Pith. sign in

REVIEW 1 cited by

Scalable Multi-Agent Inverse Reinforcement Learning via Actor-Attention-Critic

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2002.10525 v1 pith:VTPTINMX submitted 2020-02-24 cs.MA cs.LG

Scalable Multi-Agent Inverse Reinforcement Learning via Actor-Attention-Critic

classification cs.MA cs.LG
keywords multi-agentagentsinversemethodactor-attention-criticbaselineslearningma-airl
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Multi-agent adversarial inverse reinforcement learning (MA-AIRL) is a recent approach that applies single-agent AIRL to multi-agent problems where we seek to recover both policies for our agents and reward functions that promote expert-like behavior. While MA-AIRL has promising results on cooperative and competitive tasks, it is sample-inefficient and has only been validated empirically for small numbers of agents -- its ability to scale to many agents remains an open question. We propose a multi-agent inverse RL algorithm that is more sample-efficient and scalable than previous works. Specifically, we employ multi-agent actor-attention-critic (MAAC) -- an off-policy multi-agent RL (MARL) method -- for the RL inner loop of the inverse RL procedure. In doing so, we are able to increase sample efficiency compared to state-of-the-art baselines, across both small- and large-scale tasks. Moreover, the RL agents trained on the rewards recovered by our method better match the experts than those trained on the rewards derived from the baselines. Finally, our method requires far fewer agent-environment interactions, particularly as the number of agents increases.

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Symmetry-Guided Multi-Agent Inverse Reinforcement Learning

    cs.RO 2025-09 reject novelty 5.0

    Symmetry-guided data augmentation and a symmetry-aware discriminator improve sample efficiency in multi-agent inverse reinforcement learning.