Training LLM-Based Agents with Synthetic Self-Reflected Trajectories and Partial Masking

• URL: http://arxiv.org/abs/2505.20023v1
• Date: Mon, 26 May 2025 14:11:12 GMT
• Title: Training LLM-Based Agents with Synthetic Self-Reflected Trajectories and Partial Masking
• Authors: Yihan Chen, Benfeng Xu, Xiaorui Wang, Yongdong Zhang, Zhendong Mao,
• Abstract summary: We propose STeP, a novel method for improving LLM-based agent training.<n>We synthesize self-reflected trajectories that include reflections and corrections of error steps.<n>Experiments demonstrate that our method improves agent performance across three representative tasks.
• Score: 61.61356842567952
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Autonomous agents, which perceive environments and take actions to achieve goals, have become increasingly feasible with the advancements in large language models (LLMs). However, current powerful agents often depend on sophisticated prompt engineering combined with closed-source LLMs like GPT-4. Although training open-source LLMs using expert trajectories from teacher models has yielded some improvements in agent capabilities, this approach still faces limitations such as performance plateauing and error propagation. To mitigate these challenges, we propose STeP, a novel method for improving LLM-based agent training. We synthesize self-reflected trajectories that include reflections and corrections of error steps, which enhance the effectiveness of LLM agents in learning from teacher models, enabling them to become agents capable of self-reflecting and correcting. We also introduce partial masking strategy that prevents the LLM from internalizing incorrect or suboptimal steps. Experiments demonstrate that our method improves agent performance across three representative tasks: ALFWorld, WebShop, and SciWorld. For the open-source model LLaMA2-7B-Chat, when trained using self-reflected trajectories constructed with Qwen1.5-110B-Chat as the teacher model, it achieves comprehensive improvements with less training data compared to agents trained exclusively on expert trajectories.

Related papers

• Training Agents with Weakly Supervised Feedback from Large Language Models [19.216542820742607]
  This paper introduces a novel training method for LLM-based agents using weakly supervised signals from a critic LLM.<n>Our agents are trained in iterative manner, where they initially generate trajectories through environmental interaction.<n>Tests on the API-bank dataset show consistent improvement in our agents' capabilities and comparable performance to GPT-4.
  arXiv  Detail & Related papers  (2024-11-29T08:47:04Z)
• SAC-GLAM: Improving Online RL for LLM agents with Soft Actor-Critic and Hindsight Relabeling [29.29604779151457]
  This paper presents and studies an adaptation of Soft Actor-Critic and hindsight relabeling to LLM agents. Our method paves the path towards autotelic LLM agents that learn online but can also outperform on-policy methods in more classic multi-goal RL environments.
  arXiv  Detail & Related papers  (2024-10-16T11:59:27Z)
• From Words to Actions: Unveiling the Theoretical Underpinnings of LLM-Driven Autonomous Systems [59.40480894948944]
  Large language model (LLM) empowered agents are able to solve decision-making problems in the physical world. Under this model, the LLM Planner navigates a partially observable Markov decision process (POMDP) by iteratively generating language-based subgoals via prompting. We prove that the pretrained LLM Planner effectively performs Bayesian aggregated imitation learning (BAIL) through in-context learning.
  arXiv  Detail & Related papers  (2024-05-30T09:42:54Z)
• Towards Efficient LLM Grounding for Embodied Multi-Agent Collaboration [70.09561665520043]
  We propose a novel framework for multi-agent collaboration that introduces Reinforced Advantage feedback (ReAd) for efficient self-refinement of plans. We provide theoretical analysis by extending advantage-weighted regression in reinforcement learning to multi-agent systems. Experiments on Over-AI and a difficult variant of RoCoBench show that ReAd surpasses baselines in success rate, and also significantly decreases the interaction steps of agents.
  arXiv  Detail & Related papers  (2024-05-23T08:33:19Z)
• Enhancing the General Agent Capabilities of Low-Parameter LLMs through Tuning and Multi-Branch Reasoning [56.82041895921434]
  Open-source pre-trained Large Language Models (LLMs) exhibit strong language understanding and generation capabilities. When used as agents for dealing with complex problems in the real world, their performance is far inferior to large commercial models such as ChatGPT and GPT-4.
  arXiv  Detail & Related papers  (2024-03-29T03:48:12Z)
• EnvGen: Generating and Adapting Environments via LLMs for Training Embodied Agents [65.38474102119181]
  We propose EnvGen, a framework to adaptively create training environments. We train a small RL agent in a mixture of the original and LLM-generated environments. We find that a small RL agent trained with EnvGen can outperform SOTA methods, including a GPT-4 agent, and learns long-horizon tasks significantly faster.
  arXiv  Detail & Related papers  (2024-03-18T17:51:16Z)
• Trial and Error: Exploration-Based Trajectory Optimization for LLM Agents [49.85633804913796]
  We present an exploration-based trajectory optimization approach, referred to as ETO. This learning method is designed to enhance the performance of open LLM agents. Our experiments on three complex tasks demonstrate that ETO consistently surpasses baseline performance by a large margin.
  arXiv  Detail & Related papers  (2024-03-04T21:50:29Z)
• Offline Training of Language Model Agents with Functions as Learnable Weights [39.88545362699836]
  We present a novel paradigm of training Large Language Models (LLMs) agents without modifying the LLM weights. We develop Agentr that employs the LLM to update agents' functions and devise an agent training algorithm with two strategies, roll-back, and early-stop. With extensive experiments, we showcase that the agent training paradigm could significantly improve the performance of representative LLM agents.
  arXiv  Detail & Related papers  (2024-02-17T18:31:21Z)
• Large Language Model as a Policy Teacher for Training Reinforcement Learning Agents [16.24662355253529]
  Large Language Models (LLMs) can address sequential decision-making tasks through the provision of high-level instructions. LLMs lack specialization in tackling specific target problems, particularly in real-time dynamic environments. We introduce a novel framework that addresses these challenges by training a smaller, specialized student RL agent using instructions from an LLM-based teacher agent.
  arXiv  Detail & Related papers  (2023-11-22T13:15:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.