Pith. sign in

REVIEW 3 major objections 4 minor 4 cited by

The paper proposes the first fully first-order algorithm—using only gradient evaluations, never Hessians or matrix inverses—that provably finds an ε-stationary Stackelberg equilibrium when several followers play a strongly monotone game.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review

2026-08-04 21:09 UTC pith:4XICZQ3Q

load-bearing objection The extension to multiple followers is a sensible idea, but the proof of the main rate has two load-bearing gaps; as written the central claim is not established. the 3 major comments →

arxiv 2509.08161 v1 pith:4XICZQ3Q submitted 2025-09-09 math.OC cs.GT

Finding a Multiple Follower Stackelberg Equilibrium: A Fully First-Order Method

April Niu , Kai Wang , Juba Ziani This is my paper
classification math.OC cs.GT MSC 91A6590C3090C33
keywords Stackelberg gamesmulti-follower gamesbilevel optimizationfully first-order methodLagrangian reformulationstrongly monotone gamesε-stationary equilibriumconvergence rate
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Stackelberg games model a leader committing to a strategy before several followers play a game in response. Computing such equilibria normally requires second derivatives and matrix inversions of the followers' equilibrium map, or exact best responses. This paper claims the first algorithm for multiple followers that needs only first-order gradient evaluations and still comes with a convergence guarantee: it reaches an ε-stationary equilibrium in O(k^2 ε^{-6-α}) gradient evaluations for any α>0. The method converts the two-level problem into a single constrained optimization problem through a Lagrangian penalty, then alternates between gradient-descent approximation of the followers' equilibrium and gradient-descent updates of the leader's strategy, while growing the penalty parameters. If the claim holds, multi-follower Stackelberg problems become solvable where Hessian-based implicit differentiation is impractical.

Core claim

The paper's central claim: a fully first-order method can find an ε-stationary Stackelberg equilibrium with one leader and k followers, if the followers' subgame is strongly monotone and smooth. The proof reformulates the bilevel problem as a single-level constrained program via the Lagrangian L_λ(x,y)=f(x,y)+Σ_i λ_i(g_i(x,y_i,y*_{-i}(x))-g_i(x,y*(x))), showing that as λ grows, the Lagrangian minimizer approaches y*(x) and its gradient approaches ∇F(x). Algorithm 1 alternates gradient-descent approximation of the followers' equilibrium, gradient-descent approximation of the Lagrangian minimizer, and leader updates using the Lagrangian gradient, with λ_t=t^ρ. The error splits into three terms

What carries the argument

The load-bearing object is the Lagrangian L_λ(x,y)=f(x,y)+Σ_i λ_i(g_i(x,y_i,y*_{-i}(x))-g_i(x,y*(x))). It replaces the constraint that y be a Nash equilibrium of the followers' game with a weighted penalty. Because the followers' game is strongly monotone, L_λ is strongly convex in y for λ large enough, so gradient descent on y converges linearly. Its minimizer y*_λ(x) lies within O(1/λ) of the true follower equilibrium y*(x), and the gap between ∇L*_λ(x) and the true leader gradient ∇F(x) is O(k^2/λ). The algorithm grows λ over time so this finite-penalty error becomes summable, while inner-loop iteration counts control the other two error terms from approximating the followers' equilibrium

Load-bearing premise

The argument rests on the followers' subgame being strongly monotone for every leader strategy, which guarantees a unique, smoothly varying follower equilibrium; if that fails, the Lagrangian may not be strongly convex and the convergence proof no longer applies.

What would settle it

Fix a leader strategy x, choose a large λ, compute y*_λ(x) by high-precision gradient descent on L_λ, and compare ∇L*_λ(x) with the true leader gradient ∇F(x) obtained by solving the followers' game exactly. If the difference does not shrink like O(1/λ) as Lemma 4.2 predicts, the Lagrangian proxy fails. Alternatively, run Algorithm 1 on a known strongly monotone two-follower game and check whether empirical gradient-evaluation counts track ε^{-6}; a much faster or divergent growth would indicate the theorem's bound is loose or wrong.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Hessian-free scalability: leader and follower strategy spaces can be high-dimensional because no Hessian or matrix inverse is ever formed; per-iteration work is a few gradient evaluations.
  • Approximate followers suffice: the leader can update even when followers' best responses are only approximated by gradient descent, so the method works when the lower-level game cannot be solved exactly.
  • First guarantee of its kind: for k>1 followers, this is the first convergence rate (O(k^2 ε^{-6-α})) with only first-order oracles, at quadratic cost in the number of followers.
  • Parameter-growth blueprint: the proof shows how to schedule penalty growth (λ_t=t^ρ, ρ>1) and inner-loop iterations so the cumulative gradient error stays finite and the outer loop needs only T=O(ε^{-2}) steps.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The O(ε^{-6-α}) rate looks conservative: the E1 term only needs O(log t) inner steps while the E2 term needs O(t^{2+α/2}), and the α-vs-ε' trade-off is an artifact of the proof's schedule. A tighter coupling may push the rate below ε^{-5}.
  • The method assumes exact gradient oracles for f and g_i. A stochastic variant with noisy or sampled gradients is the natural next step; the paper anticipates complexity blow-up, but variance-reduction techniques might control the inflation.
  • Because the followers' game must be strongly monotone, the method does not apply to games with multiple isolated equilibria or to weakly monotone subgames; a regularized or perturbed variant could broaden the class, paying a bias-complexity trade-off.
  • A practical test: on a smooth market-pricing or security game with k=2–5 followers, record gradient-evaluation counts versus ε; the observed power law can reveal whether the ε^{-6} factor is a real bottleneck or a proof artifact.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper studies Stackelberg games with one leader and k followers in continuous action spaces. It proposes a bilevel reformulation and a Lagrangian-penalty single-level surrogate, then an algorithm (Algorithm 1) that alternates between (i) obtaining an approximate follower equilibrium z_{t+1} with an outer monotone-game solver, (ii) minimizing the surrogate Lagrangian by gradient descent, and (iii) updating the leader by an approximate gradient. The main claim, Theorem 5.1, is a convergence rate of O(k^2 ε^{-6-α}) gradient evaluations to an ε-stationary Stackelberg equilibrium for any α>0, under strong monotonicity and smoothness of the followers' subgame. The proof attempts to decompose the gradient error into three terms E1, E2, E3, control them via inner-loop iteration counts M_{y,t} and M_{z,t}, and then sum over T=O(ε^{-2}) outer iterations.

Significance. If the main theorem were correct, the contribution would be notable: it would provide the first fully first-order method for multi-follower Stackelberg games with a convergence guarantee, avoiding Hessian inversion and implicit differentiation, and the use of the Cai-Zheng monotone-game solver as a black box is a legitimate design choice. The paper also contains no fitted parameters and makes a genuine effort to provide full proofs in the appendix. However, the central convergence analysis has two load-bearing gaps: the bound on E2 is arithmetically inconsistent with the stated choice of M_{z,t}, and Lemma 5.5 contracts the wrong objective (the true Lagrangian instead of the surrogate actually minimized in the algorithm). These gaps invalidate Theorem 5.1 as stated. The result may be salvageable with a substantially different analysis and a weaker rate, but the claimed O(k^2 ε^{-6-α}) guarantee is not established.

major comments (3)
  1. [§5.2, 'Bounding E2' and Proof of Theorem 5.1] The sufficient condition for E2 is derived incorrectly. Corollary 5.4 gives E2 = 2k^2 λ_t^2 ||z_{t+1}-y*(x_t)||^2. With λ_t = t^ρ and the manuscript's own inner-solver bound ||z-z*|| ≤ C_z/(μ_g √M_{z,t}), E2 ≤ const · k^2 t^{2ρ} C_z^2 / M_{z,t}. Requiring E2 ≤ t^{-(1+ε')} therefore needs M_{z,t} ≥ const · k^2 C_z^2 t^{2ρ+1+ε'}. Since the text states C_z = O(k), this is M_{z,t} ≥ O(k^4 t^{2ρ+1+ε'}). The paper instead states M_{z,t} ≥ C_z k t^{ρ+ε'+1}/μ_g and then sets M_{z,t}=O(k^2 t^{2+α/2}). With the paper's relation ρ = 1 + α/2 - ε', the correct exponent is 2ρ+1+ε' = 3 + α - ε', not 2+α/2, and the k-dependence is k^4, not k^2. Thus E2 is not summable under the stated inner-loop counts, and the total complexity O(k^2 ε^{-6-α}) does not follow. The text also alternates between ||Vz|| ≤ C_z/M and ||z-z*|| ≤ C_z/(μ√M); either convention preserves the mismatch.
  2. [§5.2, Lemma 5.5 and Algorithm 1, line 3] Lemma 5.5 conflates the surrogate \(\tilde L\) with the true Lagrangian \(L\). Algorithm 1 line 3 minimizes \(\tilde L_{\lambda_t}(x_t,\cdot,z_{t+1}) = f(x_t,\cdot) + \lambda_t \sum_i [g_i(x_t,\cdot,z_{-i,t+1}) - g_i(x_t,z_{t+1})]\), where \(z_{t+1}\) is only an approximate follower equilibrium. The proof of Lemma 5.5, however, invokes strong convexity and smoothness of the true Lagrangian \(L\) and asserts a contraction of \(y_{t+1}\) to \(y^*_{\lambda_t}(x_t)\). Gradient descent on \(\tilde L\) contracts to \(\arg\min_y \tilde L\), not to \(y^*_{\lambda_t}\). The distance between these two minimizers depends on \(\|z_{t+1}-y^*(x_t)\|\), and this drift term is never bounded in Lemma 5.5. Since the drift enters E1 in Corollary 5.4 multiplied by \(\lambda_t^2\), the claimed summability of E1, and hence the overall rate, is not established. This is a substantive gap, not a notational issue
  3. [Lemma 5.5] The statement '... ≤ C_y = O(t^{1-ρ})' is internally inconsistent for ρ>1: the left-hand side is positive, while O(t^{1-ρ}) tends to 0. In the proof, B_t is O(t^{-ρ}), so its partial sum from u=1 to t is O(1) (bounded by ζ(ρ)), not O(t^{1-ρ}). As written, Lemma 5.5 does not provide the uniform constant needed to choose M_{y,t}=O(log t). This is likely repairable, but it is another load-bearing error in the proof of the E1 bound.
minor comments (4)
  1. [Definition 2.4] The follower suboptimality condition writes 'g_i(x,y) ≤ min_{y'∈Y_i} g(x,y'_i,y_{-i})+ε'; the argument of g on the left should be (x,y_i,y_{-i}), not (x,y).
  2. [Algorithm 1] Line 3 mentions an accuracy parameter ε_{y,t} that is never defined or used in the analysis; the proof only tracks M_{y,t}. Either define ε_{y,t} and relate it to M_{y,t}, or remove it.
  3. [§5.2, 'Bounding E2'] The inner-solver rate is quoted inconsistently: one sentence gives ||Vz(x,z_{M_z,t})|| ≤ C_z/M_{z,t}, and the next uses ||z-z*|| ≤ C_z/(μ_g√M_{z,t}). Please state the exact Cai-Zheng last-iterate rate in gradient norm and derive the distance bound consistently.
  4. [Throughout] The notation alternates between y, y_t, y_{t+1}, z_t, z_{t+1}, and the subscripting in the surrogate \(\tilde L\) is easy to confuse with the true Lagrangian \(L\). A table of notation or a displayed definition immediately before Lemma 5.5 would help.

Circularity Check

0 steps flagged

No circularity: the convergence proof reduces to stated assumptions and an external black-box solver; no fitted parameter is relabeled as a prediction.

full rationale

The paper's main result (Theorem 5.1) is not circular. The error decomposition in Corollary 5.4 is derived from the definitions of F, L_λ, and \tilde L using the paper's own smoothness and strong-monotonicity assumptions (Lemmas B.1, B.5). E3 is controlled by increasing λ_t = t^ρ; E2 is controlled by taking enough iterations of the external Cai-Zheng [2023] monotone-game solver; E1 is controlled by the contraction in Lemma 5.5. No parameter is fitted to the claimed O(k^2 ε^{-6-α}) rate; the inner-loop counts M_z,t and M_y,t are chosen to make the displayed error terms summable. The Cai-Zheng result is an independent, external black-box theorem, and the only self-citation (Wang et al. 2022, by co-author Kai Wang) is used in related work to contrast with second-order methods, not as evidence for any proof step. The genuine weakness is a correctness gap, not circularity: Lemma 5.5 invokes strong convexity/smoothness of the true Lagrangian L to assert contraction of the iterate, while Algorithm 1 line 3 minimizes the surrogate \tilde L; the drift ∥arg min \tilde L - y*_λ∥ is not bounded, so the E1 bound as written does not follow. This is an omitted argument, not an equivalence-by-construction or a renamed fit, so it does not raise the circularity score.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The algorithm uses no invented physical or mathematical objects. It depends on standard smoothness/strong-convexity assumptions and on the external solver rate. The free parameters are hyperparameters of the algorithm and iteration counts, none of which are fitted to data. The main burden is the inner-loop count inconsistency which affects the proof, not the axiomatic input.

free parameters (4)
  • ρ = any ρ > 1
    Penalty exponent in λ_t = t^ρ. Chosen by hand; the final complexity exponent α is tied to ρ via ρ = 1+α/2-ε'. The choice affects the inner-loop requirements.
  • λ_0 = not specified
    Initial penalty value. The proof requires λ_t ≥ 2ℓ_f1/μ_g for strong convexity of the Lagrangian (Lemma B.6), but no specific choice is given for λ_0.
  • η_t = 1/ℓ_F1
    Leader step size fixed to inverse smoothness of F. This is theory-determined, not fitted to data.
  • M_{y,t}, M_{z,t} = O(log t), O(k^2 t^{2+α/2})
    Inner-loop iteration counts chosen to satisfy error bounds. The stated M_{z,t} is inconsistent with Corollary 5.4's E2 bound, making this a fitted value that does not achieve the claimed decay.
axioms (5)
  • domain assumption Followers' subgame is μ_g-strongly monotone and each g_i is μ_g-strongly convex in (x,y_i) (Assumption 1).
    Used for uniqueness of y*(x), strong convexity of the Lagrangian (Lemma B.6), and the inner solver's convergence.
  • domain assumption Joint smoothness of f and g, twice differentiability of g, and ℓ_g,2-Lipschitz Hessian of g (Assumption 2).
    Used throughout the error bounds: smoothness controls gradient differences and Lipschitzness of the implicit Jacobian (Lemma B.3).
  • domain assumption Bounded gradients: ||∇_x f|| ≤ ℓ_f0 and ||∇_x g_i|| ≤ ℓ_g0 (Assumption 3).
    Used in Lemma B.9 to bound ||x_t - x_{t-1}|| and in the Lagrangian error bounds.
  • standard math Cai-Zheng 2023 black-box: an accelerated method for strongly monotone games achieves ||Vz(z_M)|| ≤ C_z/√M with C_z = O(k).
    The paper relies on this external theorem as the inner solver for Line 2 of the algorithm. It is a cited result, not derived in the paper, but accepted as a background theorem.
  • standard math For a μ-strongly monotone, ℓ-smooth VI, ||DV|| ≤ ℓ and ||DV^{-1}|| ≤ 1/μ (Facchinei and Pang 2003).
    Invoked to bound the implicit Jacobian and the sensitivity of y*(x). Standard variational analysis.

pith-pipeline@v1.3.0-alltime-deepseek · 19053 in / 19576 out tokens · 204591 ms · 2026-08-04T21:09:36.041642+00:00 · methodology

0 comments
read the original abstract

In this work, we propose the first fully first-order method to compute an epsilon stationary Stackelberg equilibrium with convergence guarantees. To achieve this, we first reframe the leader follower interaction as single level constrained optimization. Second, we define the Lagrangian and show that it can approximate the leaders gradient in response to the equilibrium reached by followers with only first-order gradient evaluations. These findings suggest a fully first order algorithm that alternates between (i) approximating followers best responses through gradient descent and (ii) updating the leaders strategy via approximating the gradient using Lagrangian.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Finite-Time Analysis of Q-Value Iteration for General-Sum Stackelberg Games

    cs.LG 2026-04 unverdicted novelty 7.0

    Provides the first finite-time convergence guarantees for Q-value iteration in general-sum Stackelberg Markov games.

  2. Shared Infrastructure Investment and Pricing: Stackelberg Equilibria in Risk-Aware Take-or-Pay Contracts

    cs.GT 2026-06 conditional novelty 5.0

    The paper proves existence of a Stackelberg equilibrium and sufficient conditions for uniqueness of the follower variational equilibrium in a shared-infrastructure pricing game with risk-averse, congestion-coupled fir...

  3. Continuity of VaR and Continuous Differentiability of CVaR under Decision-Dependent Losses

    cs.GT 2026-06 unverdicted novelty 5.0

    Sufficient local conditions are derived for continuity of decision-dependent VaR and C1 differentiability of CVaR, plus an explicit gradient formula.

  4. Shared Infrastructure Investment and Pricing: Stackelberg Equilibria in Risk-Aware Take-or-Pay Contracts

    cs.GT 2026-06 unverdicted novelty 4.0

    Formalizes shared infrastructure as a risk-aware Stackelberg game with take-or-pay contracts, proves equilibrium existence, gives a polynomial-time approximation algorithm, derives a PoP lower bound, and simulates ris...

Reference graph

Works this paper leans on

36 extracted references · 27 canonical work pages · cited by 3 Pith papers · 1 internal anchor

  1. [1]

    Differentiable convex optimization layers

    Akshay Agrawal, Brandon Amos, Shane Barratt, Stephen Boyd, Steven Diamond, and J Zico Kolter. Differentiable convex optimization layers. Advances in neural information processing systems, 32, 2019

    2019

  2. [2]

    Optnet: Differentiable optimization as a layer in neural networks

    Brandon Amos and J Zico Kolter. Optnet: Differentiable optimization as a layer in neural networks. In International conference on machine learning, pages 136--145. PMLR, 2017

    2017

  3. [3]

    Stackelberg security games (ssg) basics and application overview

    Bo An, Milind Tambe, and Arunesh Sinha. Stackelberg security games (ssg) basics and application overview. Improving Homeland Security Decisions, 2: 0 485, 2017

    work page 2017

  4. [4]

    Sample-efficient learning of stackelberg equilibria in general-sum games

    Yu Bai, Chi Jin, Huan Wang, and Caiming Xiong. Sample-efficient learning of stackelberg equilibria in general-sum games. In Proceedings of the 35th International Conference on Neural Information Processing Systems, NIPS '21, Red Hook, NY, USA, 2021. Curran Associates Inc. ISBN 9781713845393

    work page 2021

  5. [5]

    Ba s ar and R

    T. Ba s ar and R. Srikant. A stackelberg network game with a large number of followers. Journal of Optimization Theory and Applications, 0 (3): 0 479--490, 2002

    work page 2002

  6. [6]

    On Stackelberg--Nash Equilibria in Bilevel Optimization Games, pages 27--51

    Damien Bazin, Ludovic Julien, and Olivier Musy. On Stackelberg--Nash Equilibria in Bilevel Optimization Games, pages 27--51. Springer International Publishing, 2020

    work page 2020

  7. [7]

    Doubly optimal no-regret learning in monotone games

    Yang Cai and Weiqiang Zheng. Doubly optimal no-regret learning in monotone games. In Proceedings of the 40th International Conference on Machine Learning, ICML'23. JMLR.org, 2023

    work page 2023

  8. [8]

    Near-optimal nonconvex-strongly-convex bilevel optimization with fully first-order oracles

    Lesi Chen, Yaohua Ma, and Jingzhao Zhang. Near-optimal nonconvex-strongly-convex bilevel optimization with fully first-order oracles. Journal of Machine Learning Research, 26 0 (109): 0 1--56, 2025

    work page 2025

  9. [9]

    Finite--Dimensional Variational Inequalities and Complementarity Problems, volume 1 of Springer Series in Operations Research and Financial Engineering

    Francisco Facchinei and Jong-Shi Pang. Finite--Dimensional Variational Inequalities and Complementarity Problems, volume 1 of Springer Series in Operations Research and Financial Engineering. Springer, 2003

    work page 2003

  10. [10]

    Tanner Fiez, Benjamin Chasnov, and Lillian J. Ratliff. Convergence of learning dynamics in stackelberg games. 2019. URL https://arxiv.org/abs/1906.01217

    Pith/arXiv arXiv 2019

  11. [11]

    Implicit learning dynamics in stackelberg games: Equilibria and convergence

    Tracy Fiez, Lillian J Ratliff, and Peter Seiler. Implicit learning dynamics in stackelberg games: Equilibria and convergence. In Advances in Neural Information Processing Systems, 2020

    work page 2020

  12. [12]

    Continuous-time convergence rates in potential and monotone games

    Bolin Gao and Lacra Pavel. Continuous-time convergence rates in potential and monotone games. SIAM Journal on Control and Optimization, 60 0 (3): 0 1712--1731, 2022

    work page 2022

  13. [13]

    Oracles & followers: Stackelberg equilibria in deep multi-agent reinforcement learning

    Matthias Gerstgrasser and David C Parkes. Oracles & followers: Stackelberg equilibria in deep multi-agent reinforcement learning. In International Conference on Machine Learning, pages 11213--11236. PMLR, 2023

    work page 2023

  14. [14]

    Approximation methods for bilevel programming, 2018

    Saeed Ghadimi and Mengdi Wang. Approximation methods for bilevel programming, 2018. URL https://arxiv.org/abs/1802.02246

    Pith/arXiv arXiv 2018

  15. [15]

    Zero-sum stochastic stackelberg games

    Denizalp Goktas, Sadie Zhao, and Amy Greenwald. Zero-sum stochastic stackelberg games. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Information Processing Systems, volume 35, pages 11658--11672. Curran Associates, Inc., 2022

    work page 2022

  16. [16]

    Convex-concave zero-sum markov stackelberg games

    Denizalp Goktas, Arjun Prakash, and Amy Greenwald. Convex-concave zero-sum markov stackelberg games. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors, Advances in Neural Information Processing Systems, volume 36, pages 66818--66832. Curran Associates, Inc., 2023

    work page 2023

  17. [17]

    Tight last-iterate convergence rates for no-regret learning in multi-player games

    Noah Golowich, Sarath Pattathil, and Constantinos Daskalakis. Tight last-iterate convergence rates for no-regret learning in multi-player games. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS '20, 2020

    work page 2020

  18. [18]

    Hierarchical game theory for system-optimal control: Applications of reverse stackelberg games in regulating marketing channels and traffic routing

    Noortje Groot, Georges Zaccour, and Bart De Schutter. Hierarchical game theory for system-optimal control: Applications of reverse stackelberg games in regulating marketing channels and traffic routing. IEEE Control Systems Magazine, 37 0 (2): 0 129--152, 2017

    work page 2017

  19. [19]

    A double oracle algorithm for zero-sum security games on graphs

    Manish Jain, Dmytro Korzhyk, Ond r ej Van e k, Vincent Conitzer, Michal P e chou c ek, and Milind Tambe. A double oracle algorithm for zero-sum security games on graphs. In The 10th International Conference on Autonomous Agents and Multiagent Systems - Volume 1, AAMAS '11. International Foundation for Autonomous Agents and Multiagent Systems, 2011

    work page 2011

  20. [20]

    Bilevel optimization: Convergence analysis and enhanced design

    Kaiyi Ji, Junjie Yang, and Yingbin Liang. Bilevel optimization: Convergence analysis and enhanced design. In International conference on machine learning, pages 4882--4892. PMLR, 2021

    2021

  21. [21]

    A fully first-order method for stochastic bilevel optimization

    Jeongyeol Kwon, Dohyun Kwon, Stephen Wright, and Robert Nowak. A fully first-order method for stochastic bilevel optimization. In Proceedings of the 40th International Conference on Machine Learning, ICML'23. JMLR.org, 2023

    work page 2023

  22. [22]

    End-to-end learning and intervention in games

    Jiayang Li, Jing Yu, Yu Nie, and Zhaoran Wang. End-to-end learning and intervention in games. Advances in Neural Information Processing Systems, 33, 2020

    work page 2020

  23. [23]

    A review of dynamic stackelberg game models

    Tao Li and Suresh P Sethi. A review of dynamic stackelberg game models. Discrete & Continuous Dynamical Systems-Series B, 22 0 (1), 2017

    work page 2017

  24. [24]

    Solving Structured Hierarchical Games Using Differential Backward Induction

    Zun Li, Feiran Jia, Aditya Mate, Shahin Jabbari, Mithun Chakraborty, Milind Tambe, and Yevgeniy Vorobeychik. Solving structured hierarchical games using differential backward induction, 2022. URL https://arxiv.org/abs/2106.04663

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  25. [25]

    Finite-time last-iterate convergence for multi-agent learning in games

    Tianyi Lin, Zhengyuan Zhou, Panayotis Mertikopoulos, and Michael Jordan. Finite-time last-iterate convergence for multi-agent learning in games. In International Conference on Machine Learning, pages 6161--6171. PMLR, 2020

    work page 2020

  26. [26]

    Bome! bilevel optimization made easy: A simple first-order approach

    Bo Liu, Mao Ye, Stephen Wright, Peter Stone, and Qiang Liu. Bome! bilevel optimization made easy: A simple first-order approach. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Information Processing Systems, volume 35, pages 17248--17262. Curran Associates, Inc., 2022

    work page 2022

  27. [27]

    First-order penalty methods for bilevel optimization

    Zhaosong Lu and Sanyou Mei. First-order penalty methods for bilevel optimization. SIAM Journal on Optimization, 34 0 (2): 0 1937--1969, 2024

    work page 1937

  28. [28]

    Shankar Sasty, Lillian Ratliff, and Eric Mazumdar

    Chinmay Maheshwari, James Cheng, S. Shankar Sasty, Lillian Ratliff, and Eric Mazumdar. Follower agnostic methods for stackelberg games, 2024. URL https://arxiv.org/abs/2302.01421

    Pith/arXiv arXiv 2024

  29. [29]

    Learning in games with continuous action sets and unknown payoff functions

    Panayotis Mertikopoulos and Zhengyuan Zhou. Learning in games with continuous action sets and unknown payoff functions. Mathematical Programming, 173 0 (1): 0 465--507, 2019

    2019

  30. [30]

    Learning nash equilibria in monotone games

    Tatiana Tatarenko and Maryam Kamgarpour. Learning nash equilibria in monotone games. In 2019 IEEE 58th Conference on Decision and Control (CDC), pages 3104--3109. IEEE, 2019

    work page 2019

  31. [31]

    Reiter, and Milind Tambe

    Kai Wang, Lily Xu, Andrew Perrault, Michael K. Reiter, and Milind Tambe. Coordinating followers to reach better equilibria: End-to-end gradient descent for stackelberg games. Proceedings of the AAAI Conference on Artificial Intelligence, Jun. 2022

    work page 2022

  32. [32]

    Alternating projected sgd for equality-constrained bilevel optimization

    Quan Xiao, Han Shen, Wotao Yin, and Tianyi Chen. Alternating projected sgd for equality-constrained bilevel optimization. In Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, volume 206 of Proceedings of Machine Learning Research, pages 987--1023. PMLR, 25--27 Apr 2023

    work page 2023

  33. [33]

    A one-leader multi-follower bayesian-stackelberg game for anti-jamming transmission in uav communication networks

    Yifan Xu, Guochun Ren, Jin Chen, Yunpeng Luo, Luliang Jia, Xin Liu, Yang Yang, and Yuhua Xu. A one-leader multi-follower bayesian-stackelberg game for anti-jamming transmission in uav communication networks. IEEE Access, 6, 2018

    work page 2018

  34. [34]

    Haikuo Yang, Luo Luo, Chris Junchi Li, and Michael I. Jordan. Accelerating inexact hypergradient descent for bilevel optimization, 2023. URL https://arxiv.org/abs/2307.00126

    Pith/arXiv arXiv 2023

  35. [35]

    Multi-agent reinforcement learning: A selective overview of theories and algorithms

    Kaiqing Zhang, Zhuoran Yang, and Tamer Ba s ar. Multi-agent reinforcement learning: A selective overview of theories and algorithms. Handbook of reinforcement learning and control, pages 321--384, 2021

    work page 2021

  36. [36]

    An introduction to bi-level optimization: Foundations and applications in signal processing and machine learning, 2023

    Yihua Zhang, Prashant Khanduri, Ioannis Tsaknakis, Yuguang Yao, Mingyi Hong, and Sijia Liu. An introduction to bi-level optimization: Foundations and applications in signal processing and machine learning, 2023. URL https://arxiv.org/abs/2308.00788

    Pith/arXiv arXiv 2023

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.