Discuz! Board

 找回密码
 立即注册
搜索
热搜: 活动 交友 discuz
查看: 3|回复: 0

以四川腊肉为灵感的国产威士忌亮相进博会

[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
以四川腊肉为灵感的国产威士忌亮相进博会

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_XNjU4ODkxOTcwOA==.html
https://v.youku.com/v_show/id_XNjU3OTIzNzMyOA==.html
https://v.youku.com/v_show/id_XNjU4ODkxNzk4OA==.html
https://v.youku.com/v_show/id_XNjU4ODkxMzQ1Ng==.html
https://v.youku.com/v_show/id_XNjU4ODkxNjQzMg==.html
https://v.youku.com/v_show/id_XNjU3OTIzNDM4MA==.html
https://v.youku.com/v_show/id_XNjU3OTIzNzA5Ng==.html
https://v.youku.com/v_show/id_XNjU4ODkxNDU1Ng==.html
https://v.youku.com/v_show/id_XNjU3OTIzMjc3Ng==.html
https://v.youku.com/v_show/id_XNjU4ODkxNDUxNg==.html
https://v.youku.com/v_show/id_XNjU3OTIzNTkwMA==.html
https://v.youku.com/v_show/id_XNjU3OTIzMTExNg==.html
https://v.youku.com/v_show/id_XNjU3OTIzMjUyMA==.html
https://v.youku.com/v_show/id_XNjU4ODkxMTM4MA==.html
https://v.youku.com/v_show/id_XNjU3OTIzNTYzMg==.html
https://v.youku.com/v_show/id_XNjU4ODkwMzk2MA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNjk2OA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNjgzMg==.html
https://v.youku.com/v_show/id_XNjU4ODkwNTQ4NA==.html
https://v.youku.com/v_show/id_XNjU4ODkwOTY2MA==.html
https://v.youku.com/v_show/id_XNjU4ODkwOTY1Mg==.html
https://v.youku.com/v_show/id_XNjU4ODkwMzg2NA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNjY4MA==.html
https://v.youku.com/v_show/id_XNjU3OTIzMDU2NA==.html
https://v.youku.com/v_show/id_XNjU3OTIyOTEzMg==.html
https://v.youku.com/v_show/id_XNjU3OTIzMDQzNg==.html
https://v.youku.com/v_show/id_XNjU3OTIyNzg2NA==.html
https://v.youku.com/v_show/id_XNjU4ODkwMDY0OA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNTE3Mg==.html
https://v.youku.com/v_show/id_XNjU4ODkwNTE2NA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNTE2MA==.html
https://v.youku.com/v_show/id_XNjU4ODkwNTE1Ng==.html
https://v.youku.com/v_show/id_XNjU3OTIxOTk5Mg==.html
https://v.youku.com/v_show/id_XNjU3OTIxOTk4OA==.html
https://v.youku.com/v_show/id_XNjU4ODkwMzUxNg==.html
https://v.youku.com/v_show/id_XNjU3OTIxOTk2OA==.html
https://v.youku.com/v_show/id_XNjU3OTIyNTg3Ng==.html
https://v.youku.com/v_show/id_XNjU3OTIyNDI4OA==.html
https://v.youku.com/v_show/id_XNjU3OTIyMTE0MA==.html
https://v.youku.com/v_show/id_XNjU3OTIyNzI2NA==.html
https://v.youku.com/v_show/id_XNjU3OTIyNTY2NA==.html
https://v.youku.com/v_show/id_XNjU3OTIyMjYyNA==.html
https://v.youku.com/v_show/id_XNjU3OTIyNzIyNA==.html
https://v.youku.com/v_show/id_XNjU3OTIxNTExMg==.html
https://v.youku.com/v_show/id_XNjU3OTIxNjY4OA==.html
https://v.youku.com/v_show/id_XNjU4ODg5NzU5Ng==.html
https://v.youku.com/v_show/id_XNjU3OTIxNTA4MA==.html
https://v.youku.com/v_show/id_XNjU3OTIxNDkzMg==.html
https://v.youku.com/v_show/id_XNjU3OTIxODEyOA==.html
https://v.youku.com/v_show/id_XNjU4ODg5NTY2MA==.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 13:48 , Processed in 0.046802 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