pith. sign in

arxiv: 2605.30392 · v2 · pith:7SR7NYL5new · submitted 2026-05-28 · 💻 cs.MA · cs.GT· math.DS

Delayed Repression and Emergent Instability in Adaptive Multi-Agent Systems

Igor Itkin This is my paper

Pith reviewed 2026-06-29 00:12 UTC · model grok-4.3

classification 💻 cs.MA cs.GTmath.DS
keywords delayed replicator equationHopf bifurcationsupercritical bifurcationmulti-agent systemsinstitutional delayQ-learningadaptive agentsradical behavior
0
0 comments X

The pith

Institutional processing delays alone can destabilize otherwise stable multi-agent systems through a supercritical Hopf bifurcation.

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

The paper asks whether lags in regulatory observation and intervention can destabilize populations of autonomous agents that would remain stable without shocks, coordination, or malice. In a delayed replicator model, agents gain from radical actions but receive punishment after an institutional lag; the analysis yields a closed-form critical delay at which the interior equilibrium loses stability via Hopf bifurcation. Center manifold reduction establishes that the bifurcation remains supercritical across the full sigmoid response family, producing bounded oscillations. Network simulations with 240 agents then compare decision rules and show that immediate reactivity to the lagged signal, rather than learning, produces the instability.

Core claim

In the delayed replicator equation with lagged institutional alarm, a closed-form critical delay exists beyond which the unique interior equilibrium loses stability through a Hopf bifurcation; center manifold reduction proves the bifurcation is supercritical for every sigmoid response function, yielding bounded large-amplitude oscillations. Simulations confirm that fixed-policy agents remain stable at all delays, reactive threshold agents reach 96 percent runaway by delay 8, and Q-learning agents reach only 66 percent runaway at delay 20, because value functions encode punishment memory that buffers immediate exploitation of low-alarm windows.

What carries the argument

The delayed replicator equation with lagged institutional punishment signal, whose local stability is analyzed by locating the Hopf bifurcation point and applying center manifold reduction to classify its type.

If this is right

  • Beyond the critical delay the interior equilibrium loses stability.
  • The resulting oscillations remain bounded rather than growing without limit.
  • Fixed-policy agents exhibit zero runaway at every tested delay.
  • Reactive threshold agents exhibit 96 percent runaway once delay reaches or exceeds 8.
  • Q-learning agents exhibit intermediate resilience because value functions retain memory of past punishments.

Where Pith is reading between the lines

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

  • If agents can coordinate their responses to the lagged signal, the effective critical delay may decrease.
  • Institutions that shorten processing time or add forward-looking components could raise the stability threshold without changing agent rules.
  • The same lag mechanism could be tested in other domains where agents respond to delayed negative feedback, such as market regulation or content platforms.

Load-bearing premise

Agents gain from radical behavior and receive punishment only from a lagged institutional signal, with no external shocks, coordination among agents, or malicious intent.

What would settle it

Compute the model's closed-form critical delay for a chosen sigmoid response, then run the replicator or agent simulation at delays just below and just above that value to check whether large-amplitude oscillations appear exactly at the predicted threshold.

Figures

Figures reproduced from arXiv: 2605.30392 by Igor Itkin.

Figure 1
Figure 1. Figure 1: The delay-destabilization mechanism. (a) Without delay, the institution observes the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Solution pipeline. Stage 1 derives analytical predictions from the delayed replicator [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bifurcation diagram for the delayed replicator ODE (Experiment 1). Horizontal axis: [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Representative ODE trajectories at six delay levels (Experiment 1). Below [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Effect of sharpness k on ODE dynamics at fixed ∆/∆c = 1.5 (Experiment 1). All panels are at the same distance beyond the stability boundary, yet higher k produces sharper, larger-amplitude limit cycles saturating near the population boundaries. Conclusion: sharpness amplifies the nonlinear consequences of delay-induced instability, confirming Hypothesis 2. 4.4 Central experiment: crossed delay × architectu… view at source ↗
Figure 6
Figure 6. Figure 6: Phase diagram of the discrete mean-field system in [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Discrete mean-field trajectories at ∆=1.2∆c for several η values (Experiment 2). Small η: bounded oscillations matching the ODE limit cycle. Large η: overshooting dynamics exceeding the ODE envelope. Conclusion: the full simulation’s discrete time steps (η ≈ 1) introduce additional instability beyond the continuous theory. 0 10 20 30 40 50 60 delay steps 0.0 0.2 0.4 0.6 0.8 1.0 Tail amplitude =1.0 0 10 20 … view at source ↗
Figure 8
Figure 8. Figure 8: Stability margin vs. discrete step size η (Experiment 2). As η → 0, the margin recovers ∆c from the continuous theory. Conclusion: discretization reduces the stability margin monotonically. This provides a quantitative bridge between the ODE prediction and the simulation’s inherent discrete dynamics. 4.6 Summary of findings The central result is the two-factor crossing (Experiment 5): reactive agents colla… view at source ↗
Figure 9
Figure 9. Figure 9: Delay sweep (Experiment 3; k=20, Q-learning, 50 seeds). Left: runaway rate with 95% Wilson band. Right: excess runaway over the ∆=0 baseline. Conclusion: monotonic increase con￾firms Hypothesis 1: delay alone produces a large, dose-dependent destabilization in the networked simulation. k Stable Oscillatory Runaway 3 70% 14% 16% 5 56% 24% 20% 7 40% 32% 28% 10 26% 22% 52% 15 26% 14% 60% 20 24% 16% 60% 30 20%… view at source ↗
Figure 10
Figure 10. Figure 10: Runaway rate vs. sharpness k at fixed delay=15 (Experiment 4; 50 seeds). Vertical lines: k=10 (used in Experiment 5) and k=20 (Experiment 3). Conclusion: the sharp transition at k=7–10 marks the system crossing ∆c (Equation 10). Above this threshold, further sharpening produces diminishing returns, a ceiling effect. behavior rather than oscillating between extremes), and spatial heterogeneity (network str… view at source ↗
Figure 11
Figure 11. Figure 11: Runaway rate vs. delay for three agent architectures (Experiment 5, the central ex [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Fine-resolution delay sweep for reactive agents ( [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Regime distribution for three governance configurations (Experiment 6; 500 steps, 50 [PITH_FULL_IMAGE:figures/full_fig_p023_13.png] view at source ↗
read the original abstract

Regulatory institutions (from content moderation platforms to financial supervisors) observe, deliberate, and intervene only after a characteristic delay. We ask whether this processing lag alone can destabilize a multi-agent system that would otherwise remain stable, without exogenous shocks, coordination among agents, or malicious actors. We study this in two stages. First, we analyze a delayed replicator equation in which autonomous agents benefit from radical behavior but face punishment based on a lagged institutional alarm signal. We derive a closed-form critical delay beyond which the unique interior equilibrium loses stability through a Hopf bifurcation, and prove via center manifold reduction that the bifurcation is supercritical (bounded oscillations, not explosive growth) for the entire sigmoid response family. Second, we embed N=240 agents on a network with reinforcement learning (tabular Q-learning) and cross institutional delay with three decision architectures: fixed-policy, reactive (a memoryless threshold heuristic), and Q-learning. The hierarchy is opposite to the naive expectation that learning amplifies instability. Reactive agents are perfectly stable without delay yet collapse once delay is introduced (96% runaway by delay >= 8); fixed-policy agents are immune (0% at all delays); Q-learning agents are only partially resilient (66% at delay 20). The destabilizing ingredient is reactivity to delayed signals, not learning: agents that immediately exploit low-alarm windows trigger oscillatory feedback loops, while learning buffers this through punishment memory encoded in value functions. Throughout, "runaway" denotes bounded large-amplitude oscillation crossing a radical-fraction threshold, consistent with the supercritical bifurcation, not unbounded growth.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper studies whether institutional processing delays alone can destabilize multi-agent systems. It first analyzes a delayed replicator equation where agents gain from radical behavior but receive lagged punishment via a sigmoid institutional response. It derives a closed-form critical delay at which the interior equilibrium loses stability via Hopf bifurcation and uses center-manifold reduction to prove the bifurcation is supercritical for the full sigmoid family. It then runs N=240 agent simulations on a network with tabular Q-learning, comparing fixed-policy, reactive threshold, and Q-learning architectures across delays; reactive agents show 96% runaway at delay >=8, fixed-policy remain stable at 0%, and Q-learning reach 66% at delay 20. The key conclusion is that reactivity to delayed signals, not learning per se, drives the instability, producing bounded large-amplitude oscillations consistent with the supercritical bifurcation.

Significance. If the closed-form delay and center-manifold proof hold, the work supplies a rigorous, parameter-free mechanism by which pure lag in regulatory feedback can induce endogenous oscillatory instability in otherwise stable multi-agent populations. The explicit demonstration that the bifurcation remains supercritical across the sigmoid family is a strength, as is the simulation hierarchy that isolates reactivity as the destabilizing factor and shows learning can buffer rather than amplify instability. These elements would be of interest to researchers in multi-agent systems, evolutionary game theory, and institutional design.

major comments (1)
  1. [Abstract / replicator analysis] Abstract and replicator-analysis section: the central claim is a closed-form critical delay for Hopf bifurcation together with a center-manifold proof of supercriticality for the entire sigmoid family, yet the characteristic equation, the explicit expression for the critical delay, and the normal-form coefficients obtained from the reduction are not supplied, so the algebraic correctness and coverage of the sigmoid family cannot be verified.
minor comments (1)
  1. [Simulation stage] Simulation stage: the reported runaway percentages (96%, 0%, 66%) are given without error bars, number of independent runs, or exclusion criteria, which limits assessment of robustness even though these results are secondary to the analytic claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the thorough review and for identifying the need for greater algebraic transparency in the replicator analysis. The comment correctly notes that the characteristic equation, explicit critical-delay formula, and normal-form coefficients are not supplied in the current version. We will revise the manuscript to include these derivations so that the Hopf bifurcation result and the supercriticality claim for the full sigmoid family can be verified directly.

read point-by-point responses

  1. Referee: [Abstract / replicator analysis] Abstract and replicator-analysis section: the central claim is a closed-form critical delay for Hopf bifurcation together with a center-manifold proof of supercriticality for the entire sigmoid family, yet the characteristic equation, the explicit expression for the critical delay, and the normal-form coefficients obtained from the reduction are not supplied, so the algebraic correctness and coverage of the sigmoid family cannot be verified.

    Authors: We agree that the characteristic equation, the closed-form expression for the critical delay, and the normal-form coefficients from the center-manifold reduction are absent from the submitted manuscript. This prevents independent checking of the bifurcation threshold and of the claim that the bifurcation remains supercritical for every member of the sigmoid family. In the revised manuscript we will insert the full derivation: the linearized delayed replicator equation, the resulting characteristic equation, the explicit formula for the critical delay τ* obtained from the Hopf condition, and the normal-form coefficients computed via center-manifold reduction that establish supercriticality uniformly across the sigmoid parameter range. These additions will be placed in the main replicator-analysis section (or a short appendix if length constraints require it) with all intermediate algebraic steps shown. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivations are direct from model equations

full rationale

The paper derives the closed-form critical delay and proves supercritical Hopf bifurcation via center manifold reduction explicitly from the delayed replicator equation and sigmoid response family. These steps are standard mathematical analysis applied to the stated model, not reductions to fitted inputs, self-definitions, or self-citation chains. Simulations report empirical outcomes from agent architectures under varying delays, without evidence that results are forced by parameter choice or prior author work. The derivation chain is self-contained against the model assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption of a delayed replicator equation with lagged punishment and the absence of external shocks or coordination; no free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption The system follows a delayed replicator equation in which agents gain from radical behavior but receive lagged punishment from an institutional alarm signal.
    This modeling choice underpins both the Hopf bifurcation derivation and the simulation stage.

pith-pipeline@v0.9.1-grok · 5812 in / 1449 out tokens · 33912 ms · 2026-06-29T00:12:08.829473+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Delayed Verification Destabilizes Multi-Agent LLM Belief: Instability Thresholds and Optimal Corrector Placement

    cs.MA 2026-06 unverdicted novelty 7.0

    Models delayed verification in multi-agent LLMs as graph consensus, derives stability thresholds (inverse golden ratio for delay two) via grounded Laplacian, and gives a supermodular greedy rule for corrector placemen...

Reference graph

Works this paper leans on

35 extracted references · 25 canonical work pages · cited by 1 Pith paper

  1. [1]

    Stability of evolutionarily stable strategies in discrete replicator dynamics with time delay

    Jan Alboszta and Jacek Miekisz. Stability of evolutionarily stable strategies in discrete replicator dynamics with time delay. Journal of Theoretical Biology, 231 0 (2): 0 175--179, 2004. doi:10.1016/j.jtbi.2004.06.012

    work page doi:10.1016/j.jtbi.2004.06.012 2004

  2. [2]

    Discrete and continuous distributed delays in replicator dynamics

    Nesrine Ben-Khalifa, Rachid El-Azouzi, and Yezekael Hayel. Discrete and continuous distributed delays in replicator dynamics. Dynamic Games and Applications, 8 0 (4): 0 713--732, 2018. doi:10.1007/s13235-017-0225-7

    work page doi:10.1007/s13235-017-0225-7 2018

  3. [3]

    Reinforcement learning with random delays

    Yann Bouteiller, Simon Ramstedt, Giovanni Beltrame, Christopher Pal, and Jonathan Binas. Reinforcement learning with random delays. In International Conference on Learning Representations (ICLR), 2021

    2021

  4. [4]

    Delay-aware multi-agent reinforcement learning for cooperative and competitive environments

    Baiming Chen, Mengdi Xu, Zuxin Liu, Liang Li, and Ding Zhao. Delay-aware multi-agent reinforcement learning for cooperative and competitive environments. arXiv preprint arXiv:2005.05441, 2020

    work page arXiv 2005

  5. [5]

    Acting in delayed environments with non-stationary markov policies

    Esther Derman, Gal Dalal, and Shie Mannor. Acting in delayed environments with non-stationary markov policies. In International Conference on Learning Representations (ICLR), 2021

    2021

  6. [6]

    Linton C. Freeman. A set of measures of centrality based on betweenness. Sociometry, 40 0 (1): 0 35--41, 1977. doi:10.2307/3033543

    work page doi:10.2307/3033543 1977

  7. [7]

    Michelle Girvan and Mark E. J. Newman. Community structure in social and biological networks. Proceedings of the National Academy of Sciences, 99 0 (12): 0 7821--7826, 2002. doi:10.1073/pnas.122653799

    work page doi:10.1073/pnas.122653799 2002

  8. [8]

    Multi-agent deep reinforcement learning: A survey.Artificial Intelligence Review, 55(2):895–943, 2022

    Sven Gronauer and Klaus Diepold. Multi-agent deep reinforcement learning: A survey. Artificial Intelligence Review, 55: 0 895--943, 2022. doi:10.1007/s10462-021-09996-w

    work page doi:10.1007/s10462-021-09996-w 2022

  9. [9]

    Introduction to Functional Differential Equations, volume 99 of Applied Mathematical Sciences

    Jack K Hale and Sjoerd M Verduyn Lunel. Introduction to Functional Differential Equations, volume 99 of Applied Mathematical Sciences. Springer, 1993

    1993

  10. [10]

    Theory and Applications of Hopf Bifurcation, volume 41 of London Mathematical Society Lecture Note Series

    Brian D Hassard, Nicholas D Kazarinoff, and Yieh-Hei Wan. Theory and Applications of Hopf Bifurcation, volume 41 of London Mathematical Society Lecture Note Series. Cambridge University Press, 1981

    1981

  11. [11]

    Evolutionary Games and Population Dynamics

    Josef Hofbauer and Karl Sigmund. Evolutionary Games and Population Dynamics. Cambridge University Press, 1998

    1998

  12. [12]

    1983 , issn =

    Paul W. Holland, Kathryn Blackmond Laskey, and Samuel Leinhardt. Stochastic blockmodels: First steps. Social Networks, 5 0 (2): 0 109--137, 1983. doi:10.1016/0378-8733(83)90021-7

    work page doi:10.1016/0378-8733(83)90021-7 1983

  13. [13]

    On delayed discrete evolutionary dynamics

    Ryota Iijima. On delayed discrete evolutionary dynamics. Journal of Theoretical Biology, 300: 0 1--6, 2012. doi:10.1016/j.jtbi.2012.01.001

    work page doi:10.1016/j.jtbi.2012.01.001 2012

  14. [14]

    Delay Differential Equations with Applications in Population Dynamics

    Yang Kuang. Delay Differential Equations with Applications in Population Dynamics. Academic Press, 1993

    1993

  15. [15]

    Erez Lieberman, Christoph Hauert, and Martin A. Nowak. Evolutionary dynamics on graphs. Nature, 433 0 (7023): 0 312--316, 2005. doi:10.1038/nature03204

    work page doi:10.1038/nature03204 2005

  16. [16]

    Fixation probabilities in evolutionary dynamics under weak selection

    Alex McAvoy and Benjamin Allen. Fixation probabilities in evolutionary dynamics under weak selection. Journal of Mathematical Biology, 82: 0 14, 2021. doi:10.1007/s00285-021-01568-4

    work page doi:10.1007/s00285-021-01568-4 2021

  17. [17]

    Evolutionary game theory and population dynamics

    Jacek Miekisz. Evolutionary game theory and population dynamics. In Multiscale Problems in the Life Sciences, volume 1940 of Lecture Notes in Mathematics, pages 269--316. Springer, 2008. doi:10.1007/978-3-540-78362-6_5

    work page doi:10.1007/978-3-540-78362-6_5 1940

  18. [18]

    Evolutionary dynamics of the delayed replicator--mutator equation: Limit cycle and cooperation

    Sourabh Mittal, Archan Mukhopadhyay, and Sagar Chakraborty. Evolutionary dynamics of the delayed replicator--mutator equation: Limit cycle and cooperation. Physical Review E, 101 0 (4): 0 042410, 2020. doi:10.1103/PhysRevE.101.042410

    work page doi:10.1103/physreve.101.042410 2020

  19. [19]

    Bifurcation analysis of replicator dynamics with logistic growth and strategy-dependent time delays in snowdrift game

    Javad Mohamadichamgavi and Marek Bodnar. Bifurcation analysis of replicator dynamics with logistic growth and strategy-dependent time delays in snowdrift game. Dynamic Games and Applications, 2025. doi:10.1007/s13235-025-00671-1

    work page doi:10.1007/s13235-025-00671-1 2025

  20. [20]

    Martin A. Nowak. Five rules for the evolution of cooperation. Science, 314 0 (5805): 0 1560--1563, 2006. doi:10.1126/science.1133755

    work page doi:10.1126/science.1133755 2006

  21. [21]

    Hisashi Ohtsuki, Christoph Hauert, Erez Lieberman, and Martin A. Nowak. A simple rule for the evolution of cooperation on graphs and social networks. Nature, 441 0 (7092): 0 502--505, 2006. doi:10.1038/nature04605

    work page doi:10.1038/nature04605 2006

  22. [22]

    Flor \'i a, and Yamir Moreno

    Matja z Perc, Jes \'u s G \'o mez-Garde \ n es, Attila Szolnoki, Luis M. Flor \'i a, and Yamir Moreno. Evolutionary dynamics of group interactions on structured populations: A review. Journal of the Royal Society Interface, 10 0 (80): 0 20120997, 2013. doi:10.1098/rsif.2012.0997

    work page doi:10.1098/rsif.2012.0997 2013

  23. [23]

    Jordan, David G

    Matja z Perc, Jillian J. Jordan, David G. Rand, Zhen Wang, Stefano Boccaletti, and Attila Szolnoki. Statistical physics of human cooperation. Physics Reports, 687: 0 1--51, 2017. doi:10.1016/j.physrep.2017.05.004

    work page doi:10.1016/j.physrep.2017.05.004 2017

  24. [24]

    Santos and Jorge M

    Francisco C. Santos and Jorge M. Pacheco. Scale-free networks provide a unifying framework for the emergence of cooperation. Physical Review Letters, 95 0 (9): 0 098104, 2005. doi:10.1103/PhysRevLett.95.098104

    work page doi:10.1103/physrevlett.95.098104 2005

  25. [25]

    Critical Transitions in Nature and Society

    Marten Scheffer. Critical Transitions in Nature and Society. Princeton University Press, 2009

    2009

  26. [26]

    Sutton and Andrew G

    Richard S. Sutton and Andrew G. Barto. Reinforcement Learning: An Introduction. MIT Press, 2018

    2018

  27. [27]

    Evolutionary games on graphs

    Gy \"o rgy Szab \'o and G \'a bor F \'a th. Evolutionary games on graphs. Physics Reports, 446 0 (4--6): 0 97--216, 2007. doi:10.1016/j.physrep.2007.04.004

    work page doi:10.1016/j.physrep.2007.04.004 2007

  28. [28]

    Taylor and Leo B

    Peter D. Taylor and Leo B. Jonker. Evolutionary stable strategies and game dynamics. Mathematical Biosciences, 40 0 (1--2): 0 145--156, 1978. doi:10.1016/0025-5564(78)90077-9

    work page doi:10.1016/0025-5564(78)90077-9 1978

  29. [29]

    Stochastic dynamics of invasion and fixation

    Arne Traulsen, Martin A Nowak, and Jorge M Pacheco. Stochastic dynamics of invasion and fixation. Physical Review E, 74 0 (1): 0 011909, 2006. doi:10.1103/PhysRevE.74.011909

    work page doi:10.1103/physreve.74.011909 2006

  30. [30]

    Hopf bifurcations in delayed rock--paper--scissors replicator dynamics

    Elizabeth Wesson and Richard Rand. Hopf bifurcations in delayed rock--paper--scissors replicator dynamics. Dynamic Games and Applications, 6 0 (1): 0 139--156, 2016. doi:10.1007/s13235-015-0138-2

    work page doi:10.1007/s13235-015-0138-2 2016

  31. [31]

    Rand, and David G

    Elizabeth Wesson, Richard H. Rand, and David G. Rand. Hopf bifurcations in two-strategy delayed replicator dynamics. International Journal of Bifurcation and Chaos, 26 0 (1): 0 1650006, 2016. doi:10.1142/S0218127416500061

    work page doi:10.1142/s0218127416500061 2016

  32. [32]

    Wettergren

    Thomas A. Wettergren. Replicator dynamics of evolutionary games with different delays on costs and benefits. Applied Mathematics and Computation, 458: 0 128228, 2023. doi:10.1016/j.amc.2023.128228

    work page doi:10.1016/j.amc.2023.128228 2023

  33. [33]

    Cooperator driven oscillation in a time-delayed feedback-evolving game

    Fang Yan, Xiaojie Chen, Zhipeng Qiu, and Attila Szolnoki. Cooperator driven oscillation in a time-delayed feedback-evolving game. New Journal of Physics, 23: 0 053017, 2021. doi:10.1088/1367-2630/abf205

    work page doi:10.1088/1367-2630/abf205 2021

  34. [34]

    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. In Handbook of Reinforcement Learning and Control, Studies in Systems, Decision and Control, pages 321--384. Springer, 2021

    2021

  35. [35]

    Multi-agent reinforcement learning with reward delays

    Yuyang Zhang, Runyu Zhang, Yuantao Gu, and Na Li. Multi-agent reinforcement learning with reward delays. In Proceedings of the 5th Annual Learning for Dynamics and Control Conference (L4DC), volume 211 of PMLR, pages 692--704, 2023

    2023