Parallel convolution method improves multi-agent system learning efficiency
From Recursion to Convolution: Parallel Self-Improvement for Multi-Agent Systems
Artificial IntelligenceComputation and LanguageMachine Learning
Summary
Many computer programs use groups of agents that learn and improve by working together. Usually, improving these agents step by step takes a lot of time and can cause errors that slow things down. The authors propose a new way to let these agents learn and improve all at once using a technique called convolution, which helps them share information better and faster. They tested their approach on a math problem dataset and found it improved performance while making the process quicker and cheaper. This method also works for different kinds of teamwork setups among agents.
What this means in practice
- •For machine learning engineers: Reduce computation time and improve accuracy in collaborative agent-based learning tasks using parallel convolution techniques.
- •For distributed system architects: Design scalable multi-agent coordination systems that handle complex interaction topologies without sequential processing bottlenecks.
Authors
Zheng Wang, Yuang Liu, Sachith Abeywickrama, Yangkai Ding, Tao Yu, Xinyu Zhang, Hongbo Wang, Qi Ye, Xuefeng Jin, Zhangchun Zhao
Abstract
Multi-agent self-improvement enables agents to improve through interaction and feedback. However, sequential dependencies across agents and refinement rounds can propagate errors, increase computational costs, and limit parallelism. Moreover, refining individual features or agents in isolation can overlook their dependencies and restrict collective coordination. To address these challenges, we propose ConvRSI, a framework for parallelizing multi-agent self-improvement through convolution. ConvRSI couples intra-agent convolution, which jointly refines related agent features, with inter-agent graph convolution, which aggregates messages according to the collaboration topology. Both stages operate in latent space, enabling parallel updates without repeatedly decoding and re-encoding intermediate messages. Experiments on MATH-500 show that ConvRSI achieves 5.04% performance improvement while reducing 17.27% inference cost, and 45.11% latency compared with RecursiveMAS. It also supports diverse collaboration patterns by representing their topologies as adjacency matrices. These results demonstrate the potential of convolutional co-evolution for efficient multi-agent self-improvement. Our code and datasets are provided in https://github.com/LYA2333/ConvRSI.