# Reducing Variance Caused by Communication in Decentralized Multi-agent Deep Reinforcement Learning

arXiv: https://arxiv.org/abs/2502.06261 · HTML: https://mlsanity.com/paper/2502.06261/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 10 of its 12 references that are in the map.
- First posted on arXiv 10 Feb 2025.

## Builds on (most cited first, 10 of 10)

- [Learning to Communicate with Deep Multi-Agent Reinforcement Learning](https://mlsanity.com/paper/1605.06676/) (2016)
- [Actor-Attention-Critic for Multi-Agent Reinforcement Learning](https://mlsanity.com/paper/1810.02912/) (2018)
- [Learning Attentional Communication for Multi-Agent Cooperation](https://mlsanity.com/paper/1805.07733/) (2018)
- [A Survey of Multi-Agent Deep Reinforcement Learning with Communication](https://mlsanity.com/paper/2203.08975/) (2022)
- [The Optimal Reward Baseline for Gradient-Based Reinforcement Learning](https://mlsanity.com/paper/1301.2315/) (2013)
- [Multi-Agent Game Abstraction via Graph Attention Neural Network](https://mlsanity.com/paper/1911.10715/) (2019)
- [Variance Reduction for Policy Gradient with Action-Dependent Factorized Baselines](https://mlsanity.com/paper/1803.07246/) (2018)
- [Contrasting Centralized and Decentralized Critics in Multi-Agent Reinforcement Learning](https://mlsanity.com/paper/2102.04402/) (2021)
- [Settling the Variance of Multi-Agent Policy Gradients](https://mlsanity.com/paper/2108.08612/) (2021)
- [Learning to Communicate in Multi-Agent Reinforcement Learning : A Review](https://mlsanity.com/paper/1911.05438/) (2019)

## Built on by (most cited first, 1 of 1)

- [Dynamic Generation of Multi-LLM Agents Communication Topologies with Graph Diffusion Models](https://mlsanity.com/paper/2510.07799/) (2025)

## Abstract (arXiv)

In decentralized multi-agent deep reinforcement learning (MADRL), communication can help agents to gain a better understanding of the environment to better coordinate their behaviors. Nevertheless, communication may involve uncertainty, which potentially introduces variance to the learning of decentralized agents. In this paper, we focus on a specific decentralized MADRL setting with communication and conduct a theoretical analysis to study the variance that is caused by communication in policy gradients. We propose modular techniques to reduce the variance in policy gradients during training. We adopt our modular techniques into two existing algorithms for decentralized MADRL with communication and evaluate them on multiple tasks in the StarCraft Multi-Agent Challenge and Traffic Junction domains. The results show that decentralized MADRL communication methods extended with our proposed techniques not only achieve high-performing agents but also reduce variance in policy gradients during training.