Discuz! Board

 找回密码
 立即注册
搜索
热搜: 活动 交友 discuz
查看: 5|回复: 0

百视通亮相2026移动网络大会 ALL In AI探索视听新生态

[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
百视通亮相2026移动网络大会 ALL In AI探索视听新生态

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_XNjU4ODQ1MTEwOA==.html
https://v.youku.com/v_show/id_XNjU3ODczOTg2OA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MTA3Mg==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MjQ5Mg==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MjQ4NA==.html
https://v.youku.com/v_show/id_XNjU3ODczOTgyNA==.html
https://v.youku.com/v_show/id_XNjU3ODc0MTI2NA==.html
https://v.youku.com/v_show/id_XNjU3ODczNjcwNA==.html
https://v.youku.com/v_show/id_XNjU3ODczNjY5Mg==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MzY2NA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MzY1Mg==.html
https://v.youku.com/v_show/id_XNjU3ODc0MTE4NA==.html
https://v.youku.com/v_show/id_XNjU3ODc0MTEzMg==.html
https://v.youku.com/v_show/id_XNjU4ODQ0OTQwMA==.html
https://v.youku.com/v_show/id_XNjU4ODQ0Nzk2OA==.html
https://v.youku.com/v_show/id_XNjU3ODc0MTEyMA==.html
https://v.youku.com/v_show/id_XNjU3ODc0MTExNg==.html
https://v.youku.com/v_show/id_XNjU3ODczODE0NA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MjIxNg==.html
https://v.youku.com/v_show/id_XNjU3ODc0MjQzMg==.html
https://v.youku.com/v_show/id_XNjU3ODczODEwOA==.html
https://v.youku.com/v_show/id_XNjU3ODczMzQzNg==.html
https://v.youku.com/v_show/id_XNjU4ODQ0OTI0MA==.html
https://v.youku.com/v_show/id_XNjU3ODc0MDk3Ng==.html
https://v.youku.com/v_show/id_XNjU3ODczNjQ4NA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MjEwNA==.html
https://v.youku.com/v_show/id_XNjU3ODczMzM5Mg==.html
https://v.youku.com/v_show/id_XNjU3ODczOTQ5Ng==.html
https://v.youku.com/v_show/id_XNjU3ODczNTEwMA==.html
https://v.youku.com/v_show/id_XNjU3ODczOTQ4NA==.html
https://v.youku.com/v_show/id_XNjU3ODczNTA3Ng==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MDYxNg==.html
https://v.youku.com/v_show/id_XNjU3ODczOTQzMg==.html
https://v.youku.com/v_show/id_XNjU3ODczNTAyNA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MDU4OA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MjAyNA==.html
https://v.youku.com/v_show/id_XNjU4ODQ1MDU1Mg==.html
https://v.youku.com/v_show/id_XNjU3ODczNDk4MA==.html
https://v.youku.com/v_show/id_XNjU3ODczNDk2MA==.html
https://v.youku.com/v_show/id_XNjU3ODczNzg4MA==.html
https://v.youku.com/v_show/id_XNjU3ODczOTM2OA==.html
https://v.youku.com/v_show/id_XNjU3ODczNzg0NA==.html
https://v.youku.com/v_show/id_XNjU3ODczMzIzMg==.html
https://v.youku.com/v_show/id_XNjU4ODQ0ODk5Ng==.html
https://v.youku.com/v_show/id_XNjU3ODczNjE2OA==.html
https://v.youku.com/v_show/id_XNjU4ODQ0ODk4NA==.html
https://v.youku.com/v_show/id_XNjU3ODczNjE1Ng==.html
https://v.youku.com/v_show/id_XNjU3ODczNjEzMg==.html
https://v.youku.com/v_show/id_XNjU4ODQ0NzU0NA==.html
https://v.youku.com/v_show/id_XNjU4ODQ0NDY4OA==.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 13:52 , Processed in 0.043660 second(s), 19 queries .

Powered by Discuz! X3.4

Copyright © 2001-2020, Tencent Cloud.

快速回复 返回顶部 返回列表