ACC-Collab: An Actor-Critic Approach to Multi-Agent LLM Collaboration

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Estornell, Andrew, Ton, Jean-Francois, Yao, Yuanshun, Liu, Yang
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912261769003008
author Estornell, Andrew
Ton, Jean-Francois
Yao, Yuanshun
Liu, Yang
author_facet Estornell, Andrew
Ton, Jean-Francois
Yao, Yuanshun
Liu, Yang
contents Large language models (LLMs) have demonstrated a remarkable ability to serve as general-purpose tools for various language-based tasks. Recent works have demonstrated that the efficacy of such models can be improved through iterative dialog between multiple models. While these paradigms show promise in improving model efficacy, most works in this area treat collaboration as an emergent behavior, rather than a learned behavior. In doing so, current multi-agent frameworks rely on collaborative behaviors to have been sufficiently trained into off-the-shelf models. To address this limitation, we propose ACC-Collab, an Actor-Critic based learning framework to produce a two-agent team (an actor-agent and a critic-agent) specialized in collaboration. We demonstrate that ACC-Collab outperforms SotA multi-agent techniques on a wide array of benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2411_00053
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ACC-Collab: An Actor-Critic Approach to Multi-Agent LLM Collaboration
Estornell, Andrew
Ton, Jean-Francois
Yao, Yuanshun
Liu, Yang
Computation and Language
Artificial Intelligence
Large language models (LLMs) have demonstrated a remarkable ability to serve as general-purpose tools for various language-based tasks. Recent works have demonstrated that the efficacy of such models can be improved through iterative dialog between multiple models. While these paradigms show promise in improving model efficacy, most works in this area treat collaboration as an emergent behavior, rather than a learned behavior. In doing so, current multi-agent frameworks rely on collaborative behaviors to have been sufficiently trained into off-the-shelf models. To address this limitation, we propose ACC-Collab, an Actor-Critic based learning framework to produce a two-agent team (an actor-agent and a critic-agent) specialized in collaboration. We demonstrate that ACC-Collab outperforms SotA multi-agent techniques on a wide array of benchmarks.
title ACC-Collab: An Actor-Critic Approach to Multi-Agent LLM Collaboration
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2411.00053