential game, pure Nash equilibria correspond to local maxima of the potential, and a unique global maximum does not rule out additional local maxima that are also equilibria. The proofs assert that a stable point must be the global maximizer, but that is exactly what needs to be proven. No example of a natural condition on delta, the customers' payments, or the social network that implies the assumed unique global maximum or minimum is given, so the uniqueness claims are vacuous.","section":"Appendix B.1, Corollary 1; Appendix B.2, Corollary 2"},{"comment":"The reported 23% payment reduction is built into the model by construction. Appendix C defines the discount factor as delta(d_k^s)=0.5 e^{-(|Y|/|X|)(d_k^s-1)}+0.4, with a maximum discount of 60%, so any task with d_k^s>1 receives a strictly lower payment than the independent baseline. Comparing 'Customer-led' (which applies this delta) with 'Company-led' (which does not) therefore guarantees a payment reduction; the experiment does not test whether the proposed equilibrium mechanism itself reduces payments. The claimed revenue increase is similarly confounded because the customer-led treatment additionally allows company team formation while the baseline does not.","section":"Appendix C and Section 5.2"},{"comment":"The paper labels rows in Tables 2-7 as 'at Nash equilibrium', but the experiments use a greedy team-formation algorithm and a similarity threshold rather than computing equilibria of the game defined by Definition 2. No verification of the no-unilateral-deviation condition is reported, so the tables do not provide empirical evidence for the theoretical equilibrium claims. In addition, the headline revenue multiplier is internally inconsistent: the abstract states a 6.7-fold increase, the introduction states 'increase company 14 times revenue', and Table 2 reports +900% (a 9-fold increase) for the baseline scenario. These numbers should be reconciled.","section":"Section 5.2 and Appendix E"}],"minor_comments":[{"comment":"The symbol TAM is used in Eq. (1) and in Definition 4 but is never defined in the main text; please define it (presumably the set of teams able to perform a task).","section":"Section 4.2, Eq. (1)"},{"comment":"The variable u_x is called 'utility' but customers are described as minimizing it; the text also uses 'payment' and 'cost' interchangeably. The sign convention should be made consistent, especially since Theorem 2 refers to customers 'lowering their cost'.","section":"Section 4.2, Eq. (1)"},{"comment":"The proof of Proposition 3 cites Lemma 2.9 of [Roughgarden and Tardos, 2002] as if it were a general statement about social welfare in potential games; that lemma concerns the price of anarchy in selfish routing and does not justify the claimed equality of total revenues across arbitrary strategy profiles.","section":"Appendix B.3, Proposition 3"},{"comment":"The algorithm calls the similarity measure 'Jacobian distance'; the formula is the Jaccard distance. Please correct the terminology.","section":"Appendix D.1, Algorithm 2"},{"comment":"The displayed formula for the discount factor uses the ambiguous notation 'e(-(...))' and describes |Y|/|X| as 'total service demands and the number of services offered', while |Y| and |X| are previously defined as the numbers of companies and customers. Please clarify the formula and the description.","section":"Appendix C"},{"comment":"The main-text Theorem 1 is labelled Theorem 3 in Appendix B.1, Theorem 2 is labelled Theorem 4 in Appendix B.2, and Propositions 1-3 are labelled Propositions 4-6. Renumber or provide a mapping so readers can cross-reference the statements.","section":"Appendix B"}],"recommendation":"reject","confidential_remarks":"The core existence proof contains a concrete counterexample to the claimed potential-function property, so the central theoretical contribution is unsupported. The uniqueness conditions are tautological, and the experimental comparison is confounded by construction. These are not local presentation issues; a substantially different proof strategy and a redesign of the experiments would be needed. I therefore recommend rejection rather than major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The short version: the customer-led Stackelberg setup is a sensible framing for two-sided markets, but the central theoretical claims do not hold up, and the experimental numbers are overstated. This needs substantial revision before it deserves to be cited.\n\nWhat is actually new: the combination of customer group buying with company team formation, both driven by similarity networks, is a fresh angle for market design in domains like satellite services. The Basilisk-based simulation is a concrete, reproducible piece of work, and the paper does a decent job of laying out the algorithms in the appendix.\n\nWhere it falls down: the proof of Theorem 1 is invalid. The proposed potential Φ is the sum of team expected profits. But for any task, the win probabilities of all teams that can cover it sum to 1, so Φ depends only on which tasks are covered, not on how the teams are split. A unilateral deviation can improve a company's payoff without changing coverage, leaving Φ unchanged. The stress-test counterexample captures this exactly: with three companies and one task, y1's payoff can rise from 25 to 33.3 while Φ stays 100. That kills the ordinal potential argument, and Theorem 2 inherits the gap. The uniqueness corollaries are circular: they simply assume a unique global maximizer/minimizer, which is essentially the conclusion.\n\nThe experiments compare the full model against a no-cooperation baseline where most tasks simply fail, so the revenue jumps are largely mechanical. The 23% payment reduction is built into the discount factor δ(d), which is forced below 1 for any group purchase. Also, the headline numbers disagree: the abstract says 6.7x, the introduction says 14x, the conclusion says 6.7x, and the tables show up to 1430%.\n\nThis is not a hopeless paper. The model is genuinely novel, and if the existence results can be proven by a different route, it could be a useful applied contribution. I would not cite the theory as it stands.\n\nRecommendation: send it to peer review, but with a referee who will actually check the potential-function proof. The flaws are specific and fixable in principle, and the model deserves a careful look.","headline":"Interesting model, broken proof: the main existence theorems are unsupported and the headline numbers are unreliable.","tokens_in":24540,"tokens_out":3630,"would_cite":false,"duration_ms":34365,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A customer-led Stackelberg game for two-sided markets is claimed to guarantee stable task allocation and to cut customer payments by 23% while boosting company revenue 6.7-fold.","keywords":["task allocation","two-sided markets","Stackelberg game","Nash equilibrium","satellite constellations","team formation","group buying","multi-agent systems"],"falsifier":"Take a market with two companies and two tasks, where each company can cover both tasks alone, so every team formation yields the same total expected profit $\\Phi$. Give company $y$ a unilateral deviation that moves it from its current team to a different team and strictly increases $y$'s own revenue while leaving $\\Phi$ unchanged. If such a configuration exists, then $\\Delta r_y > 0$ while $\\Delta \\Phi = 0$, disproving the ordinal-potential equivalence that Theorem 1 depends on.","tokens_in":23378,"feed_emoji":"🛰️","tokens_out":7286,"duration_ms":62742,"temperature":0.7,"pith_summary":"This paper argues that two-sided markets work better when the customer leads. It proposes a Stackelberg game in which customers initiate tasks—alone or in group-buying coalitions with similar needs—and companies respond by forming teams to serve those tasks at the lowest expected price. The paper claims two existence guarantees: a pure-strategy Nash equilibrium among companies for any feasible task set, and a Stackelberg equilibrium for the full game. A simulation of a satellite Earth-observation market, with 5,000–10,000 customers and up to 30 companies, shows customer payments falling by 23% and company revenue rising by 6.7-fold relative to a company-led baseline. The significance is a theoretical and practical template for markets where personalised customer demand, not company pricing, sets the agenda.","feed_headline":"Customer-led game model cuts payments 23%, lifts revenue 6.7x","feed_subtitle":"Reversing leader-follower roles in two-sided markets yields stable equilibria and better outcomes in satellite services.","key_machinery":"The load-bearing object is the potential function $\\Phi(\\mathrm{spf}) = \\sum_{m \\in M^{\\mathrm{spf}}_K} E_m[\\mathrm{spf}]$, the sum of expected profits over all teams formed by a company strategy profile. The paper uses this to argue that the follower game is an ordinal potential game—a game in which every profitable unilateral deviation raises a common potential value—which gives the finite improvement property and hence a pure Nash equilibrium. The leader game is then shown to have an equilibrium by finite best-response reasoning once the followers' equilibrium is fixed. In the experimental implementation, the operative mechanism is a similarity threshold on the company social network: a greedy algorithm forms teams from companies whose service offerings overlap below the threshold, and each task is allocated to the team with the minimum total offer price. The customer side uses a group-buying discount $\\delta(d^s_k)$ that decreases as more customers in a task share the same service need.","core_discovery":"The paper's central claim is that reversing the usual Stackelberg hierarchy—customers as leaders who post tasks, companies as followers who form teams—preserves the game-theoretic guarantees that market designers rely on. Theorem 1 asserts that for any given feasible set of tasks, the companies' follower game has at least one pure-strategy Nash equilibrium, and Corollary 1 gives uniqueness when the potential function has a unique global maximum. Theorem 2 asserts that the customer-led Stackelberg game has at least one pure-strategy Stackelberg equilibrium, with a uniqueness corollary under strict monotonicity. The mechanism combines customer group-buying discounts with company team formation based on service similarity, and allocates each task to the team with the lowest expected offer price. The experimental section reports that, across five scenarios of the satellite constellation market, the customer-led model cuts average customer payments by up to 23% and lifts company revenues by factors ranging from about 2.4 to 14, with the abstract citing a 6.7-fold average increase.","pith_inferences":["A concrete way to stress-test Theorem 1 is to search small market instances for a unilateral deviation that improves a company's revenue while leaving the total team-profit potential unchanged; if found, the existence proof would need a different argument even if the equilibrium itself still exists.","The similarity threshold for team formation acts as a market-design lever: the experiments locate the point of near-100% task completion at different thresholds per scenario, suggesting an operator could tune this threshold to the market's density rather than fixing it.","The model assumes a static set of feasible tasks chosen by customers; extending it to dynamic entry and exit of customers would require re-checking the equilibrium conditions, since the follower best response could shift as tasks change.","The reported revenue multipliers are computed against a company-led baseline in which no team formation occurs; a fair comparison with a company-led model that also allows teams would isolate the effect of leadership reversal from the effect of allowing cooperation."],"forward_implications":["If Theorem 1 and Theorem 2 hold, platforms in crowdsourcing, ride-sharing, and cloud computing can adopt customer-initiated design while retaining stability guarantees.","The 23% customer payment reduction indicates that group-buying among customers with similar service needs can substantially lower costs even when individual preferences are heterogeneous.","The revenue gains reported across all scenarios suggest that company team formation unlocks high-value tasks that a single provider cannot serve, which is a direct economic argument for encouraging partnerships.","The uniqueness corollaries imply that, under strict monotonicity, the market outcome is predictable, which would let platform operators compute the equilibrium in advance rather than rely on tatonnement."],"supporting_citations":[{"why":"Provides the characterization of ordinal potential games used to argue the follower game has a pure Nash equilibrium.","marker":"[V oorneveld and Norde, 1997]"},{"why":"Supplies the ordinal potential framework for the finite improvement property argument.","marker":"[Ewerhart, 2020]"},{"why":"Defines the Stackelberg game hierarchy that the customer-led model inverts.","marker":"[Stackelberg and Peacock, 1952]"},{"why":"Provides the extreme value theory used to model the probability that a team wins a task by having the minimum offer price.","marker":"[Haan and Ferreira, 2006]"},{"why":"Lemma 2.9 is used in Proposition 3 to show that all Nash strategy profiles for a given task set yield the same total company revenue.","marker":"[Roughgarden and Tardos, 2002]"},{"why":"Supplies the satellite simulation platform used to generate the experimental scenarios.","marker":"[Kenneally et al., 2020]"}],"fun_headline_variants":["Flip the market: customer-led Stackelberg cuts costs 23%, lifts revenue 6.7x","When customers lead, firms win: 6.7x revenue, 23% lower costs","Customer-led Stackelberg equilibrium: 23% savings, 6.7x revenue boost","In satellite markets, flipping the order cuts payments 23%, boosts revenue 6.7x","Reversed game roles: customers lead, firms gain 6.7x, costs drop 23%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the follower game has a pure Nash equilibrium rests on the unproved assertion that whenever a company can improve its own revenue by switching teams, the total expected profit of all teams changes in the same direction; this equivalence fails in general because the total can be identical for different team formations that serve the same set of tasks.","fun_headline_variants_meta":{"raw":{"variants":["Flip the market: customer-led Stackelberg cuts costs 23%, lifts revenue 6.7x","When customers lead, firms win: 6.7x revenue, 23% lower costs","Customer-led Stackelberg equilibrium: 23% savings, 6.7x revenue boost","In satellite markets, flipping the order cuts payments 23%, boosts revenue 6.7x","Reversed game roles: customers lead, firms gain 6.7x, costs drop 23%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000356,"raw_usage":{"total_tokens":1913,"prompt_tokens":909,"completion_tokens":1004,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":525,"completion_tokens_details":{"reasoning_tokens":879}},"tokens_in":525,"tokens_out":1004,"duration_ms":10960,"temperature":1.0,"reasoning_tokens":879,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T16:01:37.195913+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a market with two companies and two tasks, where each company can cover both tasks alone, so every team formation yields the same total expected profit $\\Phi$. Give company $y$ a unilateral deviation that moves it from its current team to a different team and strictly increases $y$'s own revenue while leaving $\\Phi$ unchanged. If such a configuration exists, then $\\Delta r_y > 0$ while $\\Delta \\Phi = 0$, disproving the ordinal-potential equivalence that Theorem 1 depends on.","supporting_citations":[{"cited_title":"A characterization of ordinal potential games","cited_arxiv_id":null,"evidence_quote":"Provides the characterization of ordinal potential games used to argue the follower game has a pure Nash equilibrium."},{"cited_title":"Ordinal potentials in smooth games","cited_arxiv_id":null,"evidence_quote":"Supplies the ordinal potential framework for the finite improvement property argument."},{"cited_title":"The theory of the market economy","cited_arxiv_id":null,"evidence_quote":"Defines the Stackelberg game hierarchy that the customer-led model inverts."},{"cited_title":"Extreme value theory: an introduction , volume","cited_arxiv_id":null,"evidence_quote":"Provides the extreme value theory used to model the probability that a team wins a task by having the minimum offer price."},{"cited_title":"How bad is selfish routing? Journal of the ACM (JACM), 49(2):236–259,","cited_arxiv_id":null,"evidence_quote":"Lemma 2.9 is used in Proposition 3 to show that all Nash strategy profiles for a given task set yield the same total company revenue."},{"cited_title":"Basilisk: A flexible, scalable and modular astrodynamics simulation framework","cited_arxiv_id":null,"evidence_quote":"Supplies the satellite simulation platform used to generate the experimental scenarios."}],"review_version":1}