PulseAugur
实时 11:08:50
English(EN) Look Ahead Before You Distill: Future Trajectory Validation of Teacher Guidance for Agentic On-Policy Distillation

新的蒸馏方法FTB通过验证教师指导来提高智能体性能

研究人员开发了一种名为FutureBridge-OPD (FTB) 的新方法,以改进智能体任务的策略内蒸馏 (OPD)。标准的OPD在教师访问的状态上监督学生,但学生的偏差可能导致随时间的指导效果不佳。FTB通过观察学生的后续轨迹来评估在高分歧状态下教师指导的好处,从而解决这个问题。在ALFWorld、WebShop和ScienceWorld上的实验中,FTB在现有方法上显示出显著的性能提升,当使用Qwen3-32B作为Qwen3-1.7B的教师时,其性能分别比标准的OPD和TCOD平均高出16.6和7.6个百分点。 AI

影响 通过提高从大模型到小模型知识转移的有效性,增强了智能体AI的性能。

排序理由 学术论文,详细介绍了一种改进智能体AI任务中蒸馏技术的新方法。

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

新的蒸馏方法FTB通过验证教师指导来提高智能体性能

报道来源 [2]

  1. arXiv cs.CL TIER_1 English(EN) · Chishui Chen, Yaoyou Fan, Te Sun, Yi Yang, Chenghao Sun, Delin Mao, Hongbo Qiao, Zuowei Zhang, Junxi Wang, Chenxing Sun, Yangen Hu, Lu Pan, Xuyang Liu, Linfeng Zhang ·

    Look Ahead Before You Distill: Future Trajectory Validation of Teacher Guidance for Agentic On-Policy Distillation

    arXiv:2608.01953v1 Announce Type: new Abstract: On-policy distillation (OPD) provides teacher supervision on states visited by the student, reducing the distribution gap between training and inference. However, in multi-turn agentic tasks, student deviations may accumulate over t…

  2. Hugging Face Daily Papers TIER_1 English(EN) ·

    Look Ahead Before You Distill: Future Trajectory Validation of Teacher Guidance for Agentic On-Policy Distillation

    On-policy distillation (OPD) provides teacher supervision on states visited by the student, reducing the distribution gap between training and inference. However, in multi-turn agentic tasks, student deviations may accumulate over time, gradually moving the trajectory away from s…