AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

AgentTailor: A Semantic-Aware LLM-Based Multi-Agent System with Actor-Critic Structure

Peiting Yang, Jiahao Shi, Caiyi Xu, Ming Liu, Yanxia Wu, Rongsheng Li

ICML 2026 · Multi-agent systems and games › LLM agents · approach: LLM multi-agent decision making

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes AgentTailor, an LLM-based multi-agent system that uses semantic information from execution to optimize communication structure and reduce token cost, with an actor-critic structure.

Problem

LLM-based multi-agent systems incur high communication cost from redundant interactions, since existing methods optimize communication structures without measuring whether exchanged messages contribute to the final decision.

Approach

Uses semantic information observed during the execution stage to optimize multi-agent communication structure, measuring message contribution to the final decision, within an actor-critic framework to reduce token costs.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add semantic edge judgment plus virtual execution (AgentTailor) over structural/temporal pruningAgentDropout · 6-dataset averageaccuracy: 90.69 -> 91.36Improved
use AgentTailor instead of AgentDropoutmulti-agent system · total tokens across datasetstotal tokens: -21.2% to -61.6%Improved
AgentTailor vs AgentPrune-R on GSM8KAgentTailor · GSM8Kaccuracy: 95.72 -> 93.81Hurt
add virtual execution via EPN (ablation)AgentTailor · HumanEvalpass@1: 87.33 -> 90.31Improved
add virtual execution via EPN (ablation)AgentTailor · MMLUaccuracy: 84.24 -> 86.14Improved
add virtual execution via EPN (ablation)AgentTailor · AQuA/MultiArith/SVAMPaccuracy: AQuA 87.21 -> 87.34No clear effect
add virtual execution via EPN (ablation)AgentTailor · AQuAtotal tokens: 920946.2 -> 724299.8Improved
add relative-contribution normalization of edge utility (ablation)AgentTailor · AQuAaccuracy: 82.05 -> 87.34Improved
add relative-contribution normalization of edge utility (ablation)AgentTailor · MMLUaccuracy: 84.82 -> 86.14Improved
prune high-contribution edges instead of low (reverse pruning) (ablation)AgentTailor · AQuAaccuracy: 87.34 -> 83.20Hurt
remove one top-5 delta edge versus one bottom-5 delta edge (frozen graph) (ablation)AgentTailor · LOEO held-out subsetaccuracy: 79.0/81.0/81.5 vs 79.5/81.0/83.0No clear effect
combine AgentTailor with GPTSwarmGPTSwarm · AQuAaccuracy: 78.16 -> 86.98Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy