Pith. sign in

REVIEW 4 major objections 4 minor 42 references

Group-Agent Reinforcement Learning with Heterogeneous Agents

T0 review · 4 major / 4 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 Heterogeneous RL agents that share action choices and adopt each other's models learn dramatically faster than solo agents, with 96% of 129 examined Atari learners seeing a speed-up.

desk verdict Real extension of GARL to heterogeneous agents, but the headline speed-up numbers describe an oracle over three rules, not a single deployable method. read the letter →

arxiv 2501.11818 v2 pith:ARVVR7Z3 submitted 2025-01-21 cs.LG

Kaiyue Wu , Xiao-Jun Zeng , Tingting Mu This is my paper · ORCID
classification cs.LG
keywords heterogeneousagentsgroup-agentreinforcementlearningactionselectionrulesmodeladoptionknowledgesharingAtari2600sampleefficiencyactor-criticalgorithms
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 advances group-agent reinforcement learning to a heterogeneous setting: each agent in a group runs a different actor-critic algorithm (A2C, PPO, or ACER), acts in its own copy of the environment, and asynchronously exchanges two kinds of knowledge—policy/value model parameters and the accumulated reward score per episode. The claim is that using this shared knowledge, each individual agent learns faster and often reaches a higher final reward than if it trained alone. On 43 Atari 2600 games this yields, across 129 agents, a learning speed-up for 96% of them, with 72% learning 100 times faster and about 41% reaching a higher reward in less than 5% of the time steps a single agent needs. The reason this would matter is that it offers a simple, algorithm-agnostic way to accelerate RL training in settings where several agents independently solve similar tasks.

What carries the argument

The load-bearing mechanism is the combination of three action-selection rules and a model-adoption step operating under a shared-interface assumption. The rules—probability addition (sum of the group's action probabilities), probability multiplication (product of those probabilities), and the Combo rule (which restricts attention to agents with higher accumulated reward, picks the one with the highest accumulated next-state value, and applies a negative-log-likelihood confidence threshold)—turn a set of peer-suggested actions into one action per step. Model adoption, applied only with the Combo rule, replaces an agent's own policy and value networks with a peer's when that peer's policy dominated the agent's last batch of choices and holds the highest accumulated reward. The enabling assumption that backs the entire scheme is that all agents in a group share the same state space, action space, and network architecture, so parameters can be interchanged freely.

What would settle it

Recompute the speed-up statistics from the paper's Table 6 using one fixed selection rule for all agents (say Combo only) instead of the best of the three rules per game; if the 96% and 72% headline percentages fall sharply, the central quantitative claim is an artifact of per-game rule selection.

Watch

Extended reading notes

Core claim

The central discovery is a set of group-learning mechanisms that let heterogeneous RL agents improve one another without synchronizing. At each time step an agent collects the action suggestions of all group members, then selects one action using either probability addition, probability multiplication, or the Combo rule (which filters suggestions by accumulated reward, next-state value, and action confidence). When Combo is used, an agent may also adopt the entire policy and value model of a peer whose suggestions it has been following for at least half of a training batch and who has the highest accumulated reward. The paper reports that on 43 Atari games the resulting HGARL agents reach their best scores in a fraction of the steps a solo agent needs: 96% of the 129 examined agents achieve a speed-up, 72% learn at least 100 times faster, and about 41% achieve a higher reward score within 5% of the time steps required by single-agent learning.

Load-bearing premise

The whole sharing scheme rests on every agent in the group using the same state space, action space, and policy-network architecture; if a real group mixes architectures or action spaces, the action-selection and model-adoption steps can no longer be executed.

Editorial extensions

If this is right

  • A2C, PPO, and ACER agents trained as a group can match or beat their solo performance using a small slice of the environment steps, with 41% of tests reaching a higher final reward in at most 5% of the steps.
  • The speed-up is not limited to one rule: the best rule varies by game, and the reported 96% figure is the best-of-three per game, so a practical deployment needs a way to pick or switch rules.
  • Model adoption accounts for the extreme speed-ups, including the 'inf' entries where solo agents never reach the group performance; without adoption, only action selection remains.
  • The group protocol is asynchronous and each agent keeps its own environment, so it is compatible with decentralized training workloads.

Reading between the lines

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

  • Beyond the paper: because the headline figures choose the best rule per game, a fixed-rule deployment would likely show lower average speed-up; a natural extension is an online rule-selection meta-policy that picks between PA, PM, and Combo from observed group performance.
  • Beyond the paper: the shared-architecture assumption means 'heterogeneous' is limited to algorithm differences; a variant that shares only action advice, without model parameters, could test whether the speed-up comes mainly from better action choices or from parameter adoption.
  • Beyond the paper: the Combo rule's threshold is tuned once across games; one could expect per-agent or per-game threshold adaptation to improve the conservative-filtering failures seen on games like Alien.
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

4 major / 4 minor

Summary. The paper introduces HGARL, a framework for asynchronous knowledge sharing among reinforcement learning agents that run different algorithms (A2C, PPO, and ACER) in separate environments. Knowledge is shared as action suggestions from peer policies and as full policy/value model parameters. Three action-selection rules are defined (probability addition, probability multiplication, and reward-value-likelihood combination, called Combo), along with a model-adoption rule. Experiments on 43 Atari games are used to claim that 96% of the 129 agents achieve a learning speed-up, 72% learn over 100 times faster, and 41% reach a higher reward within 5% of the time steps needed by a single agent. The central quantitative claims are computed by selecting the best of the three rules per agent and game, by counting 'inf' speed-ups as successes, and by fine-tuning the Combo threshold on the test games.

Significance. If validated, HGARL would be a practically relevant contribution: it would show that agents running different RL algorithms can improve one another's sample efficiency by sharing policies and models asynchronously without a shared environment. The experimental scale is substantial (43 games, three agents, four seeds, over 10M core hours), and the idea of combining heterogeneous actor-critic algorithms through action-selection and model-adoption mechanisms is plausible. However, the paper does not provide code, proofs, or a parameter-free derivation; the contribution is empirical. As reported, the evaluation does not establish the headline claims because of the issues detailed in the major comments. The underlying direction is worth pursuing, but the current evidence is insufficient for the strength of the claims.

major comments (4)
  1. [§4.2, Tables 1 and 2] The notes to Table 1 and Table 2 state that for each agent type and game the authors report the performance of the rule that gives the best performance among PA, PM, and Combo. Consequently, the headline numbers (96% speed-up, 72% >100x, 41% reward within 5% time) are maxima over three rules, and no single rule or concrete algorithm is evaluated. Since the paper does not supply a rule-selection mechanism for a new game, these percentages describe an oracle rather than a deployable method. Please report per-rule statistics separately, and if a combined method is claimed, define how the rule is chosen without access to the test game.
  2. [§4.2, Table 2 and Table 6] The 'inf' entries in Table 6 mean that the single agent never reached the group agent's highest performance within the finite training budget. In Table 2 these are counted in the 'inf' bin and treated as speed-ups greater than 100x. This is a censored observation, not an infinite speed-up. For instance, the ACER column in Table 2 contains 29 of 43 'inf' entries, which heavily inflates the aggregate statistics. The authors should analyze performance at a fixed time budget (for example, comparing final scores at equal numbers of steps) and report cases where the single agent never reached the group agent's score as 'not reached', not as >1000x or infinite.
  3. [§4.1 and Appendix A] The threshold phi in Eq. (8) is described as 'fine-tuned' on the test games in Appendix A. This is test-set selection, so the Combo-rule results cannot be interpreted as a validation of the method's generalization. The authors should either select phi on a validation split before evaluating on the test games, or provide a sensitivity analysis over phi values chosen a priori and show that the reported improvements are not artifacts of a particular tuned threshold.
  4. [§4.2, Table 1, Table 2, and Table 6] The treatment of 'nan' values in Table 6 is not explained in the statistics. The paper notes that 'nan values mean the group agents never reached a satisfactory level of performance', but it does not specify whether these are counted as failures, excluded from the percentages, or assigned to one of the speed-up bins. Since the totals in Table 2 sum to 43 for each agent, at least some 'nan' entries must have been omitted or absorbed into an existing bin; this handling should be stated explicitly and justified.
minor comments (4)
  1. [§3.2, Eq. (8)] The symbol phi is introduced in Eq. (8) without stating that it is a threshold on the negative log likelihood; the definition appears only in the following paragraph. Please define the threshold at its first use.
  2. [Figure 1 and Algorithm 1] In Figure 1, the label 'aStep' appears to be a fragment; it should probably be 'Action Step' or 'Apply'. Also, Algorithm 1 uses 'AN D' with a space in lines 14-16, which should be 'AND'.
  3. [§3, introductory paragraph] The assumption of identical state and action spaces and identical policy/value architectures is stated in Section 3, but the title and abstract do not qualify the meaning of 'heterogeneous agents' beyond algorithmic differences. Please scope the terminology throughout the paper to avoid overclaiming generality.
  4. [§4.1, experiment setting] The paper states that the experiments consumed over 10M core hours but does not provide a precise breakdown per run. A short table listing the number of runs, environment steps per run, and hardware details would improve transparency and reproducibility.

Circularity Check

3 steps flagged · score 6.0 of 10

Headline speed-up and reward statistics are constructed by per-game best-of-three rule selection, a test-set-tuned Combo threshold, and coding 'never reached' as infinite speed-up; the central quantitative claims are therefore not those of any single fixed method.

  1. fitted input called prediction [Section 4.2, Table 1 note (repeated in Table 2 note)]
    "Note that for each agent type and test case (game), we report the performance of the rule that gives the best performance among the three rules. This is because the three rules are supposed to show performance strengths for different test cases, and we're happy as long as there is at least one rule that works well for each case."

    The headline statistics (96.12% speed-up, 41.09% improved reward) are computed from this per-game maximum over PA, PM and Combo. No rule-selection mechanism is defined for a new game, so the reported improvement is not the performance of any single fixed HGARL algorithm; it is an oracle bound obtained by choosing, after seeing the outcome, whichever of the three rules wins on that game. The claimed success rate is therefore true by construction for the reported test set but does not predict performance for an unseen deployment, making the fitted per-game rule choice stand in for a predictive claim.

  2. fitted input called prediction [Section 4.1 and Appendix A]
    "The thresholdϕ for the negative log likelihood is set as 80% of the initial negative log likelihood, which is fine-tuned. ... we then try a reasonably large likelihood threshold ϕ2, and gradually narrow the interval [ϕ1, ϕ2] till locating a best likelihood threshold ϕ in the middle."

    The only hyperparameter of the Combo rule is selected by searching on the same 43 Atari games used for the reported evaluation, and no separate validation or test split is described. The Combo-rule entries in Tables 3-6 and the conclusion that 'the Combo rule shows superb performance' are therefore produced using test-set information; the evaluation is not an independent test of a fixed rule. This is a fitted parameter being reported as an achieved result on the data that selected it.

1 more flagged steps
  1. self definitional [Section 4.2, Table 2 and Table 6 note]
    "The inf value means that the single agents were never able to reach the same performance."

    Speed-up is defined as r = T/TG, where T is the number of steps the single agent takes to reach the group agent's performance. For the entries coded 'inf', T is not a measured finite number because the single agent never reached that performance within the training budget. Counting these censored observations as speed-ups, and aggregating them into the headline that 72% of agents learn over 100 times faster, makes that fraction partly an artifact of the metric's definition: a non-achievement is encoded as an infinite positive ratio. The claimed speed-up distribution is thus not fully measured from observed times.

full rationale

This is an empirical paper with no formal derivation chain, so the circularity analysis concerns whether the reported quantitative claims are self-contained tests of a fixed method. The central claims reduce to three construction choices. First, every per-game success is the maximum over the three proposed rules, without any specified way to choose a rule for a new game; the 96% and 41% figures therefore describe an oracle, not a deployable algorithm. Second, the Combo rule's threshold phi is fine-tuned on the same 43 test games, so the Combo results include test-set selection. Third, 'inf' speed-ups are censored cases in which the single agent never matched the group agent within budget; treating them as infinite speed-ups inflates the 'over 100 times faster' statistic by definition. None of these requires speculation about intent; each is directly quoted from the paper's tables and appendix. The self-citation to Wu and Zeng [2023] for the GARL concept is not load-bearing: the novel heterogeneous mechanism is evaluated independently of that citation, and there is no imported uniqueness theorem. Some individual fixed-rule results in the raw tables (e.g., Combo on Breakout or RoadRunner) do show large genuine improvements, so the method is not wholly vacuous; but the headline numbers as stated are not established for any single fixed rule with a fixed threshold. Score 6 reflects that the paper's central quantitative claim reduces by construction to per-game and per-threshold selection on the test data, while some independent empirical content remains.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a few domain assumptions about shared architecture and reward as a quality signal, plus two tuned or selected parameters (the Combo threshold and the post hoc rule choice). No new physical or mathematical entities are introduced.

free parameters (2)
  • Combo rule threshold phi = 80% of initial negative log likelihood
    Sets the confidence cutoff for accepting another agent's action; reported as 'fine-tuned' on the test games (Section 4.1, Appendix A).
  • Best rule per game selection = PA, PM, or Combo chosen post hoc per game
    Headline statistics are computed from the best-performing rule among the three per agent and game, so the reported method is not fixed a priori (Table 1 and Table 2 captions).
assumptions (4)
  • domain assumption All agents in a group share the same action space, state space, and policy network architecture.
    Required for model parameter sharing and adoption (Section 3, paragraph 2).
  • domain assumption Accumulated reward score is reliable evidence for assessing the quality of an agent's suggestions.
    Used by the Combo rule to filter advisors (Section 3.2, Combo Rule discussion).
  • domain assumption The three algorithms A2C, PPO, and ACER represent agents with different learning strengths.
    Motivates the heterogeneous setup but is not verified (Section 3.1).
  • domain assumption Applying an action suggested by a better-performing agent to the target agent's environment will improve learning.
    Core to the action selection rules; not proven (Section 3.2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Group-Agent Reinforcement Learning with Heterogeneous Agents." pith.science (2026). https://pith.science/paper/ARVVR7Z3

@misc{pith2026250111818,
  author       = {Pith},
  title        = {Pith review of: Group-Agent Reinforcement Learning with Heterogeneous Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ARVVR7Z3}},
  note         = {Machine review of arXiv:2501.11818}
}
read the original abstract

Group-agent reinforcement learning (GARL) is a newly arising learning scenario, where multiple reinforcement learning agents study together in a group, sharing knowledge in an asynchronous fashion. The goal is to improve the learning performance of each individual agent. Under a more general heterogeneous setting where different agents learn using different algorithms, we advance GARL by designing novel and effective group-learning mechanisms. They guide the agents on whether and how to learn from action choices from the others, and allow the agents to adopt available policy and value function models sent by another agent if they perform better. We have conducted extensive experiments on a total of 43 different Atari 2600 games to demonstrate the superior performance of the proposed method. After the group learning, among the 129 agents examined, 96% are able to achieve a learning speed-up, and 72% are able to learn over 100 times faster. Also, around 41% of those agents have achieved a higher accumulated reward score by learning in less than 5% of the time steps required by a single agent when learning on its own.

Figures

Figures reproduced from arXiv: 2501.11818 by the authors.

Figure 1
Figure 1. Flowchart: The agents of different types share [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Performance comparison for different agents and [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Action Probabilities: The probability that each [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Atari 2600 Games: Part 1. The Combo rule shows superb performance for all three agents of A2C, ACER and [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Atari 2600 Games: Part 2. The ACER agents are greatly improved under the Combo rule. [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Atari 2600 Games: Part 3. The PA or PM rule shows great performance. [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Atari 2600 Games: Part 4. The three rules produce similar performance that are still much better than single agents. [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 24 canonical work pages

  1. [1]

    Averaged-dqn: Variance reduction and stabilization for deep reinforcement learning

    Oron Anschel, Nir Baram, and Nahum Shimkin. Averaged-dqn: Variance reduction and stabilization for deep reinforcement learning. In International conference on machine learning, pages 176--185. PMLR, 2017

    work page 2017

  2. [2]

    The arcade learning environment: An evaluation platform for general agents

    Marc G Bellemare, Yavar Naddaf, Joel Veness, and Michael Bowling. The arcade learning environment: An evaluation platform for general agents. Journal of Artificial Intelligence Research, 47: 0 253--279, 2013

    2013

  3. [3]

    Bagging predictors

    Leo Breiman. Bagging predictors. Machine learning, 24: 0 123--140, 1996

    work page 1996

  4. [4]

    Openai gym, 2016

    Greg Brockman, Vicki Cheung, Ludwig Pettersson, Jonas Schneider, John Schulman, Jie Tang, and Wojciech Zaremba. Openai gym, 2016

    2016

  5. [5]

    Ucb exploration via q-ensembles

    Richard Y Chen, Szymon Sidor, Pieter Abbeel, and John Schulman. Ucb exploration via q-ensembles. arXiv preprint arXiv:1706.01502, 2017

    arXiv 2017

  6. [6]

    Ensemble network architecture for deep reinforcement learning

    Xi-liang Chen, Lei Cao, Chen-xi Li, Zhi-xiong Xu, and Jun Lai. Ensemble network architecture for deep reinforcement learning. Mathematical Problems in Engineering, 2018, 2018

    work page 2018

  7. [7]

    Shared experience actor-critic for multi-agent reinforcement learning

    Filippos Christianos, Lukas Sch \"a fer, and Stefano Albrecht. Shared experience actor-critic for multi-agent reinforcement learning. Advances in neural information processing systems, 33: 0 10707--10717, 2020

    work page 2020

  8. [8]

    Off-policy actor-critic

    Thomas Degris, Martha White, and Richard Sutton. Off-policy actor-critic. In International Conference on Machine Learning, 2012

    work page 2012

Show all 42 references
  1. [9]

    Openai baselines

    Prafulla Dhariwal, Christopher Hesse, Oleg Klimov, Alex Nichol, Matthias Plappert, Alec Radford, John Schulman, Szymon Sidor, Yuhuai Wu, and Peter Zhokhov. Openai baselines. https://github.com/openai/baselines, 2017

    2017

  2. [10]

    Ensembles for continuous actions in reinforcement learning

    Siegmund Duell and Steffen Udluft. Ensembles for continuous actions in reinforcement learning. In ESANN. Citeseer, 2013

    2013

  3. [11]

    Ensemble methods for reinforcement learning with function approximation

    Stefan Fau er and Friedhelm Schwenker. Ensemble methods for reinforcement learning with function approximation. Multiple Classifier Systems, pages 56--65, 2011

    2011

  4. [12]

    Neural network ensembles in reinforcement learning

    Stefan Fau er and Friedhelm Schwenker. Neural network ensembles in reinforcement learning. Neural Processing Letters, 41: 0 55--69, 2015 a

    2015

  5. [13]

    Selective neural network ensembles in reinforcement learning: taking the advantage of many agents

    Stefan Fau er and Friedhelm Schwenker. Selective neural network ensembles in reinforcement learning: taking the advantage of many agents. Neurocomputing, 169: 0 350--357, 2015 b

    2015

  6. [14]

    A short introduction to boosting

    Yoav Freund, Robert Schapire, and Naoki Abe. A short introduction to boosting. Journal-Japanese Society For Artificial Intelligence, 14 0 (771-780): 0 1612, 1999

    1999

  7. [15]

    Random decision forests

    Tin Kam Ho. Random decision forests. In Proceedings of 3rd international conference on document analysis and recognition, volume 1, pages 278--282. IEEE, 1995

    1995

  8. [16]

    Sunrise: A simple unified framework for ensemble learning in deep reinforcement learning

    Kimin Lee, Michael Laskin, Aravind Srinivas, and Pieter Abbeel. Sunrise: A simple unified framework for ensemble learning in deep reinforcement learning. In International Conference on Machine Learning, pages 6131--6141. PMLR, 2021

    2021

  9. [17]

    Multi-agent actor-critic for mixed cooperative-competitive environments

    Ryan Lowe, Yi I Wu, Aviv Tamar, Jean Harb, OpenAI Pieter Abbeel, and Igor Mordatch. Multi-agent actor-critic for mixed cooperative-competitive environments. Advances in neural information processing systems, 30, 2017

    2017

  10. [18]

    Maven: Multi-agent variational exploration

    Anuj Mahajan, Tabish Rashid, Mikayel Samvelyan, and Shimon Whiteson. Maven: Multi-agent variational exploration. Advances in neural information processing systems, 32, 2019

    2019

  11. [19]

    Albert bandura's social learning theory

    Saul McLeod. Albert bandura's social learning theory. Simply Psychology. London, 2011

    2011

  12. [20]

    Playing atari with deep reinforcement learning

    Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin Riedmiller. Playing atari with deep reinforcement learning. arXiv preprint arXiv:1312.5602, 2013

    2013 arXiv

  13. [21]

    Asynchronous methods for deep reinforcement learning

    Volodymyr Mnih, Adria Puigdomenech Badia, Mehdi Mirza, Alex Graves, Timothy Lillicrap, Tim Harley, David Silver, and Koray Kavukcuoglu. Asynchronous methods for deep reinforcement learning. In International conference on machine learning, pages 1928--1937. PMLR, 2016

    1928

  14. [22]

    Safe and efficient off-policy reinforcement learning

    R \'e mi Munos, Tom Stepleton, Anna Harutyunyan, and Marc Bellemare. Safe and efficient off-policy reinforcement learning. Advances in neural information processing systems, 29, 2016

    2016

  15. [23]

    Split feature space ensemble method using deep reinforcement learning for algorithmic trading

    Marcell N \'e meth and G \'a bor Sz u cs. Split feature space ensemble method using deep reinforcement learning for algorithmic trading. In Proceedings of the 2022 8th International Conference on Computer Technology Applications, pages 188--194, 2022

    2022

  16. [24]

    Deep exploration via bootstrapped dqn

    Ian Osband, Charles Blundell, Alexander Pritzel, and Benjamin Van Roy. Deep exploration via bootstrapped dqn. Advances in neural information processing systems, 29, 2016

    2016

  17. [25]

    Multiagent bidirectionally-coordinated nets: Emergence of human-level coordination in learning to play starcraft combat games

    Peng Peng, Ying Wen, Yaodong Yang, Quan Yuan, Zhenkun Tang, Haitao Long, and Jun Wang. Multiagent bidirectionally-coordinated nets: Emergence of human-level coordination in learning to play starcraft combat games. arXiv preprint arXiv:1703.10069, 2017

    2017 arXiv

  18. [26]

    Value propagation for decentralized networked deep multi-agent reinforcement learning

    Chao Qu, Shie Mannor, Huan Xu, Yuan Qi, Le Song, and Junwu Xiong. Value propagation for decentralized networked deep multi-agent reinforcement learning. Advances in Neural Information Processing Systems, 32, 2019

    2019

  19. [27]

    Qmix: Monotonic value function factorisation for deep multi-agent reinforcement learning

    T Rashid, CS De Witt, G Farquhar, J Foerster, S Whiteson, and M Samvelyan. Qmix: Monotonic value function factorisation for deep multi-agent reinforcement learning. In Proceedings of the 35th International Conference on Machine Learning, pages 6846--6859, 2018

    2018

  20. [28]

    Seerl: Sample efficient ensemble reinforcement learning

    Rohan Saphal, Balaraman Ravindran, Dheevatsa Mudigere, Sasikant Avancha, and Bharat Kaul. Seerl: Sample efficient ensemble reinforcement learning. In Proceedings of the 20th International Conference on Autonomous Agents and MultiAgent Systems, pages 1100--1108, 2021

    2021

  21. [29]

    Mastering atari, go, chess and shogi by planning with a learned model

    Julian Schrittwieser, Ioannis Antonoglou, Thomas Hubert, Karen Simonyan, Laurent Sifre, Simon Schmitt, Arthur Guez, Edward Lockhart, Demis Hassabis, Thore Graepel, et al. Mastering atari, go, chess and shogi by planning with a learned model. Nature, 588 0 (7839): 0 604--609, 2020

    2020

  22. [30]

    Proximal policy optimization algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  23. [31]

    Mastering the game of go with deep neural networks and tree search

    David Silver, Aja Huang, Chris J Maddison, Arthur Guez, Laurent Sifre, George Van Den Driessche, Julian Schrittwieser, Ioannis Antonoglou, Veda Panneershelvam, Marc Lanctot, et al. Mastering the game of go with deep neural networks and tree search. nature, 529 0 (7587): 0 484-...

    2016

  24. [32]

    Mastering the game of go without human knowledge

    David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai, Adrian Bolton, et al. Mastering the game of go without human knowledge. nature, 550 0 (7676): 0 354--359, 2017

    2017

  25. [33]

    A general reinforcement learning algorithm that masters chess, shogi, and go through self-play

    David Silver, Thomas Hubert, Julian Schrittwieser, Ioannis Antonoglou, Matthew Lai, Arthur Guez, Marc Lanctot, Laurent Sifre, Dharshan Kumaran, Thore Graepel, et al. A general reinforcement learning algorithm that masters chess, shogi, and go through self-play. Science, 362 0 ...

    2018

  26. [34]

    Pebl: Pessimistic ensembles for offline deep reinforcement learning

    Jordi Smit, Canmanie T Ponnambalam, Matthijs TJ Spaan, and Frans A Oliehoek. Pebl: Pessimistic ensembles for offline deep reinforcement learning. In Robust and Reliable Autonomy in the Wild Workshop at the 30th International Joint Conference of Artificial Intelligence, 2021

    2021

  27. [35]

    Learning multiagent communication with backpropagation

    Sainbayar Sukhbaatar, Rob Fergus, et al. Learning multiagent communication with backpropagation. Advances in neural information processing systems, 29, 2016

    2016

  28. [36]

    Value-decomposition networks for cooperative multi-agent learning based on team reward

    Peter Sunehag, Guy Lever, Audrunas Gruslys, Wojciech Marian Czarnecki, Vinicius Zambaldi, Max Jaderberg, Marc Lanctot, Nicolas Sonnerat, Joel Z Leibo, Karl Tuyls, et al. Value-decomposition networks for cooperative multi-agent learning based on team reward. In Proceedings of t...

    2018

  29. [37]

    Roma: multi-agent reinforcement learning with emergent roles

    Tonghan Wang, Heng Dong, Victor Lesser, and Chongjie Zhang. Roma: multi-agent reinforcement learning with emergent roles. In Proceedings of the 37th International Conference on Machine Learning, pages 9876--9886, 2020

    2020

  30. [38]

    Fcmnet: Full communication memory net for team-level cooperation in multi-agent systems

    Yutong Wang and Guillaume Sartoretti. Fcmnet: Full communication memory net for team-level cooperation in multi-agent systems. In Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems, pages 1355--1363, 2022

    2022

  31. [39]

    Sample efficient actor-critic with experience replay

    Ziyu Wang, Victor Bapst, Nicolas Heess, Volodymyr Mnih, Remi Munos, Koray Kavukcuoglu, and Nando de Freitas. Sample efficient actor-critic with experience replay. In International Conference on Learning Representations, 2017

    2017

  32. [40]

    Ensemble algorithms in reinforcement learning

    Marco A Wiering and Hado Van Hasselt. Ensemble algorithms in reinforcement learning. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 38 0 (4): 0 930--936, 2008

    2008

  33. [41]

    Group-agent reinforcement learning

    Kaiyue Wu and Xiao-Jun Zeng. Group-agent reinforcement learning. In International Conference on Artificial Neural Networks, 2023

    2023

  34. [42]

    Deep reinforcement learning for automated stock trading: An ensemble strategy

    Hongyang Yang, Xiao-Yang Liu, Shan Zhong, and Anwar Walid. Deep reinforcement learning for automated stock trading: An ensemble strategy. In Proceedings of the first ACM international conference on AI in finance, pages 1--8, 2020

    2020

Pith tools

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