Pith. sign in

REVIEW 4 major objections 6 minor 31 references

Learning Individual Intrinsic Reward in Multi-Agent Reinforcement Learning via Incorporating Generalized Human Expertise

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read LIGHT injects human-knowledge preferences as per-agent intrinsic rewards and reports state-of-the-art results on sparse-reward MARL benchmarks.

desk verdict LIGHT is a plausible new combination of decision-tree human rules and intrinsic-reward MARL with promising empirical gains, but the training objective is underspecified in a way that determines whether the method actually optimizes team reward. read the letter →

arxiv 2507.18867 v1 pith:4ARQQ6AQ submitted 2025-07-25 cs.LG cs.AIcs.MA

Xuefei Wu , Xiao Yin , Yuanyang Zhu , Chunlin Chen This is my paper · ORCID
classification cs.LGcs.AIcs.MA
keywords multi-agentreinforcementlearningintrinsicrewardhumanexpertisesparserewardsvaluedecompositioncentralizedtrainingdecentralizedexecutionLevel-BasedForagingStarCraftChallenge
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

LIGHT is a proposal for sparse-reward cooperative multi-agent reinforcement learning. The paper argues that a few soft logic rules distilled from human knowledge can be turned into dense, per-agent intrinsic rewards that guide exploration more effectively than hand-designed shaping or generic intrinsic motivation. The reward for each agent is the negative Euclidean distance between a transformed action distribution from the rules and the agent's own action-value distribution, and the whole module is trained end to end inside a value-decomposition algorithm. If the claim is right, MARL agents in tasks with only a team reward can learn faster, align their behavior with human preferences, and reuse the same knowledge across related tasks, without needing a fully specified dense reward function.

What carries the argument

The load-bearing object is the intrinsic reward in Eq. (4): $r_i^t = -\|\varphi_i(H(o_i^t)) - \varphi_i(A_i^t)\|_2$, the negative Euclidean distance between a transformed soft-logic-rule action distribution and the agent's own action-value distribution. It converts a few human-supplied rules into a dense, time-varying per-agent learning signal. The other piece of machinery is the two-loss training scheme: the mixing network is updated on the shaped team reward $R_t = r_t^{ex} + \lambda \frac{1}{N}\sum_i r_i^t$ (Eqs. 5-6), while each individual Q-network is updated on its intrinsic reward (Eq. 7), with the total loss $L = L_{TD}(\theta) + \lambda_K L_i(\theta_i)$ (Eq. 8). This is the mechanism by which human knowledge enters the credit assignment loop.

What would settle it

On the 5m SMAC sparse-reward map, retrain LIGHT with the human-knowledge rule inverted - recommend attack whenever health is below 15 instead of retreat. If the win-rate curve does not fall markedly below LIGHT with the correct rule, the knowledge signal is not what is driving the reported improvement; if it does fall, the mechanism is confirmed.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that human expertise can be carried by a differentiable distance penalty rather than by hand-coded rewards or hard constraints. LIGHT computes, for each agent, $r_i^t = -\|\varphi_i(H(o_i^t)) - \varphi_i(A_i^t)\|_2$, the negative Euclidean distance between the human-knowledge action distribution and the agent's own action-value distribution, and adds the average of these signals to the team reward for the mixing network while training each individual Q-network on its intrinsic reward. The authors report that LIGHT outperforms QMIX, VDN, QTRAN, LIIR, and MASER on Level-Based Foraging and on sparse-reward StarCraft scenarios, that plugging LIGHT onto QMIX or VDN improves the backbone, and that the trained agents' actions match the human-knowledge rules more often than QMIX's do.

Load-bearing premise

The load-bearing premise is that training each agent's Q-network on the intrinsic reward alone (Eq. 7), while training the mixing network on the shaped team reward (Eq. 6), keeps the two objectives consistent enough that individual utilities still factor the true team value; the paper offers no proof or empirical check that this preserves the IGM principle, and if the losses conflict the reported gains could vanish.

Editorial extensions

If this is right

  • LIGHT can be dropped onto any value-decomposition backbone: the paper shows LIGHT-QMIX and LIGHT-VDN both beat their fine-tuned backbones on sparse-reward SMAC maps.
  • Agents trained with LIGHT complete 4m and 2m_vs_1z episodes in fewer steps than QMIX and match human-knowledge rules more often.
  • Removing either the intrinsic reward or the human-knowledge distribution degrades performance, so both components contribute to the reported gains.
  • The method requires only a few soft logic rules as prior knowledge, not a fully specified dense reward function.
  • Because the module is end-to-end and parameterized, the same knowledge can be reused across tasks that share rules.

Reading between the lines

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

  • If the improvement is real, one natural next step is to learn the rule extraction together with the policy rather than fixing rules beforehand; that would let the intrinsic reward track the agents' evolving competence.
  • The distance signal essentially acts as a dynamic reward bias, so its effect should be equivalent to potential-based shaping only when the two distributions stay close. Quantifying that bias and testing whether LIGHT preserves the optimal policy under arbitrary rule mistakes would connect it to standard reward-shaping invariance results.
  • A direct prediction of the paper's mechanism is that LIGHT's advantage over plain QMIX should shrink as extrinsic rewards become denser, because the human-knowledge signal then mostly duplicates information the environment already provides.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes LIGHT, a reward-shaping framework for cooperative multi-agent reinforcement learning under the CTDE paradigm. LIGHT computes an individual intrinsic reward for each agent as the negative Euclidean distance between the agent's action-value distribution and a 'soft logic rule' distribution derived from human knowledge. This intrinsic reward is added, with a weighting coefficient, to the extrinsic team reward, and the overall pipeline is trained with a combined TD loss for the mixing network and an individual Q-learning loss per agent. The method is evaluated on Level-Based Foraging (LBF) and StarCraft Multi-Agent Challenge (SMAC) in both dense- and sparse-reward settings, against five baselines, and is also analyzed through ablations and behavior-alignment statistics.

Significance. If the empirical claims hold, LIGHT provides a simple, plug-in method for injecting human knowledge into value-decomposition MARL algorithms, with reported gains on sparse-reward benchmarks and better alignment of agent behavior with human rules. The paper includes five-seed comparisons, ablations, and visualizations of the learned intrinsic rewards, which are informative. However, the strength of the contribution is undercut by an ambiguity in the joint training objective that may violate the IGM principle, by the circularity of the behavior-alignment analysis, and by missing implementation details that prevent reproduction. These issues are addressable, but they require substantial revision rather than copy editing.

major comments (4)
  1. [Section III-B, Eqs. (5)-(8)] The two-loss training scheme is internally inconsistent as specified. Eq. (6) trains Q_tot on R_t = r_ext + lambda * (1/N) * sum_i r_i,t, while Eq. (7) trains each Q_i with a TD target that uses only the intrinsic reward r_i,t. For the QMIX/VDN architectures used in the paper, Q_tot is a monotone function of the individual Q_i, so under the IGM principle the greedy joint action is the tuple of individual argmax actions. If the Q_i are updated solely by Eq. (7), action selection is driven exclusively by the intrinsic reward, and the extrinsic team reward cannot influence the chosen joint action. If instead gradients from Eq. (6) flow into the individual parameters in an end-to-end manner, the paper must state this explicitly and analyze the resulting gradient flow, since Eq. (7) then acts as an auxiliary objective whose effect on credit assignment is unexamined. The manuscript provides neither a proof that the combined objective preserves IGM nor an empirical check, such as comparing argmax_a Q_tot(tau,a) with the tuple of argmax_{a_i} Q_i(tau_i,a_i) on collected states. Without this, the central claim that LIGHT 'maximizes the joint action value' is unsupported.
  2. [Section VI-D, Fig. 10] The behavior-alignment analysis is circular. Eq. (4) defines the intrinsic reward as the negative Euclidean distance between the agent's action distribution and the human-knowledge distribution H(o_i^t), and Eq. (8) trains the agents to maximize this reward. Measuring the fraction of actions consistent with the human-knowledge rules therefore confirms by construction, not empirically, that the learned policies align with the injected rules. The claim that LIGHT 'can efficiently capture the given human knowledge' is a tautology. The authors should test alignment on held-out human-labeled states or compare the consistency metric of LIGHT against a variant where the intrinsic reward is not directly derived from the same rules (as in the 'LIGHT w/o knowledge' ablation), and demonstrate a distributional difference.
  3. [Section III-B and Table II] Three implementation details that are load-bearing for reproducibility are missing. (i) The weight lambda in Eq. (5) is never given in Table II or anywhere in the text; only lambda_K = 0.02 is mentioned in Section VI. (ii) The probability p in Algorithm 1, and its role in defining the soft-logic-rule distribution H(o_i^t), is not mathematically specified; it is unclear whether p is a fixed constant, a learned quantity, or a function of health. (iii) The human-knowledge rule for the LBF experiments is never presented, so the LBF results cannot be reproduced. The authors should report lambda, define p precisely, and provide the full rule formulation for both benchmarks.
  4. [Section VI-A, Figs. 3-6] The claim that LIGHT 'achieves impressive performance on all scenarios' is not backed by statistical analysis. Only five seeds are used and no confidence intervals or significance tests are reported; in Fig. 3, VDN appears to match LIGHT before 0.5M steps, and in Fig. 5 the dense-reward results saturate for all methods. A formal comparison with error bars or paired tests is needed to support the superiority claim, especially for the sparse-reward SMAC maps.
minor comments (6)
  1. [Eq. (5)] The summation index 'ri,jt' appears to be a typo for r_i,t; please fix it.
  2. [Algorithm 1] The output line 'Get the probability p, where argmax_a p = M' is unclear; p should be defined as a distribution or a scalar, and its mapping to H(o_i^t) should be stated.
  3. [Section VI-A, Fig. 3] The phrase 'before 0.5 steps' lacks the time unit (presumably 0.5 million steps); please correct it.
  4. [Fig. 10] The y-axis 'Similarity' in Fig. 10(b) is not defined in the text; specify how action-consistency with human knowledge is computed.
  5. [Section VI-B] In Fig. 4 caption it says '3-agent & 3-food' but the surrounding text refers to '4-agent & 2-food'; please reconcile the description.
  6. [References] Reference [28] appears to have the author ordering garbled; also check that all citations in the text, such as [27] near the IGM definition, point to the correct entries.

Circularity Check

1 steps flagged · score 6.0 of 10

The behavior-alignment claim in Sec. VI-D is the training objective restated; the benchmark performance claims are independently evaluated.

  1. self definitional [Section VI-D (Behavior analysis), with Eqs. (4) and (8)]
    "we compute the intrinsic reward r_i^t as r_i^t = −\|\phi_i(H(o_i^t))−\phi_i(A_i^t)\|_2 ... The total loss function used in this work is expressed as follows: L = L_TD(θ) + λ_K L_i(θ_i) ... LIGHT has more behaviors aligning with the given human knowledge than QMIX across both scenarios"

    Equation (4) defines the intrinsic reward as the negative Euclidean distance between the agent's action distribution A_i^t and the human-knowledge distribution H(o_i^t), and Eq. (8) explicitly trains each agent's Q-network with L_i(θ_i), a TD loss whose target is r_i^t. Maximizing that reward is therefore the same as minimizing the distance to the human-knowledge distribution. Section VI-D then measures exactly whether the agent's action is 'consistent with the given human knowledge' and reports that LIGHT has more such behaviors than QMIX. That comparison is forced by the loss definition: LIGHT is optimized to align with H, while QMIX is not. The behavior-alignment 'finding' is the training objective restated, not an emergent property. The benchmark performance claims (Figs.

full rationale

Reviewing the full derivation chain, the core LIGHT construction—Eqs. (5)-(7) combine extrinsic team reward with an intrinsic reward, and Eq. (8) adds the individual loss—is benchmarked against QMIX, VDN, QTRAN, LIIR, and MASER on LBF and SMAC. Those comparisons are external and not determined by the paper's own definitions, so the central 'outperforms baselines' claim carries independent content. The only step that reduces by construction is the Section VI-D behavior-alignment analysis: since Eq. (4) defines intrinsic reward as negative distance to human-knowledge distributions and Eq. (8) trains agents to maximize that reward, observing above-chance action consistency with human knowledge is a direct report of the optimization target. No load-bearing self-citation chain was found: the authors' own prior works ([11], [12], [24], [25]) are used as supporting technique references, not as uniqueness theorems or forced ansatz justifications. Note also an internal inconsistency in Sec. VI-B ('the intrinsic rewards do not influence the learned policy and are not utilized in trajectory generation') versus Eq. (8), but that is a correctness/consistency risk, not a circularity step. Overall: partial circularity in one secondary claim, independent benchmark core, score 6.

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

The method introduces no new physical or conceptual entities; it reuses existing MARL components (Q-functions, mixing networks) and human rules. The main axioms are domain assumptions about the validity of the learned rules and the reward metric.

free parameters (4)
  • λ (Eq. 5)
    Weight balancing extrinsic and intrinsic reward in the shaped team reward; not listed in Table II, so the exact value used in experiments is unknown.
  • λ_K (Eq. 8) = 0.02
    Weight for the individual intrinsic-reward loss; stated once in Section VI.
  • Health threshold in Algorithm 1 = 15
    Hand-chosen threshold for the movement rule in SMAC 3m; likely chosen based on the map's enemy attack range, but no sensitivity analysis is given.
  • Rule probability p
    Probability attached to soft logic rules; not specified how it is derived from the decision tree. This affects the shape of H(o_i^t) and thus the intrinsic reward.
assumptions (4)
  • standard math Dec-POMDP and IGM principle (Eq. 1) hold for the value decomposition.
    The method builds on QMIX and VDN, which assume the IGM principle for consistent credit assignment.
  • domain assumption Human knowledge can be captured by a few soft logic rules extracted from offline MARL data via decision trees.
    The method depends on the availability of offline data and the validity of decision-tree extraction; the paper does not describe the dataset or the extraction process in detail.
  • domain assumption The Euclidean distance between the agent's action distribution and the rule's preference distribution is a useful intrinsic reward signal.
    No analysis or ablation supports this specific metric over alternatives; it is a design choice.
  • ad hoc to paper Training individual Q_i with the intrinsic-reward TD target (Eq. 7) while training Q_tot with the shaped reward (Eq. 6) preserves the IGM principle and yields stable credit assignment.
    The paper does not provide a theoretical or empirical argument that the two losses are consistent; this is an unflagged assumption that is load-bearing for the method's correctness.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Individual Intrinsic Reward in Multi-Agent Reinforcement Learning via Incorporating Generalized Human Expertise." pith.science (2026). https://pith.science/paper/4ARQQ6AQ

@misc{pith2026250718867,
  author       = {Pith},
  title        = {Pith review of: Learning Individual Intrinsic Reward in Multi-Agent Reinforcement Learning via Incorporating Generalized Human Expertise},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4ARQQ6AQ}},
  note         = {Machine review of arXiv:2507.18867}
}
read the original abstract

Efficient exploration in multi-agent reinforcement learning (MARL) is a challenging problem when receiving only a team reward, especially in environments with sparse rewards. A powerful method to mitigate this issue involves crafting dense individual rewards to guide the agents toward efficient exploration. However, individual rewards generally rely on manually engineered shaping-reward functions that lack high-order intelligence, thus it behaves ineffectively than humans regarding learning and generalization in complex problems. To tackle these issues, we combine the above two paradigms and propose a novel framework, LIGHT (Learning Individual Intrinsic reward via Incorporating Generalized Human experTise), which can integrate human knowledge into MARL algorithms in an end-to-end manner. LIGHT guides each agent to avoid unnecessary exploration by considering both individual action distribution and human expertise preference distribution. Then, LIGHT designs individual intrinsic rewards for each agent based on actionable representational transformation relevant to Q-learning so that the agents align their action preferences with the human expertise while maximizing the joint action value. Experimental results demonstrate the superiority of our method over representative baselines regarding performance and better knowledge reusability across different sparse-reward tasks on challenging scenarios.

Figures

Figures reproduced from arXiv: 2507.18867 by the authors.

Figure 1
Figure 1. The framework of LIGHT. (a) In the knowledge module, action distributions are generated through a knowledge base considering states and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Example of a task with the rule about human knowledge. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. An example of the intrinsic reward curves and some keyframes on [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (7 more)
Figure 3
Figure 3. Figure 3: Performance comparison with different baselines on two constructed [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 5
Figure 5. Figure 5: Performance comparison with baselines on dense-reward setting scenarios. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Performance comparison with baselines on sparse-reward setting scenarios. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: An example of the intrinsic reward curves and auxiliary illustration [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: LIGHT is plugged into two different baselines on sparse-reward scenarios. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 11
Figure 11. Figure 11: An example of the intrinsic reward curves and auxiliary illustration [PITH_FULL_IMAGE:figures/full_fig_p007_11.png]
Figure 10
Figure 10. Figure 10: A comparison of alignment with human knowledge preference [PITH_FULL_IMAGE:figures/full_fig_p007_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 27 canonical work pages

  1. [1]

    An overview of recent progress in the study of distributed multi-agent coordination,

    Y . Cao, W. Yu, W. Ren, and G. Chen, “An overview of recent progress in the study of distributed multi-agent coordination,” IEEE Trans. Industr. Inform., pp. 427–438, 2012

    work page 2012

  2. [2]

    Coordinated multi-agent reinforcement learning in networked distributed pomdps,

    C. Zhang and V . Lesser, “Coordinated multi-agent reinforcement learning in networked distributed pomdps,” inProceedings of the AAAI Conference on Artificial Intelligence , 2011

    work page 2011

  3. [3]

    Interpretation of neural networks is fragile,

    A. Ghorbani, A. Abid, and J. Zou, “Interpretation of neural networks is fragile,” in Proceedings of the AAAI Conference on Artificial Intelligence, 2019, pp. 3681–3688

    work page 2019

  4. [4]

    Guided deep reinforce- ment learning for swarm systems,

    M. Hüttenrauch, A. Šoši ´c, and G. Neumann, “Guided deep reinforce- ment learning for swarm systems,” arXiv:1709.06011, 2017

    arXiv 2017

  5. [5]

    Q-value path decomposition for deep multiagent reinforce- ment learning,

    Y . Yang, J. Hao, G. Chen, H. Tang, Y . Chen, Y . Hu, C. Fan, and Z. Wei, “Q-value path decomposition for deep multiagent reinforce- ment learning,” in Proceedings of the International Conference on Machine Learning, 2020, pp. 10 706–10 715

    work page 2020

  6. [6]

    RODE: Learning roles to decompose multi-agent tasks,

    T. Wang, T. Gupta, A. Mahajan, B. Peng, S. Whiteson, and C. Zhang, “RODE: Learning roles to decompose multi-agent tasks,” in Proceed- ings of the International Conference on Learning Representations , 2020, pp. 1–20

    work page 2020

  7. [7]

    ROMA: Multi-agent reinforcement learning with emergent roles,

    T. Wang, H. Dong, V . Lesser, and C. Zhang, “ROMA: Multi-agent reinforcement learning with emergent roles,” in Proceedings of the International Conference on Machine Learning, 2020, pp. 9876–9886

    work page 2020

  8. [8]

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

    P. Sunehag, G. Lever, A. Gruslys, W. M. Czarnecki, V . Zambaldi, M. Jaderberg, M. Lanctot, N. Sonnerat, J. Z. Leibo, K. Tuyls, et al., “Value-decomposition networks for cooperative multi-agent learning based on team reward,” in Proceedings of the International Conference on Autonomous Agents and MultiAgent Systems, 2018, pp. 2085–2087

    work page 2018

Show all 31 references
  1. [9]

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

    T. Rashid, M. Samvelyan, C. Schroeder, G. Farquhar, J. Foerster, and S. Whiteson, “QMIX: Monotonic value function factorisation for deep multi-agent reinforcement learning,” in Proceedings of the International Conference on Machine Learning, 2018, pp. 4295–4304

    2018

  2. [10]

    QPLEX: Duplex dueling multi-agent Q-learning,

    J. Wang, Z. Ren, T. Liu, Y . Yu, and C. Zhang, “QPLEX: Duplex dueling multi-agent Q-learning,” in Proceedings of the International Conference on Learning Representations , 2020, pp. 1–27

    2020

  3. [11]

    Mixrts: Toward interpretable multi-agent reinforcement learning via mixing recurrent soft decision trees,

    Z. Liu, Y . Zhu, Z. Wang, Y . Gao, and C. Chen, “Mixrts: Toward interpretable multi-agent reinforcement learning via mixing recurrent soft decision trees,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 5, pp. 4090–4107, 2025

    2025

  4. [12]

    NA 2Q: Neural attention additive model for interpretable multi-agent q-learning,

    Z. Liu, Y . Zhu, and C. Chen, “NA 2Q: Neural attention additive model for interpretable multi-agent q-learning,” in Proceedings of the International Conference on Machine Learning , vol. 202, 2023, pp. 22 539–22 558

    2023

  5. [13]

    A comprehensive survey of multiagent reinforcement learning,

    L. Busoniu, R. Babuska, and B. De Schutter, “A comprehensive survey of multiagent reinforcement learning,” IEEE Transactions on Systems, Man, and Cybernetics , vol. 38, no. 2, pp. 156–172, 2008

    2008

  6. [14]

    Dynamic agent-based reward shaping for multi-agent systems,

    M. Sadeghlou, M. R. Akbarzadeh-T, and M. B. Naghibi-S, “Dynamic agent-based reward shaping for multi-agent systems,” in Iranian Conference on Intelligent Systems , 2014, pp. 1–6

    2014

  7. [15]

    Multiagent deep reinforcement learning: Challenges and directions towards human-like approaches,

    A. Wong, T. Bäck, A. V . Kononova, and A. Plaat, “Multiagent deep reinforcement learning: Challenges and directions towards human-like approaches,” arXiv:2106.15691, 2021

    2021 arXiv

  8. [16]

    Cooperative exploration for multi-agent deep reinforcement learning,

    I.-J. Liu, U. Jain, R. A. Yeh, and A. Schwing, “Cooperative exploration for multi-agent deep reinforcement learning,” in Proceedings of the International conference on machine learning , 2021, pp. 6826–6836

    2021

  9. [17]

    Maven: Multi-agent variational exploration,

    A. Mahajan, T. Rashid, M. Samvelyan, and S. Whiteson, “Maven: Multi-agent variational exploration,” Advances in neural information processing systems, vol. 32, 2019

    2019

  10. [18]

    Liir: Learning individual intrinsic reward in multi-agent reinforcement learning,

    Y . Du, L. Han, M. Fang, J. Liu, T. Dai, and D. Tao, “Liir: Learning individual intrinsic reward in multi-agent reinforcement learning,” in Advances in neural information processing systems , vol. 32, 2019

    2019

  11. [19]

    MASER: Multi-agent reinforcement learning with subgoals generated from experience replay buffer,

    J. Jeon, W. Kim, W. Jung, and Y . Sung, “MASER: Multi-agent reinforcement learning with subgoals generated from experience replay buffer,” in Proceedings of the International Conference on Machine Learning, 2022, pp. 10 041–10 052

    2022

  12. [20]

    Individual reward assisted multi-agent reinforcement learning,

    L. Wang, Y . Zhang, Y . Hu, W. Wang, C. Zhang, Y . Gao, J. Hao, T. Lv, and C. Fan, “Individual reward assisted multi-agent reinforcement learning,” in Proceedings of the International Conference on Machine Learning. PMLR, 2022, pp. 23 417–23 432

    2022

  13. [21]

    Haven: hierarchical cooperative multi-agent reinforcement learning with dual coordination mechanism,

    Z. Xu, Y . Bai, B. Zhang, D. Li, and G. Fan, “Haven: hierarchical cooperative multi-agent reinforcement learning with dual coordination mechanism,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 10, 2023, pp. 11 735–11 743

    2023

  14. [22]

    Hierarchical reinforcement learning in starcraft ii with human expertise in subgoals selection,

    X. Xu, T. Huang, P. Wei, A. Narayan, and T.-Y . Leong, “Hierarchical reinforcement learning in starcraft ii with human expertise in subgoals selection,” arXiv preprint arXiv:2008.03444 , 2020

    2008 arXiv

  15. [23]

    Dl2: training and querying neural networks with logic,

    M. Fischer, M. Balunovic, D. Drachsler-Cohen, T. Gehr, C. Zhang, and M. Vechev, “Dl2: training and querying neural networks with logic,” in Proceedings of the International Conference on Machine Learning , 2019, pp. 1931–1941

    2019

  16. [24]

    Rule-based reinforce- ment learning for efficient robot navigation with space reduction,

    Y . Zhu, Z. Wang, C. Chen, and D. Dong, “Rule-based reinforce- ment learning for efficient robot navigation with space reduction,” IEEE/ASME Transactions on Mechatronics , vol. 27, no. 2, pp. 846– 857, 2022

    2022

  17. [25]

    Extracting decision tree from trained deep reinforcement learning in traffic signal control,

    Y . Zhu, X. Yin, and C. Chen, “Extracting decision tree from trained deep reinforcement learning in traffic signal control,” IEEE Transac- tions on Computational Social Systems, vol. 10, no. 4, pp. 1997–2007, 2023

    1997

  18. [26]

    Kogun: accelerating deep reinforcement learning via integrating hu- man suboptimal knowledge,

    P. Zhang, J. Hao, W. Wang, H. Tang, Y . Ma, Y . Duan, and Y . Zheng, “Kogun: accelerating deep reinforcement learning via integrating hu- man suboptimal knowledge,” arXiv preprint arXiv:2002.07418 , 2020

    2002 arXiv

  19. [27]

    QTRAN: Learning to factorize with transformation for cooperative multi-agent reinforcement learning,

    K. Son, D. Kim, W. J. Kang, D. E. Hostallero, and Y . Yi, “QTRAN: Learning to factorize with transformation for cooperative multi-agent reinforcement learning,” in Proceedings of the International Confer- ence on Machine Learning , 2019, pp. 5887–5896

    2019

  20. [28]

    Exploration with unreliable intrinsic reward in multi-agent reinforcement learning,

    S. W. Wendelin Böhmer, Tabish Rashid, “Exploration with unreliable intrinsic reward in multi-agent reinforcement learning,” in Proceedings of the International Conference on Machine Learning , 2019

    2019

  21. [29]

    Discovering generalizable multi-agent coordination skills from multi-task offline data,

    F. Zhang, C. Jia, Y .-C. Li, L. Yuan, Y . Yu, and Z. Zhang, “Discovering generalizable multi-agent coordination skills from multi-task offline data,” in Proceedings of the International Conference on Learning Representations, 2023

    2023

  22. [30]

    Shared experience actor- critic for multi-agent reinforcement learning,

    F. Christianos, L. Schäfer, and S. Albrecht, “Shared experience actor- critic for multi-agent reinforcement learning,” in Advances in Neural Information Processing Systems , 2020, pp. 10 707–10 717

    2020

  23. [31]

    The StarCraft Multi-Agent Challenge,

    M. Samvelyan, T. Rashid, C. Schroeder de Witt, G. Farquhar, N. Nardelli, T. G. Rudner, C.-M. Hung, P. H. Torr, J. Foerster, and S. Whiteson, “The StarCraft Multi-Agent Challenge,” in Proceedings of the International Conference on Autonomous Agents and MultiAgent Systems, 2019,...

    2019

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.