Pith. sign in

REVIEW 3 major objections 5 minor 67 references

Learnable Mixed Nash Equilibria are Collectively Rational

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A mixed Nash equilibrium that is uniformly stable under uncoupled learning dynamics must be strategically Pareto optimal: in connected, bi-directional N-player multilinear games, no joint deviation can strictly improve everyone's payoff.

desk verdict A genuinely interesting bridge between learning dynamics and collective rationality, but the main proof has a load-bearing gap (Lemma F.1 applies Theorem 2 off equilibrium) and the contraction argument in Theorem 3 doesn't justify the stated rate. read the letter →

arxiv 2510.14907 v2 pith:W6KLPX2X submitted 2025-10-16 cs.GT cs.LG

Geelon So , Yi-An Ma This is my paper · ORCID
classification cs.GTcs.LG MSC 91A1091A2637C75
keywords uniformstabilityNashequilibriumcollectiverationalitystrategicParetooptimalitysmoothedbest-responsedynamicslast-iterateconvergencemultilineargamesgameJacobian
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that, for N-player games with multilinear payoffs, the Nash equilibria that can be robustly learned by uncoupled, utility-seeking dynamics are exactly the ones that are collectively rational: no coalition can jointly deviate to strictly improve everyone's payoff, up to strategically irrelevant components. It proves this by linking a dynamical condition—uniform stability of the game Jacobian under all positive-definite block preconditioners—to a second-order economic condition called strategic Pareto stationarity, and from there to strategic Pareto optimality. It then shows that this stability notion controls last-iterate convergence of incremental smoothed best-response dynamics: stable regions admit T^{-1/2} convergence, while non-uniformly-stable equilibria are inapproximable for some regularizers. If correct, the result gives a principled sense in which bounded-rational self-interested learning avoids the collective inefficiency of prisoner's-dilemma-type outcomes, at least near mixed equilibria.

What carries the argument

Game Jacobian J(x): the block matrix of mixed second derivatives ∇²_{nm}f_n, whose block diagonals vanish because utilities are multilinear. Uniform stability of J means that every preconditioned matrix H^{-1}J, with H positive-definite and block-diagonal, has purely imaginary eigenvalues. A matrix is λ-skew when λ_n J_{nm} = -λ_m J_{mn}ᵀ, i.e., ΛJ is skew-symmetric; this expresses pairwise strict competition after rescaling. Strategic Pareto stationarity: the equilibrium Jacobian is λ-skew. The proofs turn on converting the spectral condition (uniform stability) into this algebraic λ-skew condition, then using bilinearity to lift pairwise competition to global Pareto optimality.

What would settle it

Construct an N-player multilinear game (N≥3) with connected, bi-directional interaction graph and a locally uniformly stable interior Nash equilibrium, and exhibit a joint deviation that strictly improves every player's strategic payoff. A direct way is to find a point x inside the stable neighborhood, not itself an equilibrium, where J(x) is uniformly stable but not λ-skew; Lemma F.1 and Theorem 1 both collapse if such a point exists. Equivalently, numerically check a three-player game with a trilinear term in each strategic utility for whether local uniform stability forces that term to vani

Watch

Extended reading notes

Core claim

On its own terms, the central claim is Theorem 1: if a Nash equilibrium x* of an N-player multilinear game lies in an open region where the game Jacobian J(x) is uniformly stable (H^{-1}J(x) has purely imaginary spectrum for every positive-definite block-diagonal H), and if the interaction graph is connected and bi-directional at x*, then x* is strategically Pareto optimal. The proof strategy is to show that local uniform stability forces the strategic component of the game to be bilinear, so that J is λ-skew-symmetric throughout; in a bilinear game, λ-skewness means every pair of players is in strict competition, which directly rules out joint improvements. The paper also proves the convers

Load-bearing premise

The proof of Theorem 1 assumes that uniform stability of the Jacobian on a whole neighborhood forces the Jacobian to be λ-skew at every point in that neighborhood, even though the theorem it cites is stated only at Nash equilibria; that extension is not separately proved.

Editorial extensions

If this is right

  • If a mixed Nash equilibrium is uniformly stable in a neighborhood, every smoothed best-response dynamics with sufficiently small learning rate converges to it at rate T^{-1/2}, so the dynamics are stabilizable to arbitrary accuracy.
  • If a mixed equilibrium is not pointwise uniformly stable, there exist regularizers for which no averaging rate η stabilizes the dynamics: the smoothed equilibria are unstable fixed points for small smoothing parameter β.
  • Collective rationality is not a separate assumption; it is forced by robust learnability under non-asymptotic stability, so prisoner's-dilemma-like inefficiency cannot survive as a stable mixed equilibrium.
  • For polymatrix games, uniform stability, strategic Pareto stationarity, and strategic Pareto optimality coincide under connected, bi-directional interactions.
  • Quasi-strict boundary equilibria reduce to interior ones: if the reduced game is locally uniformly stable and the regularizers are linearly steep and proper, smoothed best-response dynamics still stabilize to the equilibrium.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial: The main theorem inherits a hidden premise — Lemma F.1 applies the pointwise characterization Theorem 2 to non-equilibrium points in a neighborhood; unless uniform stability of J(x) itself implies λ-skew for arbitrary x, the bilinearity step in Theorem 1 is unsupported. A reader should check this before relying on the theorem.
  • Editorial: The results identify 'learnable mixed Nash equilibria' with Pareto-efficient ones, suggesting a mechanism for the invisible hand in bounded-rational learning; but the paper leaves open whether escape from inefficient equilibria is itself collectively rational (explicit Open Question 1).
  • Editorial: A testable extension is to relax bi-directionality to directed interactions; the theorem predicts stable mixed equilibria in such games would still have to be Pareto optimal, which could be checked directly on cyclic games.
  • Editorial: For two-player games, the characterization says uniform stability of a mixed equilibrium is equivalent to strictly competitive structure; this can be verified by direct calculation in small 2×2 games.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper introduces a spectral stability notion for mixed Nash equilibria — uniform stability, defined by requiring the eigenvalues of H^{-1}J(x*) to be purely imaginary for every positive-definite block-diagonal H — and links it to collective rationality. The main claims are: (i) local uniform stability at a Nash equilibrium implies strategic Pareto optimality (Theorem 1); (ii) strategic Pareto optimality is equivalent, under connectedness and bi-directionality, to strategic Pareto stationarity, which in turn is equivalent to pointwise uniform stability (Proposition 1 and Theorem 2); and (iii) uniform stability governs stabilization and last-iterate convergence of incremental smoothed best-response dynamics, with non-convergence in the opposite case (Proposition 2, Theorems 3 and 4). The paper also extends the convergence results to partially mixed, quasi-strict equilibria.

Significance. If the results are correct, the paper makes a valuable conceptual contribution: it connects a dynamical learnability notion for mixed equilibria to a normative economic condition, and it provides a sharp contrast with the well-known socially inefficient stabilization of strict equilibria. The definitions of uniform stability and strategic Pareto stationarity are natural, and the chain of implications is elegant. The appendices are extensive and contain several reusable linear-algebraic and game-theoretic lemmas. However, the current version has several load-bearing proof gaps, mostly in the appendices, so the claims are not yet established as written.

major comments (3)
  1. [Appendix F, Lemma F.1 / Theorem 1] Lemma F.1 begins by applying Theorem 2 to every point x in a neighborhood U of the equilibrium. But Theorem 2 is stated only for a Nash equilibrium, and its proof uses weak Pareto optimality of that equilibrium. The points x in U are not assumed to be Nash equilibria; Definition 14 only says that J(x) is a uniformly stable matrix. Without a separate argument — for example, subtracting the linear term ∇_n f_n(x)^T z_n from f_n to make x a Nash equilibrium of a modified game with the same Jacobian, or a direct matrix-theoretic lemma — the conclusion that J(x) is λ(x)-skew is unjustified. This step is load-bearing because it eliminates the third-order derivatives and is the basis of the bilinearity claim in Theorem 1.
  2. [Appendix K, Theorem 3] The contraction proof bounds ∥M_η(x)∥₂ by |1−η + iηL/β|. This is an eigenvalue modulus bound, not an operator-norm bound; M_η = (1−η)I + (η/β)H(x)^{-1}J(x) is generally non-normal, since H^{-1}J is only similar to a skew-symmetric matrix rather than skew-symmetric itself. The mean-value argument requires an operator norm, so the displayed inequality is not justified. The proof also establishes contraction only for x in an open ball U, while the theorem claims global convergence; no argument shows that orbits from arbitrary initial conditions enter U. These gaps affect the claimed stabilization result and the T^{-1/2} rate.
  3. [Appendix I, Theorem 2 proof] The displayed similarity H^{-1}J ∼ (H^{-1/2}Λ^{1/2})Σ(Λ^{1/2}H^{-1/2}) is algebraically incorrect: the conjugating factors do not cancel. The conclusion is repairable by setting K=ΛH and observing K^{1/2}(H^{-1}J)K^{-1/2}=K^{-1/2}ΣK^{-1/2}, which is skew-symmetric. In the forward direction, the proof obtains HJx=x; to show non-uniform stability one must replace H by H^{-1} so that (H^{-1})^{-1}J has eigenvalue 1. These are local algebraic slips, but they obscure one of the central equivalences.
minor comments (5)
  1. [Appendix J, Proposition 2] The condition 'β0 < 2/c' appears to be a typo: the displayed instability inequality requires β < c/2, so the threshold should be β0 < c/2.
  2. [Appendix H, Lemma H.2] In the definition of λ, the denominator is written u_1^T B v_2; it should presumably be u_1^T B v_1.
  3. [Definition 25] 'For each i∈[i]' should read 'for each i∈[k]'.
  4. [Appendix F, Lemma F.1] The quotient λ_nm(x)=λ_m(x)/λ_n(x) is only meaningful when the corresponding Jacobian blocks are nonzero; the case of zero blocks (absent edges) should be handled explicitly.
  5. [Appendix K, Theorem 3 proof] The sentence 'The rate then follows from Theorem 2' seems to refer to the preceding norm computation rather than Theorem 2; please rephrase to avoid ambiguity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the Lemma F.1 concern is a proof gap, not a definitional reduction.

full rationale

The paper's central chain—local uniform stability implies strategic Pareto optimality—is built from distinct definitions: uniform stability is a spectral condition on H^{-1}J for all positive-definite block-diagonal H, while strategic Pareto stationarity is λ-skewness of the game Jacobian. The equivalence of these notions (Theorem 2) is proved using independent linear algebra and game-theoretic arguments, not by definition. Proposition 1, Lemma 3, and the contrapositive arguments in Section I do not assume the conclusion of Theorem 1. The dependency graph is acyclic: Lemma F.1 invokes Theorem 2, but Theorem 2 does not invoke Lemma F.1 or Theorem 1. The paper contains no parameter fitting, no empirical prediction that is forced by construction, and no load-bearing self-citations by the authors. The reader-identified problem in Lemma F.1—applying Theorem 2 to arbitrary points x in a neighborhood U that are not themselves Nash equilibria—is a possible invalidity in an auxiliary lemma, but it is not circularity: it is a question of whether the lemma's hypotheses are satisfied, not a step where the theorem's conclusion is equivalent to an input by definition or by fitted values. Accordingly, the appropriate circularity score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities beyond the definitions of uniform stability and strategic Pareto stationarity. The central claims rely on domain assumptions about the game structure (bidirectionality, connectedness) and the regularizers (steepness, linear steepness), plus standard results from linear algebra and dynamical systems.

assumptions (5)
  • domain assumption Bi-directional interactions at the equilibrium (Definition 15): ker(J_nm) = ker(J_mn^T) for all n,m.
    Used in Theorem 1, Theorem 2, and Proposition 1 to ensure that the bilinear structure of the game Jacobian is sufficient to characterize Pareto optimality. Without it, the equivalence between uniform stability and strategic Pareto stationarity may fail (e.g., one-sided interactions).
  • domain assumption Connected interaction graph at the equilibrium (Definition 16).
    Used to propagate the λ-skew property across all players; the proofs of Lemma H.4 and Lemma F.1 rely on connectedness to ensure that the scaling factors are consistent across the whole graph.
  • domain assumption Steep regularizers (Definition 9) for smoothed best-response dynamics.
    Ensures that the smoothed best-response map Φβ maps into the interior of the simplex and is smooth; invoked in Lemma 2 and throughout Section 5.
  • domain assumption Quasi-strict equilibrium (Definition 23) for boundary equilibria.
    Used in Theorem 4 to ensure that unsupported actions are strictly dominated, so the reduced game is well-defined and the equilibrium becomes interior.
  • domain assumption Linearly steep regularizers (Definition 25) and proper regularizers (Definition 26).
    These quantitative steepness conditions are needed in Lemma 4 and Theorem 4 to control the probability placed on suboptimal actions and to ensure smooth extension of the Hessian pseudoinverse to the boundary.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learnable Mixed Nash Equilibria are Collectively Rational." pith.science (2026). https://pith.science/paper/W6KLPX2X

@misc{pith2026251014907,
  author       = {Pith},
  title        = {Pith review of: Learnable Mixed Nash Equilibria are Collectively Rational},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W6KLPX2X}},
  note         = {Machine review of arXiv:2510.14907}
}
read the original abstract

We extend the study of learning in games to dynamics that exhibit non-asymptotic stability. We do so through the notion of uniform stability, which is concerned with equilibria of individually utility-seeking dynamics. Perhaps surprisingly, it turns out to be closely connected to economic properties of collective rationality. Up to strategic equivalence, if a mixed equilibrium is uniformly stable, then it is weakly Pareto optimal; there is no way for all players to improve by jointly deviating from the equilibrium. This is a form of collective rationality that rules out the types of behaviors in the prisoner's dilemma or the tragedy of the commons. Moreover, we show that uniform stability determines the last-iterate convergence behavior for the family of incremental smoothed best-response dynamics, used to model individual and corporate behaviors in the markets. Unlike dynamics around strict equilibria, which can stabilize to socially-inefficient solutions, individually utility-seeking behaviors near mixed Nash equilibria lead to collective rationality.

Figures

Figures reproduced from arXiv: 2510.14907 by the authors.

Figure 1
Figure 1. Uncoupled learning dynamics in two 2 × 2 normal-form games with purely-strategic utilities f = (f1, f2). The Nash equilibria are marked by stars. The streamlines visualize the trajectories of the learning dynamics. The heatmap plots the social welfare function min{f1, f2}, measuring the utility of the player worst-off. The heatmap is white where the utilities are equal to the equilibrium; the darker the red, the wor… view at source ↗
Figure 2
Figure 2. The trajectories of β-smoothed best-response dynamics with η-learning rate toward a uniformly-stable, mixed Nash equilibrium (star) in a two-player normal-form game. (a) The β-smoothed equilibria become better approximations of the Nash equilibrium as β shrinks, but the dynamics become less stable and exhibit more cycling. The figure on the left plots the trajectories initialized at the black dot for varying smoothi… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

67 extracted references · 4 linked inside Pith

  1. [1]

    A note on strictly competitive games

    Ilan Adler, Constantinos Daskalakis, and Christos H Papadimitriou. A note on strictly competitive games. In International Workshop on Internet and Network Economics, pages 471--474. Springer, 2009

    2009

  2. [2]

    Uncertainty, evolution, and economic theory

    Armen A Alchian. Uncertainty, evolution, and economic theory. Journal of political economy, 58 0 (3): 0 211--221, 1950

    1950

  3. [3]

    The logit-response dynamics

    Carlos Al \'o s-Ferrer and Nick Netzer. The logit-response dynamics. Games and Economic Behavior, 68 0 (2): 0 413--427, 2010

    2010

  4. [4]

    On last-iterate convergence beyond zero-sum games

    Ioannis Anagnostides, Ioannis Panageas, Gabriele Farina, and Tuomas Sandholm. On last-iterate convergence beyond zero-sum games. In International Conference on Machine Learning, pages 536--581. PMLR, 2022

    2022

  5. [5]

    Acceptable points in general cooperative n -person games

    Robert J Aumann. Acceptable points in general cooperative n -person games. Contributions to the Theory of Games, 4: 0 287, 1959

    1959

  6. [6]

    Stochastic multiplicative weights updates in zero-sum games

    James P Bailey, Sai Ganesh Nagarajan, and Georgios Piliouras. Stochastic multiplicative weights updates in zero-sum games. arXiv preprint arXiv:2110.02134, 2021

    arXiv 2021

  7. [7]

    The mechanics of n -player differentiable games

    David Balduzzi, Sebastien Racaniere, James Martens, Jakob Foerster, Karl Tuyls, and Thore Graepel. The mechanics of n -player differentiable games. In International Conference on Machine Learning, pages 354--363. PMLR, 2018

    2018

  8. [8]

    Mixed equilibria and dynamical systems arising from fictitious play in perturbed games

    Michel Bena m and Morris W Hirsch. Mixed equilibria and dynamical systems arising from fictitious play in perturbed games. Games and Economic Behavior, 29 0 (1-2): 0 36--72, 1999

    1999

Show all 67 references
  1. [9]

    L yapunov theory for discrete time systems

    Nicoletta Bof, Ruggero Carli, and Luca Schenato. L yapunov theory for discrete time systems. arXiv preprint arXiv:1809.05289, 2018

    2018 arXiv

  2. [10]

    Zero-sum polymatrix games: A generalization of minmax

    Yang Cai, Ozan Candogan, Constantinos Daskalakis, and Christos Papadimitriou. Zero-sum polymatrix games: A generalization of minmax. Mathematics of Operations Research, 41 0 (2): 0 648--655, 2016

    2016

  3. [11]

    Flows and decompositions of games: H armonic and potential games

    Ozan Candogan, Ishai Menache, Asuman Ozdaglar, and Pablo A Parrilo. Flows and decompositions of games: H armonic and potential games. Mathematics of Operations Research, 36 0 (3): 0 474--503, 2011

    2011

  4. [12]

    Learning behavior and mixed-strategy N ash equilibria

    Vincent P Crawford. Learning behavior and mixed-strategy N ash equilibria. Journal of Economic Behavior & Organization, 6 0 (1): 0 69--78, 1985

    1985

  5. [13]

    Last-iterate convergence: Zero-sum games and constrained min-max optimization

    Constantinos Daskalakis and Ioannis Panageas. Last-iterate convergence: Zero-sum games and constrained min-max optimization. In 10th Innovations in Theoretical Computer Science (ITCS) conference, ITCS 2019, 2019

    2019

  6. [14]

    The complexity of computing a N ash equilibrium

    Constantinos Daskalakis, Paul W Goldberg, and Christos H Papadimitriou. The complexity of computing a N ash equilibrium. Communications of the ACM, 52 0 (2): 0 89--97, 2009

    2009

  7. [15]

    On learning algorithms for N ash equilibria

    Constantinos Daskalakis, Rafael Frongillo, Christos H Papadimitriou, George Pierrakos, and Gregory Valiant. On learning algorithms for N ash equilibria. In International Symposium on Algorithmic Game Theory, pages 114--125. Springer, 2010

    2010

  8. [16]

    Optimistic gradient descent ascent in zero-sum and general-sum bilinear games

    \'E tienne de Montbrun and J \'e r \^o me Renault. Optimistic gradient descent ascent in zero-sum and general-sum bilinear games. arXiv preprint arXiv:2208.03085, 2022

    2022 arXiv

  9. [17]

    Theory of value: An axiomatic analysis of economic equilibrium

    Gerard Debreu. Theory of value: An axiomatic analysis of economic equilibrium. Wiley, New York, 1959

    1959

  10. [18]

    Learning purified mixed equilibria

    Glenn Ellison and Drew Fudenberg. Learning purified mixed equilibria. Journal of Economic Theory, 90 0 (1): 0 84--115, 2000

    2000

  11. [19]

    An optimal bound for the spectral variation of two matrices

    Ludwig Elsner. An optimal bound for the spectral variation of two matrices. Linear algebra and its applications, 71: 0 77--80, 1985

    1985

  12. [20]

    The methodology of positive economics

    Milton Friedman. The methodology of positive economics. The Essays in Positive Economics, 1953

    1953

  13. [21]

    Learning mixed equilibria

    Drew Fudenberg and David M Kreps. Learning mixed equilibria. Games and economic behavior, 5 0 (3): 0 320--367, 1993

    1993

  14. [22]

    The theory of learning in games, volume 2

    Drew Fudenberg and David K Levine. The theory of learning in games, volume 2. MIT press, 1998

    1998

  15. [23]

    Survival of the strictest: Stable and unstable equilibria under regularized learning with partial information

    Angeliki Giannou, Emmanouil Vasileios Vlatakis-Gkaragkounis, and Panayotis Mertikopoulos. Survival of the strictest: Stable and unstable equilibria under regularized learning with partial information. In Conference on Learning Theory, pages 2147--2148. PMLR, 2021

    2021

  16. [24]

    First-order algorithm with convergence for-equilibrium in two-person zero-sum games

    Andrew Gilpin, Javier Pena, and Tuomas Sandholm. First-order algorithm with convergence for-equilibrium in two-person zero-sum games. Mathematical programming, 133 0 (1): 0 279--298, 2012

    2012

  17. [25]

    The rate of convergence of continuous fictitious play

    Steven Gjerstad. The rate of convergence of continuous fictitious play. Economic Theory, 7: 0 161--177, 1996

    1996

  18. [26]

    Quantal response equilibrium: A stochastic theory of games

    Jacob K Goeree, Charles A Holt, and Thomas R Palfrey. Quantal response equilibrium: A stochastic theory of games. In Quantal response equilibrium. Princeton University Press, 2016

    2016

  19. [27]

    The tragedy of the commons

    Garrett Hardin. The tragedy of the commons. Science, 162 0 (3859): 0 1243--1248, 1968

    1968

  20. [28]

    Games with randomly disturbed payoffs: A new rationale for mixed-strategy equilibrium points

    John C Harsanyi. Games with randomly disturbed payoffs: A new rationale for mixed-strategy equilibrium points. International journal of game theory, 2 0 (1): 0 1--23, 1973

    1973

  21. [29]

    A general theory of equilibrium selection in games

    John C Harsanyi and Reinhard Selten. A general theory of equilibrium selection in games. MIT Press Books , 1, 1988

    1988

  22. [30]

    Uncoupled dynamics do not lead to N ash equilibrium

    Sergiu Hart and Andreu Mas-Colell. Uncoupled dynamics do not lead to N ash equilibrium. American Economic Review, 93 0 (5): 0 1830--1836, 2003

    2003

  23. [31]

    Evolutionary games and population dynamics

    Josef Hofbauer and Karl Sigmund. Evolutionary games and population dynamics. Cambridge university press, 1998

    1998

  24. [32]

    Best response dynamics for continuous zero-sum games

    Josef Hofbauer and Sylvain Sorin. Best response dynamics for continuous zero-sum games. Discrete and Continuous Dynamical Systems Series B, 6 0 (1): 0 215, 2006

    2006

  25. [33]

    A note on best response dynamics

    Ed Hopkins. A note on best response dynamics. Games and Economic Behavior, 29 0 (1-2): 0 138--150, 1999

    1999

  26. [34]

    Matrix analysis

    Roger A Horn and Charles R Johnson. Matrix analysis. Cambridge university press, 2012

    2012

  27. [35]

    Differential games: a mathematical theory with applications to warfare and pursuit, control and optimization

    Rufus Isaacs. Differential games: a mathematical theory with applications to warfare and pursuit, control and optimization. Courier Corporation, 1999

    1999

  28. [36]

    Three problems in learning mixed-strategy N ash equilibria

    James S Jordan. Three problems in learning mixed-strategy N ash equilibria. Games and Economic Behavior, 5 0 (3): 0 368--386, 1993

    1993

  29. [37]

    Graphical games

    Michael Kearns. Graphical games. In The New Palgrave Dictionary of Economics, pages 1--4. Springer, 2008

    2008

  30. [38]

    Beyond the N ash equilibrium barrier

    Robert D Kleinberg, Katrina Ligett, Georgios Piliouras, and \'E va Tardos. Beyond the N ash equilibrium barrier. In ICS, volume 20, pages 125--140, 2011

    2011

  31. [39]

    Worst-case equilibria

    Elias Koutsoupias and Christos Papadimitriou. Worst-case equilibria. In Annual symposium on theoretical aspects of computer science, pages 404--413. Springer, 1999

    1999

  32. [40]

    On the convergence of fictitious play

    Vijay Krishna and Tomas Sj \"o str \"o m. On the convergence of fictitious play. Mathematics of Operations Research, 23 0 (2): 0 479--511, 1998

    1998

  33. [41]

    Linear algebra and its applications

    Peter D Lax. Linear algebra and its applications. John Wiley & Sons, 2007

    2007

  34. [42]

    Differentiable game mechanics

    Alistair Letcher, David Balduzzi, S \'e bastien Racaniere, James Martens, Jakob Foerster, Karl Tuyls, and Thore Graepel. Differentiable game mechanics. Journal of Machine Learning Research, 20 0 (84): 0 1--40, 2019

    2019

  35. [43]

    Games and decisions: Introduction and critical survey

    R Duncan Luce and Howard Raiffa. Games and decisions: Introduction and critical survey. John Wiley & Sons, Inc., 1957

    1957

  36. [44]

    Quantal response equilibria for normal form games

    Richard D McKelvey and Thomas R Palfrey. Quantal response equilibria for normal form games. Games and economic behavior, 10 0 (1): 0 6--38, 1995

    1995

  37. [45]

    Learning in games via reinforcement and regularization

    Panayotis Mertikopoulos and William H Sandholm. Learning in games via reinforcement and regularization. Mathematics of Operations Research, 41 0 (4): 0 1297--1324, 2016

    2016

  38. [46]

    Cycles in adversarial regularized learning

    Panayotis Mertikopoulos, Christos Papadimitriou, and Georgios Piliouras. Cycles in adversarial regularized learning. In Proceedings of the twenty-ninth annual ACM-SIAM symposium on discrete algorithms, pages 2703--2717. SIAM, 2018

    2018

  39. [47]

    An impossibility theorem in game dynamics

    Jason Milionis, Christos Papadimitriou, Georgios Piliouras, and Kelly Spendlove. An impossibility theorem in game dynamics. Proceedings of the National Academy of Sciences, 120 0 (41): 0 e2305349120, 2023

    2023

  40. [48]

    Potential games

    Dov Monderer and Lloyd S Shapley. Potential games. Games and economic behavior, 14 0 (1): 0 124--143, 1996

    1996

  41. [49]

    Strategically zero-sum games: the class of games whose completely mixed equilibria cannot be improved upon

    Herv \'e Moulin and J-P Vial. Strategically zero-sum games: the class of games whose completely mixed equilibria cannot be improved upon. International Journal of Game Theory, 7 0 (3-4): 0 201--221, 1978

    1978

  42. [50]

    E volutionary

    John H Nachbar. “ E volutionary” selection dynamics in games: Convergence and limit properties. International Journal of Game Theory, 19 0 (1): 0 59--89, 1990

    1990

  43. [51]

    Non-cooperative games

    John Nash. Non-cooperative games. Annals of Mathematics, 54 0 (2): 0 286--295, 1951

    1951

  44. [52]

    Payoff performance of fictitious play

    Georg Ostrovski and Sebastian van Strien. Payoff performance of fictitious play. arXiv preprint arXiv:1308.4049, 2013

    2013 arXiv

  45. [53]

    Towards a taxonomy of learning dynamics in 2 2 games

    Marco Pangallo, James BT Sanders, Tobias Galla, and J Doyne Farmer. Towards a taxonomy of learning dynamics in 2 2 games. Games and Economic Behavior, 132: 0 1--21, 2022

    2022

  46. [54]

    The complexity of finding N ash equilibria

    Christos H Papadimitriou. The complexity of finding N ash equilibria. Algorithmic game theory, 2: 0 30, 2007

    2007

  47. [55]

    Fast convergence of optimistic gradient ascent in network zero-sum extensive form games

    Georgios Piliouras, Lillian Ratliff, Ryann Sim, and Stratis Skoulakis. Fast convergence of optimistic gradient ascent in network zero-sum extensive form games. In International Symposium on Algorithmic Game Theory, pages 383--399. Springer, 2022

    2022

  48. [56]

    An iterative method of solving a game

    Julia Robinson. An iterative method of solving a game. Annals of mathematics, 54 0 (2): 0 296--301, 1951

    1951

  49. [57]

    Convex analysis, volume 28

    R Tyrrell Rockafellar. Convex analysis, volume 28. Princeton university press, 1997

    1997

  50. [58]

    Evolutionary stability in asymmetric games

    Larry Samuelson and Jianbo Zhang. Evolutionary stability in asymmetric games. Journal of economic theory, 57 0 (2): 0 363--391, 1992

    1992

  51. [59]

    Some topics in two-person games

    Lloyd S Shapley. Some topics in two-person games. Advances in game theory, 1963

    1963

  52. [60]

    Nash convergence of gradient dynamics in general-sum games

    Satinder Singh, Michael J Kearns, and Yishay Mansour. Nash convergence of gradient dynamics in general-sum games. In UAI, pages 541--548, 2000

    2000

  53. [61]

    An inquiry into the nature and causes of the wealth of nations

    Adam Smith. An inquiry into the nature and causes of the wealth of nations. London: printed for W. Strahan; and T. Cadell, 1776

  54. [62]

    On the instability of mixed-strategy N ash equilibria

    Dale O Stahl II. On the instability of mixed-strategy N ash equilibria. Journal of Economic Behavior & Organization, 9 0 (1): 0 59--69, 1988

    1988

  55. [63]

    Refinements of N ash equilibrium

    Eric Van Damme. Refinements of N ash equilibrium. In Advances in Economic Theory: Sixth World Congress, volume 1, pages 32--75. Cambridge University Press New York, 1992

    1992

  56. [64]

    No-regret learning and mixed N ash equilibria: They do not mix

    Emmanouil-Vasileios Vlatakis-Gkaragkounis, Lampros Flokas, Thanasis Lianeas, Panayotis Mertikopoulos, and Georgios Piliouras. No-regret learning and mixed N ash equilibria: They do not mix. Advances in Neural Information Processing Systems, 33: 0 1380--1391, 2020

    2020

  57. [65]

    Linear last-iterate convergence in constrained saddle-point optimization

    Chen-Yu Wei, Chung-Wei Lee, Mengxiao Zhang, and Haipeng Luo. Linear last-iterate convergence in constrained saddle-point optimization. The Ninth International Conference on Learning Representations, 2021

    2021

  58. [66]

    The ‘as if’ approach to game theory: Three positive results and four obstacles

    J \"o rgen W Weibull. The ‘as if’ approach to game theory: Three positive results and four obstacles. European Economic Review, 38 0 (3-4): 0 868--881, 1994

    1994

  59. [67]

    Satisficing, selection, and the innovating remnant

    Sidney G Winter. Satisficing, selection, and the innovating remnant. The Quarterly Journal of Economics, 85 0 (2): 0 237--261, 1971

    1971

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.