English(EN) STAPO: Selective Trajectory-Aware Policy Optimization for LLM Agent Training

新的STAPO框架通过减少轨迹忽略来改进LLM代理训练

作者 PulseAugur 编辑部 · [3 个来源] · 2026-07-06 11:50

研究人员开发了STAPO（选择性轨迹感知策略优化），一个新颖的层次强化学习框架，旨在改进大型语言模型（LLM）代理的训练。STAPO解决了“轨迹忽略”问题，即代理因稀疏或延迟的奖励而失去对任务目标的关注。通过利用新颖的“归一化熵”指标，STAPO识别并优化与被忽略轨迹相关的异常步骤，增强了代理的意识和训练稳定性。在ALFWorld、WebShop和Search-Augmented QA基准上的实验表明，STAPO取得了最先进的性能，并有效缓解了轨迹忽略问题。 AI

影响通过提高对长时任务的关注来增强LLM代理训练，并可能带来更强大、更具目标导向性的AI代理。

排序理由该集群描述了一篇详细介绍LLM代理训练新方法的最新研究论文。

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。我们如何撰写摘要 →

报道来源 [3]

arXiv cs.AI TIER_1 English(EN) · Qiuyi Qi, Tian Liang, Mutian Bao, Jinjian Zhang, Dongnan Liu, Wei Zhou, Linjian Mo, Ming Kong, Jie Liu, Feng Zhang, Qiang Zhu · 2026-07-07 04:00

STAPO：用于LLM智能体训练的选择性轨迹感知策略优化

arXiv:2607.04963v1 Announce Type: new Abstract: Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task …
arXiv cs.AI TIER_1 English(EN) · Qiang Zhu · 2026-07-06 11:50

STAPO：用于 LLM Agent 训练的选择性轨迹感知策略优化

Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task goal and interaction history at intermediate ste…
Hugging Face Daily Papers TIER_1 English(EN) · 2026-07-06 11:50

STAPO：用于 LLM 代理训练的选择性轨迹感知策略优化

Reinforcement Learning (RL) is the dominant paradigm for training Large Language Model (LLM) agents on long-horizon tasks. However, sparse and delayed rewards often lead to trajectory neglect, in which agents lose focus on the task goal and interaction history at intermediate ste…

报道来源 [3]

STAPO：用于LLM智能体训练的选择性轨迹感知策略优化

STAPO：用于 LLM Agent 训练的选择性轨迹感知策略优化

STAPO：用于 LLM 代理训练的选择性轨迹感知策略优化

相关实体

相关话题