RF-Agent: Automated Reward Function Design via Language Agent Tree Search

2026-02-27

AI Summary

This work addresses the challenges of inefficient reward function design in low-level control tasks, where historical feedback is underutilized and search efficiency is limited. It proposes a novel paradigm by modeling reward function design as a sequential decision-making problem, leveraging large language models as linguistic agents and integrating Monte Carlo Tree Search (MCTS) to optimize the generation process. This approach significantly enhances contextual reasoning capabilities and search efficiency. Evaluated across 17 diverse low-level control tasks, the method demonstrates strong performance, validating its effectiveness and generalization ability. The study thus establishes a new framework for automated reward function design in reinforcement learning settings.

Technology Category

• Search and Optimization: Learning to Search
• Multiagent Systems: Mechanism Design
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Designing efficient reward functions for low-level control tasks is a challenging problem. Recent research aims to reduce reliance on expert experience by using Large Language Models (LLMs) with task information to generate dense reward functions. These methods typically rely on training results as feedback, iteratively generating new reward functions with greedy or evolutionary algorithms. However, they suffer from poor utilization of historical feedback and inefficient search, resulting in limited improvements in complex control tasks. To address this challenge, we propose RF-Agent, a framework that treats LLMs as language agents and frames reward function design as a sequential decision-making process, enhancing optimization through better contextual reasoning. RF-Agent integrates Monte Carlo Tree Search (MCTS) to manage the reward design and optimization process, leveraging the multi-stage contextual reasoning ability of LLMs. This approach better utilizes historical information and improves search efficiency to identify promising reward functions. Outstanding experimental results in 17 diverse low-level control tasks demonstrate the effectiveness of our method. The source code is available at https://github.com/deng-ai-lab/RF-Agent.

Problem

• reward function design
• low-level control tasks
• historical feedback utilization
• search efficiency
• complex control tasks

Innovation

• Reward Function Design
• Language Agent
• Monte Carlo Tree Search
• Large Language Models
• Sequential Decision-Making