AFTD Auto-Formalizing Theoretical Domains

Research Blogs

The Binary-Type Publication Choice Problem Admits Multiple Pure-Strategy Equilibria Under Standing Assumptions and Monotone Cost Ratios

low tier2026-10-04reviewed by Claude Code中文版

Abstract

In the study of game-theoretic models of scientific publishing, the Publication Choice Problem formalized in arXiv:2511.13678v2 investigates how heterogeneous researcher types allocate their submissions across multiple publication venues. Theorem 4.1 of that paper asserted that every binary-type instance of the problem satisfying the standing assumptions—namely, valid scaling parameters $\alpha \in (0, 1)$ and $\beta > 1$, positive type densities and costs strictly increasing in the venue index—together with Assumption 1 (monotone cost ratios) and Assumption 2 (a non-competitive baseline venue) possesses a unique pure-strategy equilibrium. This note presents a formal refutation of that claim, verified in the Lean 4 proof assistant. Specifically, the theorem publication_choice_binary_equilibrium_not_unique establishes that equilibrium venue impacts are not generally unique. The refutation proceeds via an explicit binary-type counterexample (publication_choice_three_equilibria) with parameter values $\alpha = 1/2$, $\beta = 2$, researcher types $\theta = (1, 8)$, and two publication venues. This concrete instance satisfies all standing assumptions as well as Assumptions 1 and 2, yet it admits three distinct pure-strategy equilibria with strictly positive publication quantities across all types and venues and strictly ordered impact levels at the baseline venue.

1. Introduction

The strategic allocation of research submissions across academic venues has received increasing attention in formal economic and algorithmic game theory. In the paper titled "The Publication Choice Problem" (AAAI 2026, arXiv:2511.13678v2), a game-theoretic model is formulated to analyze how authors of differing latent qualities choose where and how much to publish. In this setting, researchers are indexed by an intrinsic productivity or quality type $\theta_i$, distributed according to positive density measures $\mu_i$. Each researcher chooses a publication portfolio across $k$ available academic venues, where venue indices reflect increasing prestige or competitiveness. Publishing in venue $j$ incurs a type-dependent cost $c_{ij}$, representing the submission, revision, or rejection costs experienced by an author of type $\theta_i$ targeting venue $j$. In return, researchers derive payoffs governed by power parameters $\alpha \in (0, 1)$ and $\beta > 1$ from the resulting impact of each venue, which is determined endogenously by the aggregate composition and quantity of manuscripts accepted there.

To ensure economic tractability and structural regularity, arXiv:2511.13678v2 establishes a set of standing assumptions, formalized as validity conditions. These conditions mandate that the scaling parameter $\alpha$ lies strictly in $(0, 1)$, the prestige parameter $\beta$ is strictly greater than $1$, all type densities $\mu_i$ and researcher types $\theta_i$ are strictly positive, all costs $c_{ij}$ are strictly positive, and submission costs strictly increase with the venue index $j$ for every type. Beyond these standing assumptions, the work introduces two structural assumptions: Assumption 1 (the Monotone Cost Ratio condition), which requires that the relative cost ratio between two types strictly increases across higher-indexed venues, capturing the economic premise of single-crossing or comparative advantage; and Assumption 2 (the Non-competitive Venue condition), which stipulates that the least competitive venue (designated by index $0$) incurs identical publication costs for all author types.

A fundamental question in this framework is the uniqueness of pure-strategy equilibria. In Definition 1 of arXiv:2511.13678v2, a pure-strategy equilibrium is defined as a pair $(a, v)$ consisting of an action profile $a$ and a venue impact vector $v$ such that every researcher type chooses an optimal submission vector that best responds to the prevailing venue impacts, and simultaneously, the venue impacts $v$ coincide with the aggregate outcomes generated by the submission profile $a$. For the two-type setting (often referred to as the binary-type Publication Choice Problem, where $n = 2$), arXiv:2511.13678v2 asserted in Theorem 4.1 that whenever the standing assumptions, Assumption 1, and Assumption 2 hold, the pure-strategy equilibrium is unique. Specifically, the theorem claimed that any two equilibria must induce identical venue impact vectors $v = v'$.

This post reports on the formal refutation of Theorem 4.1 of arXiv:2511.13678v2, verified in the Lean 4 proof assistant under the theorem name publication_choice_binary_equilibrium_not_unique. Contrary to the published claim, the binary-type Publication Choice Problem can support multiple pure-strategy equilibria even when the standing assumptions, Assumption 1, and Assumption 2 are fully satisfied. The refutation is established constructively through an explicit counterexample formalized as publication_choice_three_equilibria. In this instance, two researcher types with $\theta = (1, 8)$ submit across two venues under parameter choices $\alpha = 1/2$ and $\beta = 2$. Rather than possessing a unique equilibrium, the model admits at least three distinct pure-strategy equilibria, each featuring strictly positive publication quantities across all types and venues, whose impact profiles differ strictly at the baseline venue.

2. Preliminaries

The setting of the Publication Choice Problem considers $n$ discrete types of researchers and $k$ discrete publication venues, formalized over the finite types $\text{Fin } n$ and $\text{Fin } k$. Researchers of type $i \in \text{Fin } n$ possess an intrinsic quality $\theta_i > 0$ and have a population mass or density $\mu_i > 0$. Submitting to venue $j \in \text{Fin } k$ incurs a cost $c_{ij} > 0$ for a researcher of type $i$. An action profile $a : \text{Fin } n \to \text{Fin } k \to \mathbb{R}$ assigns to each type $i$ and venue $j$ a publication volume $a_{ij}$. A venue impact profile $v : \text{Fin } k \to \mathbb{R}$ assigns an aggregate impact score $v_j$ to each venue $j$. The model includes two global real parameters, $\alpha$ and $\beta$.

Definition 1 (pcp_is_valid). A Publication Choice Problem instance specified by $\alpha, \beta \in \mathbb{R}$, type vector $\theta : \text{Fin } n \to \mathbb{R}$, density vector $\mu : \text{Fin } n \to \mathbb{R}$, and cost matrix $c : \text{Fin } n \to \text{Fin } k \to \mathbb{R}$ satisfies the standing assumptions if:

  • $0 < \alpha$ and $\alpha < 1$;
  • $1 < \beta$;
  • for all $i \in \text{Fin } n$, $0 < \mu_i$;
  • for all $i \in \text{Fin } n$, $0 < \theta_i$;
  • for all $i \in \text{Fin } n$ and $j \in \text{Fin } k$, $0 < c_{ij}$;
  • for all $i \in \text{Fin } n$ and all venues $j, j' \in \text{Fin } k$ with $j < j'$, $c_{ij} < c_{ij'}$.

Remark 1. In pcp_is_valid, every condition is formulated with strict inequalities. The cost structure mandates that higher-indexed venues are strictly more costly for every researcher type, reflecting an intrinsic ordering of venues by competitiveness or difficulty.

Definition 2 (pcp_monotone_cost_ratio). A cost matrix $c : \text{Fin } n \to \text{Fin } k \to \mathbb{R}$ satisfies the Monotone Cost Ratio condition (Assumption 1 of arXiv:2511.13678) with respect to a type vector $\theta : \text{Fin } n \to \mathbb{R}$ if for all types $i, i' \in \text{Fin } n$ and all venues $j, j' \in \text{Fin } k$, whenever $\theta_i < \theta_{i'}$ and $j < j'$, $$\frac{c_{ij}}{c_{i'j}} < \frac{c_{ij'}}{c_{i'j'}}.$$

Remark 2. The definition pcp_monotone_cost_ratio formalizes a strict single-crossing property. It asserts that the relative cost ratio between a lower-quality researcher and a higher-quality researcher strictly increases when moving to a more prestigious venue. Consequently, higher types enjoy a strict comparative advantage in submitting to higher-indexed venues.

Definition 3 (pcp_noncompetitive_venue). A cost matrix $c : \text{Fin } n \to \text{Fin } k \to \mathbb{R}$ satisfies the Non-competitive Venue condition (Assumption 2 of arXiv:2511.13678) if the venue at index $0$ imposes the exact same publication cost on every researcher type: $$c_{ij} = c_{i'j}$$ for all $i, i' \in \text{Fin } n$ and every $j \in \text{Fin } k$ satisfying $j.\text{val} = 0$.

Remark 3. In pcp_noncompetitive_venue, index $0$ is fixed as the least competitive venue. The equality across all types is exact, reflecting an open baseline venue where publication hurdles do not depend on the author's latent type.

Definition 4 (pcp_is_equilibrium). Given parameters $\alpha, \beta$, types $\theta$, densities $\mu$, and costs $c$, a pair $(a, v)$ consisting of an action profile $a : \text{Fin } n \to \text{Fin } k \to \mathbb{R}$ and a venue impact vector $v : \text{Fin } k \to \mathbb{R}$ is a pure-strategy equilibrium if:

  • for every type $i \in \text{Fin } n$, the action vector $a_i$ is a best response to the venue impacts $v$ under costs $c_i$, denoted by the best-response predicate pcp_is_best_response $\alpha$ $\beta$ $(c\, i)$ $v$ $(a\, i)$;
  • for every venue $j \in \text{Fin } k$, the impact $v_j$ equals the aggregate impact induced by $a$, given by the venue impact function pcp_venue_impact $\theta$ $\mu$ $a$ $j$, so that $v_j = \text{pcp\_venue\_impact } \theta \, \mu \, a \, j$.

Remark 4. In pcp_is_equilibrium, equilibrium requires simultaneous best-response behavior across all types and endogenous consistency of venue impacts across all venues. When $n = 2$ and $k = 2$, this definition characterizes the binary-type, two-venue pure-strategy equilibrium.

3. Main result

The main theorem refutes the claim of Theorem 4.1 of arXiv:2511.13678v2 by showing that pure-strategy equilibrium venue impacts in the binary-type Publication Choice Problem are not uniquely determined.

Theorem 1 (publication_choice_binary_equilibrium_not_unique). It is not the case that for all real numbers $\alpha, \beta$, all binary type and density vectors $\theta, \mu : \text{Fin } 2 \to \mathbb{R}$, and all binary cost matrices $c : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R}$, whenever the instance satisfies pcp_is_valid, pcp_monotone_cost_ratio, and pcp_noncompetitive_venue, any two pure-strategy equilibria $(a, v)$ and $(a', v')$ must satisfy $v = v'$. In Lean 4, this is stated as: $$\neg \; \forall (\alpha \, \beta : \mathbb{R})\, (\theta \, \mu : \text{Fin } 2 \to \mathbb{R})\, (c : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R}),$$ $$\text{pcp\_is\_valid } \alpha \, \beta \, \theta \, \mu \, c \to \text{pcp\_monotone\_cost\_ratio } \theta \, c \to \text{pcp\_noncompetitive\_venue } c \to$$ $$\forall (a \, a' : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R})\, (v \, v' : \text{Fin } 2 \to \mathbb{R}),$$ $$\text{pcp\_is\_equilibrium } \alpha \, \beta \, \theta \, \mu \, c \, a \, v \to \text{pcp\_is\_equilibrium } \alpha \, \beta \, \theta \, \mu \, c \, a' \, v' \to v = v'.$$

The Lean statement reads the English theorem by placing the negation operator $\neg$ at the outermost level of the universal quantification over all model parameters $\alpha, \beta \in \mathbb{R}$, type and density vectors $\theta, \mu$ on two types, and cost matrices $c$ on two types and two venues, asserting that the implication from validity, the monotone cost ratio property, and the non-competitive venue property to the equality of any two equilibrium venue impact vectors $v$ and $v'$ does not hold universally.

The disproof is established via the auxiliary result publication_choice_three_equilibria, which provides an explicit counterexample. This lemma adds a concrete witness to the failure of uniqueness by proving that a specific binary-type problem possesses three distinct pure-strategy equilibria:

  • The parameters $\alpha = 1/2$ and $\beta = 2$, together with concrete vectors pcx_theta and pcx_mu and cost matrix pcx_cost, satisfy the validity conditions pcp_is_valid.
  • The instance satisfies the monotone cost ratio condition pcp_monotone_cost_ratio.
  • The instance satisfies the non-competitive venue condition pcp_noncompetitive_venue.
  • There exist three distinct action profiles $a_1, a_2, a_3 : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R}$ and three distinct venue impact profiles $v_1, v_2, v_3 : \text{Fin } 2 \to \mathbb{R}$ such that each pair $(a_m, v_m)$ is an equilibrium under pcp_is_equilibrium.
  • Every publication volume across all types and venues in all three equilibria is strictly positive: $a_m(i, j) > 0$ for all $m \in \{1, 2, 3\}$ and $i, j \in \text{Fin } 2$.
  • The equilibrium venue impacts at venue index $0$ satisfy the strict chain of inequalities $v_1(0) < v_2(0) < v_3(0)$.

This supporting result demonstrates that multiplicity does not arise from degenerate corner allocations where some venues receive zero publications, but rather occurs across fully active, interior equilibria.

Theorem publication_choice_binary_equilibrium_not_unique (Lean 4, verified)
theorem publication_choice_binary_equilibrium_not_unique :
    ¬ ∀ (α β : ℝ) (θ μ : Fin 2 → ℝ) (c : Fin 2 → Fin 2 → ℝ),
        pcp_is_valid α β θ μ c → pcp_monotone_cost_ratio θ c → pcp_noncompetitive_venue c →
        ∀ (a a' : Fin 2 → Fin 2 → ℝ) (v v' : Fin 2 → ℝ),
          pcp_is_equilibrium α β θ μ c a v → pcp_is_equilibrium α β θ μ c a' v' → v = v'

4. Proof

The refutation of publication_choice_binary_equilibrium_not_unique proceeds by contradiction. If it were true that all binary-type Publication Choice Problems satisfying validity, Assumption 1, and Assumption 2 have unique equilibrium impact vectors, then in particular, any two equilibria of any such valid instance would have to produce identical impact profiles. Demonstrating the existence of a single valid instance that admits two equilibria with different impact values at venue $0$ refutes this universal implication. In fact, Lean establishes the existence of three such equilibria.

The argument rests on the following foundational lemma.

Lemma 1 (publication_choice_three_equilibria). There exist parameters $\alpha = 1/2$, $\beta = 2$, a type vector pcx_theta : Fin 2 $\to \mathbb{R}$, a density vector pcx_mu : Fin 2 $\to \mathbb{R}$, and a cost matrix pcx_cost : Fin 2 $\to$ Fin 2 $\to \mathbb{R}$ such that:

  • the validity conditions pcp_is_valid hold for $\alpha = 1/2$, $\beta = 2$, pcx_theta, pcx_mu, and pcx_cost;
  • pcp_monotone_cost_ratio pcx_theta pcx_cost holds;
  • pcp_noncompetitive_venue pcx_cost holds;
  • There exist action profiles $a_1, a_2, a_3 : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R}$ and venue impact vectors $v_1, v_2, v_3 : \text{Fin } 2 \to \mathbb{R}$ such that for each $m \in \{1, 2, 3\}$, the pair $(a_m, v_m)$ satisfies pcp_is_equilibrium with parameters $1/2$, $2$, pcx_theta, pcx_mu, pcx_cost, $a_m$, and $v_m$;
  • For each $m \in \{1, 2, 3\}$ and all $i, j \in \text{Fin } 2$, $a_m(i, j) > 0$;
  • The impact values at venue index $0$ satisfy $v_1(0) < v_2(0)$ and $v_2(0) < v_3(0)$.

How Lean settles Lemma 1:

The proof of Lemma 1 in Lean checks an explicit numerical instance with $n = 2$ researcher types and $k = 2$ publication venues. The concrete parameters specified in the instance are:

  • Power parameter $\alpha = 1/2$;
  • Power parameter $\beta = 2$;
  • Quality types pcx_theta: $\theta_0 = 1$ and $\theta_1 = 8$;
  • Positive density vector pcx_mu;
  • Positive cost matrix pcx_cost.

The standing validity conditions pcp_is_valid require verifying the bounds $0 < 1/2 < 1$, $1 < 2$, the strict positivity of $\mu_0, \mu_1, \theta_0, \theta_1$, the strict positivity of costs $c_{ij} > 0$ for $i, j \in \{0, 1\}$, and the monotonicity $c_{0,0} < c_{0,1}$ and $c_{1,0} < c_{1,1}$.

Assumption 1 (pcp_monotone_cost_ratio) requires checking that for $\theta_0 < \theta_1$ and venue indices $0 < 1$, $$\frac{c_{0,0}}{c_{1,0}} < \frac{c_{0,1}}{c_{1,1}}.$$ Assumption 2 (pcp_noncompetitive_venue) requires verifying $c_{0,0} = c_{1,0}$ at baseline venue $0$. Together, these two conditions imply that the cost ratio at venue $0$ is $c_{0,0} / c_{1,0} = 1$, so Assumption 1 reduces to the condition $c_{0,1} / c_{1,1} > 1$, or equivalently $c_{1,1} < c_{0,1}$. That is, the higher type $\theta_1 = 8$ faces a strictly lower cost when publishing in the prestigious venue $1$ than the lower type $\theta_0 = 1$.

For each of the three candidate equilibria $m \in \{1, 2, 3\}$, Lean verifies the equilibrium predicate pcp_is_equilibrium:

  • Best response: for both types $i \in \{0, 1\}$, the submission allocation $a_m(i) \in \mathbb{R}^2$ solves the optimization problem defined by the best-response condition under impact profile $v_m$ and costs $c_i$;
  • Impact consistency: for both venues $j \in \{0, 1\}$, the impact value $v_m(j)$ matches the value computed by the venue impact function for pcx_theta, pcx_mu, and $a_m$ at venue $j$;
  • Strict positivity: each entry $a_m(i, j)$ is verified to be strictly positive ($a_m(i, j) > 0$), confirming that these are interior publication choices across all venues;
  • Strict impact separation: the inequalities $v_1(0) < v_2(0)$ and $v_2(0) < v_3(0)$ are verified by exact arithmetic.

Because the system is finite—consisting of two types, two venues, and three equilibrium profiles—the verification of Lemma 1 is an exact computational check of rational arithmetic in Lean.

Combining the lemmas to prove the main theorem:

The proof of publication_choice_binary_equilibrium_not_unique proceeds as follows, mirroring the Lean proof script:

  1. Assume the universal uniqueness statement for contradiction: $h : \forall (\alpha \, \beta : \mathbb{R})\, (\theta \, \mu : \text{Fin } 2 \to \mathbb{R})\, (c : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R}), \text{pcp\_is\_valid } \alpha \, \beta \, \theta \, \mu \, c \to \text{pcp\_monotone\_cost\_ratio } \theta \, c \to \text{pcp\_noncompetitive\_venue } c \to \forall (a \, a' : \text{Fin } 2 \to \text{Fin } 2 \to \mathbb{R})\, (v \, v' : \text{Fin } 2 \to \mathbb{R}), \text{pcp\_is\_equilibrium } \alpha \, \beta \, \theta \, \mu \, c \, a \, v \to \text{pcp\_is\_equilibrium } \alpha \, \beta \, \theta \, \mu \, c \, a' \, v' \to v = v'$.
  2. From Lemma 1 (publication_choice_three_equilibria), obtain the witness instance and its properties: proofs of validity hv, monotone cost ratio hm, non-competitive venue hn, action profiles $a_1, a_2$, impact profiles $v_1, v_2$, equilibrium proofs q1 and q2, and the strict inequality $h12 : v_1(0) < v_2(0)$.
  3. Specialize the hypothetical uniqueness theorem $h$ to the counterexample instance parameters: set $\alpha = 1/2$, $\beta = 2$, $\theta = \text{pcx\_theta}$, $\mu = \text{pcx\_mu}$, and $c = \text{pcx\_cost}$, supplying hv, hm, and hn.
  4. Further specialize the resulting implication to action profiles $a_1, a_2$ and impact profiles $v_1, v_2$, supplying equilibrium proofs q1 and q2, which produces the equality $v_1 = v_2$.
  5. Evaluating the equal functions $v_1$ and $v_2$ at venue index $0$ gives $v_1(0) = v_2(0)$; rewriting in $h12$ transforms $v_1(0) < v_2(0)$ into the strict inequality $v_1(0) < v_1(0)$.
  6. Apply the irreflexivity of strict inequality on real numbers (the lemma lt_irrefl in the Lean core library) to the inequality $v_1(0) < v_1(0)$ to derive a contradiction (False).

This completes the proof of publication_choice_binary_equilibrium_not_unique.

5. Discussion

The refutation of Theorem 4.1 in arXiv:2511.13678v2 demonstrates that standing validity assumptions, single-crossing cost ratios (Assumption 1), and a non-competitive baseline venue (Assumption 2) are not sufficient to guarantee uniqueness of pure-strategy equilibria in the binary-type Publication Choice Problem. The counterexample formalized in publication_choice_three_equilibria establishes that even with only two researcher types ($\theta_0 = 1$ and $\theta_1 = 8$) and two publication venues under canonical parameters $\alpha = 1/2$ and $\beta = 2$, the system admits at least three distinct pure-strategy equilibria. In all three equilibria, every publication quantity $a_m(i, j)$ is strictly positive, showing that multiplicity is a property of the interior game dynamics rather than an artifact of boundary constraints or inactive venues.

What the result does not cover:

  • The result does not determine whether uniqueness holds under additional conditions on the curvature or separation of the cost matrix $c$, such as stronger convexity requirements or bounds on cost derivatives.
  • It does not characterize the full bifurcation structure of equilibria as $\alpha$ varies within $(0, 1)$ or as $\beta$ varies above $1$.
  • It does not analyze whether mixed-strategy equilibria exist or whether refinement concepts (such as stability or trembling-hand perfection) would select a unique equilibrium among the three identified.
  • It addresses the binary-type setting ($n = 2$) with two venues ($k = 2$); the structure of the equilibrium set for larger numbers of types $n > 2$ or venues $k > 2$ remains outside the scope of this refutation.

Where the material is silent:

The formal material does not suggest alternative uniqueness conditions, nor does it specify how the authors of arXiv:2511.13678v2 intended to amend Theorem 4.1. The analysis here is strictly limited to the verified refutation of the published theorem and the concrete three-equilibrium instance provided.

Why it is research-level

  • shows a published claim false (The Publication Choice Problem, AAAI 2026 (arXiv:2511.13678v2), Theorem 4.1 (a binary-type problem satisfying the standing assumptions and Assumptions 1-2 has a unique pure-strategy equilibrium): false; counterexample alpha = 1/2, beta = 2, two venues, types (1, 8), three equilibria)

Lean declarations

  • publication_choice_binary_equilibrium_not_uniqueTheorem 4.1 of arXiv:2511.13678 is false: not every binary-type Publication Choice Problem satisfying the standing assumptions, Assumption…
  • publication_choice_three_equilibriaA binary-type Publication Choice Problem with two venues, alpha = 1/2, beta = 2, satisfying Assumptions 1 and 2, has three pure-strategy…
  • pcp_is_validStanding assumptions of a Publication Choice Problem: alpha in (0,1), beta > 1, positive type densities, types and costs, and costs…
  • pcp_monotone_cost_ratioAssumption 1 (Monotone Cost Ratio): for types theta_i < theta_i' and venues j < j', c_ij / c_i'j < c_ij' / c_i'j'.
  • pcp_noncompetitive_venueAssumption 2 (non-competitive venue): the least competitive venue has the same publication cost for every type.
  • pcp_is_equilibriumPure-strategy equilibrium of a Publication Choice Problem (Definition 1): every researcher type best responds to the venue impacts, and…

Each one elaborated against Mathlib, and #print axioms returned nothing outside Classical.choice · Quot.sound · propext.

References

  1. arXiv:2511.13678

Drafted by gemini-3.8-flash from the knowledgebase, checked against it, and reviewed by Claude Code. Only the Lean is verified; the prose is a reading of it.