GraphPlanner: Graph Memory-Augmented Agentic Routing for Multi-Agent LLMs

TL;DR

GraphPlanner introduces a graph memory-augmented routing framework for multi-agent LLMs, enhancing task accuracy, efficiency, and adaptability through reinforcement learning and historical memory integration.

Contribution

It presents a novel heterogeneous graph-based agentic router that generates workflows for multi-agent LLMs, supporting inductive and transductive inference with improved performance and efficiency.

Findings

Outperforms existing routers, increasing accuracy by up to 9.3%.

Reduces GPU memory usage from 186.26 GiB to 1.04 GiB.

Demonstrates strong zero-shot generalization to unseen tasks and models.

Abstract

LLM routing has achieved promising results in integrating the strengths of diverse models while balancing efficiency and performance. However, to support more realistic and challenging applications, routing must extend into agentic LLM settings, where task planning, multi-round cooperation among heterogeneous agents, and memory utilization are indispensable. To address this gap, we propose GraphPlanner, a heterogeneous graph memory-augmented agentic router for multi-agent LLMs that generates routing workflows for each query and supports both inductive and transductive inference. GraphPlanner formulates workflow generation as a Markov Decision Process (MDP), where at each step it selects both the LLM backbone and the agent role, including Planner, Executor, and Summarizer. By leveraging a heterogeneous graph, denoted as GARNet, to capture interaction memories among queries, agents, and responses, GraphPlanner integrates historical memory and workflow memory into richer state representations. The entire pipeline is optimized with reinforcement learning, jointly improving task-specific performance and computational efficiency. We evaluate GraphPlanner across 14 diverse LLM tasks and demonstrate that: (1) GraphPlanner outperforms strong single-round and multi-round routers, improving accuracy by up to 9.3% while reducing GPU cost from 186.26 GiB to 1.04 GiB; (2) GraphPlanner generalizes robustly to unseen tasks and LLMs, exhibiting strong zero-shot capabilities; and (3) GraphPlanner effectively leverages historical memories, supporting both inductive and transductive inference for more adaptive routing. Our code for GraphPlanner is released at https://github.com/ulab-uiuc/GraphPlanner.