Predicting Multi-Agent Specialization via Task Parallelizability

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mieczkowski, Elizabeth, Mon-Williams, Ruaridh, Bramley, Neil, Lucas, Christopher G., Velez, Natalia, Griffiths, Thomas L.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915500001329152
author Mieczkowski, Elizabeth
Mon-Williams, Ruaridh
Bramley, Neil
Lucas, Christopher G.
Velez, Natalia
Griffiths, Thomas L.
author_facet Mieczkowski, Elizabeth
Mon-Williams, Ruaridh
Bramley, Neil
Lucas, Christopher G.
Velez, Natalia
Griffiths, Thomas L.
contents When should we encourage specialization in multi-agent systems versus train generalists that perform the entire task independently? We propose that specialization largely depends on task parallelizability: the potential for multiple agents to execute task components concurrently. Drawing inspiration from Amdahl's Law in distributed systems, we present a closed-form bound that predicts when specialization improves performance, depending only on task concurrency and team size. We validate our model on two standard MARL benchmarks that represent opposite regimes -- StarCraft Multi-Agent Challenge (SMAC, unlimited concurrency) and Multi-Particle Environment (MPE, unit-capacity bottlenecks) -- and observe close alignment between the bound at each extreme and an empirical measure of specialization. Three follow-up experiments in Overcooked-AI demonstrate that the model works in environments with more complex spatial and resource bottlenecks that allow for a range of strategies. Beyond prediction, the bound also serves as a diagnostic tool, highlighting biases in MARL training algorithms that cause sub-optimal convergence to specialist strategies with larger state spaces.
format Preprint
id arxiv_https___arxiv_org_abs_2503_15703
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Predicting Multi-Agent Specialization via Task Parallelizability
Mieczkowski, Elizabeth
Mon-Williams, Ruaridh
Bramley, Neil
Lucas, Christopher G.
Velez, Natalia
Griffiths, Thomas L.
Multiagent Systems
Artificial Intelligence
When should we encourage specialization in multi-agent systems versus train generalists that perform the entire task independently? We propose that specialization largely depends on task parallelizability: the potential for multiple agents to execute task components concurrently. Drawing inspiration from Amdahl's Law in distributed systems, we present a closed-form bound that predicts when specialization improves performance, depending only on task concurrency and team size. We validate our model on two standard MARL benchmarks that represent opposite regimes -- StarCraft Multi-Agent Challenge (SMAC, unlimited concurrency) and Multi-Particle Environment (MPE, unit-capacity bottlenecks) -- and observe close alignment between the bound at each extreme and an empirical measure of specialization. Three follow-up experiments in Overcooked-AI demonstrate that the model works in environments with more complex spatial and resource bottlenecks that allow for a range of strategies. Beyond prediction, the bound also serves as a diagnostic tool, highlighting biases in MARL training algorithms that cause sub-optimal convergence to specialist strategies with larger state spaces.
title Predicting Multi-Agent Specialization via Task Parallelizability
topic Multiagent Systems
Artificial Intelligence
url https://arxiv.org/abs/2503.15703