Discuz! Board

 找回密码
 立即注册
搜索
热搜: 活动 交友 discuz
查看: 2|回复: 0

科研人员实现单原子层金属氧化物可控制备

[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 5 天前 | 显示全部楼层 |阅读模式
科研人员实现单原子层金属氧化物可控制备

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_XNjU4NzcwMTg0OA==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTA2OA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMDMzMg==.html
https://v.youku.com/v_show/id_XNjU4NzY5NTc1Mg==.html
https://v.youku.com/v_show/id_XNjU3ODAwMDQwNA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMDMyNA==.html
https://v.youku.com/v_show/id_XNjU4NzY5ODY5Ng==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTA1Ng==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTA2MA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMDMxMg==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTA1Mg==.html
https://v.youku.com/v_show/id_XNjU4NzcwMzQwOA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMDMwOA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMzQwNA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMzQwMA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NzI5Mg==.html
https://v.youku.com/v_show/id_XNjU3ODAwMTYyNA==.html
https://v.youku.com/v_show/id_XNjU3ODAwMDQwMA==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTA0NA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTgyOA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NDMwMA==.html
https://v.youku.com/v_show/id_XNjU4NzY5Mjc5Mg==.html
https://v.youku.com/v_show/id_XNjU4NzY5Mjc5Ng==.html
https://v.youku.com/v_show/id_XNjU4NzY5MjgwMA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTgxNg==.html
https://v.youku.com/v_show/id_XNjU3ODAwMDM4OA==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTAyOA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NzI4MA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMzM3Mg==.html
https://v.youku.com/v_show/id_XNjU4NzY5Mjc4NA==.html
https://v.youku.com/v_show/id_XNjU4NzY5ODY2MA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTgxMg==.html
https://v.youku.com/v_show/id_XNjU3ODAwMDM4NA==.html
https://v.youku.com/v_show/id_XNjU3Nzk5OTAyNA==.html
https://v.youku.com/v_show/id_XNjU4NzY5Mjc1Mg==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTgwNA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTc3Mg==.html
https://v.youku.com/v_show/id_XNjU4NzY5MjczNg==.html
https://v.youku.com/v_show/id_XNjU4NzY5NzI0MA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NDI2NA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMzMzNg==.html
https://v.youku.com/v_show/id_XNjU4NzY5NTY4OA==.html
https://v.youku.com/v_show/id_XNjU3ODAwMDMyMA==.html
https://v.youku.com/v_show/id_XNjU4NzY5ODYyNA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMDIyNA==.html
https://v.youku.com/v_show/id_XNjU4NzcwMTc2OA==.html
https://v.youku.com/v_show/id_XNjU4NzY5MjcwMA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NTY2OA==.html
https://v.youku.com/v_show/id_XNjU4NzY5NTY3Mg==.html
https://v.youku.com/v_show/id_XNjU4NzY5ODYxNg==.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 01:15 , Processed in 0.045200 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