Pith. sign in

REVIEW 3 major objections 6 minor 38 references

Why Study Emergent Behavior When You Can Regulate It? Aligning Multi-Agent Systems with Reward Prediction

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read MARP learns a shared reward model from episode-level preferences over collective outcomes and uses it to align decentralized agents with social objectives such as efficiency, equality, sustainability, and peace.

desk verdict A genuine but single-environment proof of concept for preference-based social reward modeling; the baseline comparison is too thin to support the headline claim as written. read the letter →

arxiv 2608.07280 v1 pith:SFNUCTTN submitted 2026-08-07 cs.MA

Assaf Caftory , Almog Zemach , Moshe Butman , Doron Friedman This is my paper · ORCID
classification cs.MA
keywords multi-agentrewardpredictionpreference-basedlearningsequentialsocialdilemmasmetricsemergentbehaviorregulationHarvestGamedecentralizedMARLmodeling
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

This paper argues that emergent collective behavior in multi-agent systems need not be left to unfold; it can be regulated by learning what to reward. The authors introduce Multi-Agent Reward Prediction (MARP), which trains a shared reward model from episode-level judgments about collective outcomes, then uses that model to give each decentralized agent a dense per-step reward signal. In the Harvest Game, a common-pool resource dilemma, MARP agents move social metrics in the intended direction for efficiency, equality, sustainability, and peace, and can handle composite objectives by changing only the evaluation metric. Because the empirical validation is confined to one environment, the paper presents MARP as a proof of concept rather than a fully general alignment mechanism.

What carries the argument

The load-bearing object is the shared reward model $\hat r_\theta$, trained with a Bradley–Terry preference loss over pairwise comparisons of sequences. A preference oracle ranks full episodes by a social metric, and the model is trained so that the sum of its per-step predicted rewards preserves that ranking. MARP offers two inference strategies: Joint-Episode Inference concatenates all agent trajectories from an episode into one sequence, while Local-Trajectory Inference trains on individual egocentric trajectories but inherits the episode-level ordering. The predicted rewards then serve as the RL reward for each agent's independent PPO policy, bridging global, episodic evaluation and local, step-wise control.

What would settle it

On a held-out set of episodes, compute the rank correlation between MARP's summed predicted rewards and the target social metric; a near-zero correlation would show the learned reward model does not actually capture the global objective and would falsify the alignment claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single learning framework can replace handcrafted social rewards: episodic preferences over collective outcomes are converted into local reward predictions that guide decentralized agents toward specified social objectives. In the Harvest Game, both MARP variants—Joint-Episode, which conditions on the full multi-agent episode, and Local-Trajectory, which learns from individual egocentric trajectories—escape the tragedy-of-the-commons equilibrium that on-policy PPO reaches with the environment reward, and they outperform that baseline on the target metrics. The framework also supports multiple and composite objectives, such as efficiency combined with equality or peace, by changing only the high-level metric used by the preference oracle. The learned reward model captures incidental environmental structure, assigning lower predicted reward to consuming apples when local apple density is low, which amounts to an emergent norm of restraint under scarcity.

Load-bearing premise

The load-bearing premise is that the four scalar social metrics—efficiency, Gini equality, reward timing for sustainability, and untagged time for peace—faithfully represent the objectives they name, because the preference oracle is built from those same metrics; if a metric misrepresents the goal, MARP aligns agents to the metric instead.

Editorial extensions

If this is right

  • If the central claim is right, social objectives can be swapped by changing only the evaluation metric, so one training setup regulates multiple goals without re-engineering rewards.
  • Combining objectives, such as efficiency with equality or peace, raises the secondary metric without degrading efficiency, so composite norms are within reach of the same framework.
  • The learned reward model encodes environmental structure, such as assigning lower reward to consuming apples under local scarcity, producing an implicit norm of restraint without explicit programming.
  • On-policy PPO trained on the environment reward collapses into the tragic equilibrium, while MARP escapes it, suggesting that reward prediction changes the learning dynamics and not just the reward values.
  • Because the validation is limited to a single environment, the method should be read as a proof of concept until reproduced in other multi-agent settings.

Reading between the lines

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

  • A sharper comparison would add experience replay to the PPO baseline; if replay closes the gap, the headline result may be about exploration rather than reward modeling itself.
  • Since the preference oracle is built from the same scalar metrics used for evaluation, MARP could silently optimize the metric while missing the underlying objective; metric choice is therefore part of the design, not a neutral input.
  • The per-step reward predictions could be reused as an interpretability tool to inspect which local behaviors the framework treats as prosocial.
  • Heterogeneous agents or stochastic human feedback would stress the shared-reward assumption and bound the framework's scope; failure there would not refute the Harvest proof of concept but would limit its reach.
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 / 6 minor

Summary. The paper introduces Multi-Agent Reward Prediction (MARP), a framework that trains a shared reward model from episode-level preferences over social metrics and uses it to provide dense reward signals to decentralized agents in a sequential social dilemma. Two inference strategies are proposed: Local-Trajectory Inference, which assigns each episode-level preference to every individual trajectory, and Joint-Episode Inference, which trains on aggregated episode representations. Experiments in the Harvest Game (Commons Game) show that both variants can steer behavior toward efficiency, equality, sustainability, and peace, and toward composite objectives, outperforming a PPO baseline trained on the raw environment reward. The paper explicitly frames the contribution as a proof of concept in a single environment and acknowledges limitations in a dedicated section.

Significance. The central idea is timely and relevant: treating emergent multi-agent behavior as a target of learned regulation, rather than only as an object of analysis, addresses a real gap between global normative objectives and local decentralized control. The paper ships code, reports error bars over five runs, and includes a useful interpretability analysis showing that the learned reward assigns lower value to consuming apples under local scarcity. As a proof of concept in a single environment, the contribution is valuable, but the empirical claims are currently broader than the evidence supports, and the experimental design does not fully isolate the mechanism from the metric supervision.

major comments (3)
  1. [§5.1 and Abstract] The headline claim that MARP produces behavior "more closely aligned with target social metrics than standard reward-based baselines" is supported only by comparison to PPO trained on the raw environment reward. No handcrafted social-reward baseline is tested, despite the paper citing inequity aversion (Hughes et al., 2018) and social influence bonuses (Jaques et al., 2019) in Section 2. Because the preference oracle defined in Section 4.1 is built from the same social metrics used for evaluation in Section 5.2, the comparison does not isolate whether preference-based credit assignment, rather than direct access to metric supervision, drives the result. The paper should either add at least one reward-based baseline that encodes a social objective, or explicitly restrict the claim to "compared to PPO with environment reward only."
  2. [§4.1, §5.2, §6.1] There is a circularity between the preference oracle and the evaluation metrics: the oracle ranks episodes using the same scalar social metrics (Eqs. 1-5) that later measure success. Training the reward model to reproduce these rankings and then showing that agents optimized on the learned reward improve those same metrics is partly a consequence of fitting the model to the evaluation metric. The paper acknowledges in Section 6.1 that "alignment depends on predefined social metrics," but does not address the experimental circularity. A more convincing test would evaluate MARP on a metric not used to generate preferences, or with a noisy human-like oracle, to show that the mechanism, rather than the metric itself, generalizes.
  3. [§5.2 and Appendix B, Figures 10 and 11] The claim that "for all methods and all metric pairs, the dual-objective setting yields a higher trend for the secondary metric, without degrading efficiency" is based on visual inspection of error bars over five runs; no significance tests are reported for the social metrics. Given the small number of runs and the load-bearing nature of the multi-objective claim, the paper should either report statistical tests (e.g., bootstrap or paired tests across runs) or present these as qualitative observations rather than as a consistent result.
minor comments (6)
  1. [Abstract and §5.1] The text uses the plural "standard reward-based baselines," but only a single baseline (PPO with environment reward) is evaluated; the plural should either be replaced with "baseline" or additional baselines should be added.
  2. [§5.1] There is a typo in "two-way repeated-measures ANOV A" (extra space before "A"); please correct to "ANOVA".
  3. [References] The reference list for Hughes et al. (2018) contains encoding artifacts in the author names ("Du ´e~nez-Guzm\'an"); please clean up the citation encoding.
  4. [§4.1, Definition 4] The trajectory-level oracle assigns every trajectory in the winning episode a preference over every trajectory in the losing episode; this strong assumption is acknowledged in Section 6.1 but should be stated more prominently in the method section when the oracle is defined.
  5. [Experimental Setup] The paper does not explicitly state the number of agents and episode length used in the experiments, although these parameters are central to the social metrics in Eqs. (1)-(5); please specify them in Section 5 or Appendix A.
  6. [Appendix B, Figure 8] Figure 8 compares DQN with environment rewards to MARP for a single run; the caption should clarify that this is an illustrative reproduction of prior work, not a systematic comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MARP's metric-based oracle is the stated objective, and the learned local reward model's ability to steer decentralized behavior is an independent empirical result.

full rationale

The only potentially closed-loop element is that the preference oracle (Definition 4, Section 4.1) ranks episodes by the same social metrics (Eqs. 1-5) used for evaluation (Section 5.2). This is not a hidden circularity: the social metric is the stated objective, and MARP is an optimization procedure for that objective. The chain from metric to preference labels to the learned reward model to decentralized policies and back to the metric is not an equality by construction. The reward model is a separately parameterized local function trained with a Bradley-Terry loss (Eqs. 6-8); whether policies maximizing the summed predicted rewards actually increase the metric is an empirical fact demonstrated in Figures 5-6 and Appendix B, and it could have failed. The paper also tests the same framework across multiple objectives (efficiency, equality, sustainability, peace, and composites), so the central claim has independent content beyond a single fitted value. Section 6.1 honestly discloses the dependence on predefined metrics and the lack of identifiability of the per-step decomposition; these are limitations, not circular reductions. No load-bearing self-citation, imported uniqueness theorem, or ansatz-smuggled-via-citation is present. The weak baseline comparison (PPO with environment reward only) is an experimental-design concern about whether the advantage comes from metric supervision, but it does not make the derivation circular under the quoted-equation standard applied here.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the choice of social metrics, the Bradley-Terry preference model, the deterministic metric oracle, the IPPO learner, and the trajectory-level label inheritance. None of these is derived from first principles; each is either standard practice or a stated simplifying assumption.

free parameters (1)
  • Reward model parameters theta = Learned from preference buffer B via Eq. (8)
    These fitted parameters are the mechanism by which episode-level preferences become local rewards; the claim that MARP aligns behavior depends on this learned function, not on a hand-derived reward.
assumptions (5)
  • standard math Bradley-Terry preference model converts reward sums into preference probabilities (Eq. 7)
    Standard in preference-based RL (Christiano et al., 2017); used without modification.
  • domain assumption The social metrics in Eqs. (1)-(5) faithfully represent efficiency, equality, sustainability, and peace
    The preference oracle and the evaluation are both built from these scalar metrics; if they mis-specify the objectives, MARP aligns to the wrong target. Acknowledged in Section 6.1.
  • domain assumption Deterministic metric-based preference oracle is a valid stand-in for human feedback
    Real human preferences are noisy and context-dependent; the paper uses exact metric comparisons and flags this as a simplification in Section 6.1.
  • domain assumption Independent PPO reliably optimizes the learned reward
    The paper relies on IPPO without convergence guarantees; the behavioral results depend on this choice of learner.
  • ad hoc to paper Local-Trajectory oracle assigns each episode-level preference to every trajectory in the episode
    This is a coarse credit-assignment assumption specific to MARP's Local-Trajectory variant; the paper acknowledges the resulting noisy training signal in Section 6.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Why Study Emergent Behavior When You Can Regulate It? Aligning Multi-Agent Systems with Reward Prediction." pith.science (2026). https://pith.science/paper/SFNUCTTN

@misc{pith2026260807280,
  author       = {Pith},
  title        = {Pith review of: Why Study Emergent Behavior When You Can Regulate It? Aligning Multi-Agent Systems with Reward Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SFNUCTTN}},
  note         = {Machine review of arXiv:2608.07280}
}
read the original abstract

Multi-agent simulations are widely used to study complex social and ecological systems, where rich and often unexpected emergent behaviors arise from local interactions. A large body of prior work has focused on analyzing such emergent dynamics across domains. In this paper, we move beyond analyzing emergent behavior and introduce a learning-based mechanism for actively shaping it via social reward modeling. We introduce Multi-Agent Reward Prediction (MARP), a simple framework that extends preference-based reward modeling to multi-agent reinforcement learning. While the framework is designed to be applicable across multi-agent settings, the present empirical validation is limited to a single environment, and we therefore present MARP as a proof of concept within the studied domain. Rather than relying on handcrafted rewards, MARP learns a shared reward model from episode-level evaluations of collective outcomes, enabling decentralized agents to align their behavior with global social objectives. We study MARP in the Harvest Game, a canonical sequential social dilemma modeling common-pool resource management and related real-world challenges. Our results show that MARP can be tuned to produce behavior that is more closely aligned with target social metrics than standard reward-based baselines, while the learned reward model captures subtle environmental structure without explicit programming. Crucially, MARP supports multiple and composite social objectives within a single training regime. By modifying only the high-level evaluation metric, the same framework seamlessly aligns agent behavior with diverse goals, including sustainability, equality, and peace, as well as combinations of individual and group-level objectives. These findings demonstrate that emergent multi-agent behavior can be treated not only as a phenomenon to study, but as a target of principled, data-driven regulation.

Figures

Figures reproduced from arXiv: 2608.07280 by the authors.

Figure 1
Figure 1. (Left) Full environment state showing green apples, [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Toy illustration of MARP credit assignment. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Centralized Multi-Agent Reward Prediction (MARP) framework architecture. Multiple agents interact with the simu [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Reward prediction using the Joint-Episode method. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Equality under MARP for efficiency-only and [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 5
Figure 5. Figure 5: Social metrics under Efficiency-only optimization [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Reward model architecture. Observations are processed through a CNN-based feature extractor and flattened, while [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Comparison between DQN with environment [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Reward prediction using the Local-Trajectory [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: For each learning objective, we report all social metrics, comparing the two MARP methods. Results are based on [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: For each social metric, we compare the two methods under two objectives: optimizing Efficiency alone and optimiz [PITH_FULL_IMAGE:figures/full_fig_p014_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 27 canonical work pages

  1. [1]

    Advances in neural information processing systems , volume=

    A multi-agent reinforcement learning model of common-pool resource appropriation , author=. Advances in neural information processing systems , volume=

    work page

  2. [3]

    2018 , eprint=

    Inequity aversion improves cooperation in intertemporal social dilemmas , author=. 2018 , eprint=

    work page 2018

  3. [4]

    Advances in neural information processing systems , volume=

    Deep reinforcement learning from human preferences , author=. Advances in neural information processing systems , volume=

  4. [5]

    International conference on machine learning , pages=

    Social influence as intrinsic motivation for multi-agent deep reinforcement learning , author=. International conference on machine learning , pages=. 2019 , organization=

    work page 2019

  5. [6]

    Advances in neural information processing systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in neural information processing systems , volume=

  6. [8]

    Advances in Neural Information Processing Systems , volume=

    Learning to summarize with human feedback , author=. Advances in Neural Information Processing Systems , volume=

  7. [10]

    Artificial Intelligence Review , volume=

    Multi-agent deep reinforcement learning: a survey , author=. Artificial Intelligence Review , volume=. 2022 , publisher=

    2022

  8. [11]

    Journal of Economic Literature , volume=

    Agent-based modeling in economics and finance: Past, present, and future , author=. Journal of Economic Literature , volume=. 2025 , publisher=

    work page 2025

Show all 38 references
  1. [12]

    PLOS Computational Biology , volume=

    Covasim: an agent-based model of COVID-19 dynamics and interventions , author=. PLOS Computational Biology , volume=. 2021 , publisher=

    2021

  2. [13]

    Simulating social phenomena , pages=

    Advancing the art of simulation in the social sciences , author=. Simulating social phenomena , pages=. 1997 , publisher=

    1997

  3. [14]

    2019 , publisher =

    Vinitsky, Eugene and Jaques, Natasha and Leibo, Joel and Castenada, Antonio and Hughes, Edward , title =. 2019 , publisher =

    2019

  4. [15]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Decoding global preferences: Temporal and cooperative dependency modeling in multi-agent preference-based reinforcement learning , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  5. [16]

    Science , volume=

    Exposure to ideologically diverse news and opinion on Facebook , author=. Science , volume=

  6. [17]

    Colorado Technology Law Journal , volume=

    Algorithmic harms beyond Facebook and Google: Emergent challenges of computational agency , author=. Colorado Technology Law Journal , volume=

  7. [18]

    Social theory re-wired , pages=

    The age of surveillance capitalism , author=. Social theory re-wired , pages=. 2023 , publisher=

    2023

  8. [19]

    Building a foundation for data-driven, interpretable, and robust policy design using the

    Trott, Alexander and Srinivasa, Sunil and van der Wal, Douwe and Haneuse, Sebastien and Zheng, Stephan , journal=. Building a foundation for data-driven, interpretable, and robust policy design using the

  9. [20]

    and Socher, Richard , journal =

    Zheng, Stephan and Trott, Alexander and Srinivasa, Sunil and Parkes, David C. and Socher, Richard , journal =. The. 2022 , volume =. doi:10.1126/sciadv.abk2607 , url =

    2022 doi

  10. [22]

    Advancing the art of simulation in the social sciences

    Robert Axelrod. Advancing the art of simulation in the social sciences. In Simulating social phenomena , pages 21--40. Springer, 1997

    1997

  11. [23]

    Agent-based modeling in economics and finance: Past, present, and future

    Robert L Axtell and J Doyne Farmer. Agent-based modeling in economics and finance: Past, present, and future. Journal of Economic Literature , 63(1):197--287, 2025

    2025

  12. [24]

    Exposure to ideologically diverse news and opinion on facebook

    Eytan Bakshy, Solomon Messing, and Lada A Adamic. Exposure to ideologically diverse news and opinion on facebook. Science , 348(6239):1130--1132, 2015

    2015

  13. [25]

    Deep reinforcement learning from human preferences

    Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learning from human preferences. Advances in neural information processing systems , 30, 2017

    2017

  14. [26]

    Multi-agent deep reinforcement learning: a survey

    Sven Gronauer and Klaus Diepold. Multi-agent deep reinforcement learning: a survey. Artificial Intelligence Review , 55(2):895--943, 2022

    2022

  15. [27]

    Leibo, Matthew G

    Edward Hughes, Joel Z. Leibo, Matthew G. Phillips, Karl Tuyls, Edgar A. Du \'e \ n ez-Guzm \'a n, Antonio Garc \'i a Casta \ n eda, Iain Dunning, Tina Zhu, Kevin R. McKee, Raphael Koster, Heather Roff, and Thore Graepel. Inequity aversion improves cooperation in intertemporal ...

    2018

  16. [28]

    Social influence as intrinsic motivation for multi-agent deep reinforcement learning

    Natasha Jaques, Angeliki Lazaridou, Edward Hughes, Caglar Gulcehre, Pedro Ortega, DJ Strouse, Joel Z Leibo, and Nando De Freitas. Social influence as intrinsic motivation for multi-agent deep reinforcement learning. In International conference on machine learning , pages 3040-...

    2019

  17. [29]

    Covasim: an agent-based model of covid-19 dynamics and interventions

    Cliff C Kerr, Robyn M Stuart, Dina Mistry, Romesh G Abeysuriya, Katherine Rosenfeld, Gregory R Hart, Rafael C N \'u \ n ez, Jamie A Cohen, Prashanth Selvaraj, Brittany Hagedorn, et al. Covasim: an agent-based model of covid-19 dynamics and interventions. PLOS Computational Bio...

    2021

  18. [30]

    Multi-agent reinforcement learning in sequential social dilemmas

    Joel Z Leibo, Vinicius Zambaldi, Marc Lanctot, Janusz Marecki, and Thore Graepel. Multi-agent reinforcement learning in sequential social dilemmas. arXiv preprint arXiv:1702.03037 , 2017

    2017 arXiv

  19. [31]

    Scalable agent alignment via reward modeling: a research direction

    Jan Leike, David Krueger, Tom Everitt, Miljan Martic, Vishal Maini, and Shane Legg. Scalable agent alignment via reward modeling: a research direction. arXiv preprint arXiv:1811.07871 , 2018

    2018 arXiv

  20. [32]

    The alignment problem from a deep learning perspective

    Richard Ngo, Lawrence Chan, and S \"o ren Mindermann. The alignment problem from a deep learning perspective. arXiv preprint arXiv:2209.00626 , 2022

    2022 arXiv

  21. [33]

    Training language models to follow instructions with human feedback

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. Advances in neural information processing systems , 35:277...

    2022

  22. [34]

    A multi-agent reinforcement learning model of common-pool resource appropriation

    Julien Perolat, Joel Z Leibo, Vinicius Zambaldi, Charles Beattie, Karl Tuyls, and Thore Graepel. A multi-agent reinforcement learning model of common-pool resource appropriation. Advances in neural information processing systems , 30, 2017

    2017

  23. [35]

    Learning to summarize with human feedback

    Nisan Stiennon, Long Ouyang, Jeffrey Wu, Daniel Ziegler, Ryan Lowe, Chelsea Voss, Alec Radford, Dario Amodei, and Paul F Christiano. Learning to summarize with human feedback. Advances in Neural Information Processing Systems , 33:3008--3021, 2020

    2020

  24. [36]

    Building a foundation for data-driven, interpretable, and robust policy design using the AI E conomist

    Alexander Trott, Sunil Srinivasa, Douwe van der Wal, Sebastien Haneuse, and Stephan Zheng. Building a foundation for data-driven, interpretable, and robust policy design using the AI E conomist. arXiv preprint arXiv:2108.02904 , 2021

    2021 arXiv

  25. [37]

    Algorithmic harms beyond facebook and google: Emergent challenges of computational agency

    Zeynep Tufekci. Algorithmic harms beyond facebook and google: Emergent challenges of computational agency. Colorado Technology Law Journal , 13:203--218, 2015

    2015

  26. [38]

    An open source implementation of sequential social dilemma games

    Eugene Vinitsky, Natasha Jaques, Joel Leibo, Antonio Castenada, and Edward Hughes. An open source implementation of sequential social dilemma games. https://github.com/eugenevinitsky/sequential_social_dilemma_games/issues/182, 2019. GitHub repository

    2019

  27. [39]

    On the importance of replay for policy optimization in sparse-reward and discrete action spaces

    Yifan Wang, Yuchen Zhang, Ziming Xu, and Shuo Chen. On the importance of replay for policy optimization in sparse-reward and discrete action spaces. arXiv preprint arXiv:2405.16383 , 2024

    2024 arXiv

  28. [40]

    Parkes, and Richard Socher

    Stephan Zheng, Alexander Trott, Sunil Srinivasa, David C. Parkes, and Richard Socher. The AI E conomist: Taxation policy design via two-level deep multiagent reinforcement learning. Science Advances , 8(18):eabk2607, 2022

    2022

  29. [41]

    Decoding global preferences: Temporal and cooperative dependency modeling in multi-agent preference-based reinforcement learning

    Tianchen Zhu, Yue Qiu, Haoyi Zhou, and Jianxin Li. Decoding global preferences: Temporal and cooperative dependency modeling in multi-agent preference-based reinforcement learning. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 38, pages 17202--17210, 2024

    2024

  30. [42]

    The age of surveillance capitalism

    Shoshana Zuboff. The age of surveillance capitalism. In Social theory re-wired , pages 203--213. Routledge, 2023

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.