Discuz! Board

 找回密码
 立即注册
搜索
热搜: 活动 交友 discuz
查看: 8|回复: 0

华夏文化科技:正与日本多家数据中心运营商磋商潜在租赁基础设施容量

[复制链接]

7万

主题

7万

帖子

21万

积分

论坛元老

Rank: 8Rank: 8

积分
219781
发表于 4 天前 | 显示全部楼层 |阅读模式
华夏文化科技:正与日本多家数据中心运营商磋商潜在租赁基础设施容量

机器人控制测试发现智能体会误报任务成功

RoboARCade把长序列操作任务放进同一套MuJoCo仿真环境,涉及三种机器人本体。团队扩大了原任务的初始条件范围,让智能体必须在执行过程中利用记忆、判断和视觉信息。综合测试中,GPT-6 Astra得分49.8%,Claude Fable 5.1为43.6%。更丰富的视觉界面带来1.5至6.8个百分点的改善,使用成本同时增至原来的1.7至2.3倍。 Satvik Sharma等人的这项研究收录在11月6日更新的计算机科学预印本列表中。

错误分析显示,主要困难在于选择操作策略,或把正确策略执行得足够精确。智能体在尝试过的子任务中,有24%至39%出现错误宣告成功的情况。在另一个五任务子集上,提示中加入一段成功示范后,Astra从53.5%提高到73.0%。后者属于子集测试,不能当作整套基准上的得分。团队公开了代码、视频与执行轨迹。

https://v.youku.com/v_show/id_XNjU3OTE3OTMyNA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NDgxNg==.html
https://v.youku.com/v_show/id_XNjU3OTE3OTMxNg==.html
https://v.youku.com/v_show/id_XNjU3OTE3OTMwOA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MzMyOA==.html
https://v.youku.com/v_show/id_XNjU3OTE3OTI4NA==.html
https://v.youku.com/v_show/id_XNjU3OTE3OTIyNA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MjM2OA==.html
https://v.youku.com/v_show/id_XNjU4ODg1Mzg1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDg0OA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjI3Mg==.html
https://v.youku.com/v_show/id_XNjU4ODg1NTMzMg==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjI2NA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MzgxNg==.html
https://v.youku.com/v_show/id_XNjU4ODg1Njg4MA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjIzNg==.html
https://v.youku.com/v_show/id_XNjU4ODg1NTI3Ng==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDc1Ng==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjE2OA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjE0NA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDcxMg==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjEwOA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjExMg==.html
https://v.youku.com/v_show/id_XNjU3OTE3NjEwMA==.html
https://v.youku.com/v_show/id_XNjU4ODg0Nzk1Ng==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDY4NA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDY1Ng==.html
https://v.youku.com/v_show/id_XNjU3OTE3MTg5Mg==.html
https://v.youku.com/v_show/id_XNjU4ODg0NzkxNg==.html
https://v.youku.com/v_show/id_XNjU4ODg1MzYwNA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MTg4OA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MTg4NA==.html
https://v.youku.com/v_show/id_XNjU4ODg0NzkwNA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDY0MA==.html
https://v.youku.com/v_show/id_XNjU4ODg0Nzg5Ng==.html
https://v.youku.com/v_show/id_XNjU4ODg0Nzg4OA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MTg1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODg1MjEyMA==.html
https://v.youku.com/v_show/id_XNjU3OTE3NDQwNA==.html
https://v.youku.com/v_show/id_XNjU4ODg1MjA3Ng==.html
https://v.youku.com/v_show/id_XNjU4ODg1MjA2MA==.html
https://v.youku.com/v_show/id_XNjU4ODg0OTM0MA==.html
https://v.youku.com/v_show/id_XNjU4ODg0OTMzNg==.html
https://v.youku.com/v_show/id_XNjU3OTE2ODcwNA==.html
https://v.youku.com/v_show/id_XNjU3OTE2ODcwMA==.html
https://v.youku.com/v_show/id_XNjU4ODg0NjM1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODg0OTI5Ng==.html
https://v.youku.com/v_show/id_XNjU4ODg1MDUyNA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MDI0MA==.html
https://v.youku.com/v_show/id_XNjU3OTE3MDIzNg==.html
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

Archiver|手机版|小黑屋|DiscuzX

GMT+8, 2026-11-15 14:43 , Processed in 0.042966 second(s), 18 queries .

Powered by Discuz! X3.4

Copyright © 2001-2020, Tencent Cloud.

快速回复 返回顶部 返回列表