arXivDaily arXiv每日学术速递 周一至周五更新

大厂专区

Xiaohongshu(小红书)

2026-08-20 至 2026-08-20 共收录 1
2608.18852 2026-08-20 cs.AI 新提交

SkillGate: Training In-Policy Skill Selection in Long-Horizon Agents

SkillGate:训练长视距智能体中的策略内技能选择

Qingyao Li, Wenxiang Jiao, Shuai Shao, Kangning Zhang, Yuan Lu, Yi Guo, Weiwen Liu, Weinan Zhang, Yong Yu

机构 * Shanghai Jiao Tong University(上海交通大学) Xiaohongshu Inc.(小红书科技有限公司)

AI总结 SkillGate 针对长视距智能体策略内技能选择的选择器信用匮乏问题,通过划分信用通道提升了 9B 策略的试验成功率,减少了误导性候选接触并降低技能读取次数。

Comments 17 pages, 6 figures, 6 tables. Code: this https URL (https://github.com/DeepExperience/SkillGate). Models: this https URL (https://huggingface.co/simonlqy/SkillGate-9B)

详情

展开后加载摘要…

URL PDF HTML 收藏