MoRAgent: Parameter Efficient Agent Tuning with Mixture-of-Roles

TL;DR

MoRAgent addresses the inefficiency of full-parameter fine-tuning for LLM-based agents by introducing a triad of roles—reasoner, executor, and summarizer—and a Mixture-of-Roles with role-specific LoRA adapters. The method couples a rule-based gate and token-aware routers to allocate trainable capacity across roles, guided by balance and orthogonality losses, and a multi-role data generation pipeline to create reliable, role-consistent training data. Across multiple models and benchmarks (StableToolBench, BFCL, GSM8K, MATH), MoRAgent achieves substantial gains with modest trainable parameters, demonstrating the practicality of PEFT for agent tasks. The work advances the state of PEFT for tool-use and function-calling tasks and provides a scalable path toward efficient, flexible agent systems.

Abstract

Despite recent advancements of fine-tuning large language models (LLMs) to facilitate agent tasks, parameter-efficient fine-tuning (PEFT) methodologies for agent remain largely unexplored. In this paper, we introduce three key strategies for PEFT in agent tasks: 1) Inspired by the increasingly dominant Reason+Action paradigm, we first decompose the capabilities necessary for the agent tasks into three distinct roles: reasoner, executor, and summarizer. The reasoner is responsible for comprehending the user's query and determining the next role based on the execution trajectory. The executor is tasked with identifying the appropriate functions and parameters to invoke. The summarizer conveys the distilled information from conversations back to the user. 2) We then propose the Mixture-of-Roles (MoR) framework, which comprises three specialized Low-Rank Adaptation (LoRA) groups, each designated to fulfill a distinct role. By focusing on their respective specialized capabilities and engaging in collaborative interactions, these LoRAs collectively accomplish the agent task. 3) To effectively fine-tune the framework, we develop a multi-role data generation pipeline based on publicly available datasets, incorporating role-specific content completion and reliability verification. We conduct extensive experiments and thorough ablation studies on various LLMs and agent benchmarks, demonstrating the effectiveness of the proposed method. This project is publicly available at https://mor-agent.github.io.

Figures (12)

• Figure 1: Workflow example of multiple roles collaborate to accomplish one agent task.
• Figure 2: The framework of our method. The capabilities necessary for agent are decomposed into three distinct roles: reasoner, executor, and summarizer. Each role consists of different number of LoRAs according to their learning difficulty. The rule-based role-aware gate and learnable token-aware routers are introduced to more reasonably allocate LoRAs.
• Figure 3: The JSON format on different scenarios in our fine-tuning datasets.
• Figure 4: The loss and levenshtein accuracy of respective roles with different number of LoRAs.
• Figure 5: Visualization of the similarity of routed LoRAs without (top) and with (bottom) orthogonal loss. From left to right are reasoner, executor and summarizer.