上海发布“人工智能+交通运输”行动方案
上海发布“人工智能+交通运输”行动方案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://lightroom.adobe.com/shares/ef9bb3b65aeffb478b938f96b3286920
https://lightroom.adobe.com/shares/671749aa68fbd24882d3c98c8f4d31de
https://lightroom.adobe.com/shares/114db1c71322be4c87118f5cc0130fc8
https://lightroom.adobe.com/shares/5f58c6ba377cc344820e816608efa6eb
https://lightroom.adobe.com/shares/b0f360ede81ef842ae58033feb411ec3
https://lightroom.adobe.com/shares/bbd2acf630cd3b4ab950e98578e5e225
https://lightroom.adobe.com/shares/fd283e70692e1143ab4d3da3542e3694
https://lightroom.adobe.com/shares/0aae46eeaae3e64ba364de6ebfa2af1f
https://lightroom.adobe.com/shares/22e6ed7186e1cd47a4ef8c4a3bf2e235
https://lightroom.adobe.com/shares/3bba784710de994c9585761582057266
https://lightroom.adobe.com/shares/0afa6bebdca29e4fb768bb8b99d0c957
https://lightroom.adobe.com/shares/bdf0e6214dce6644acbaa2d2519d9ae0
https://lightroom.adobe.com/shares/9d7e609acf18aa49966a689f7de1f946
https://lightroom.adobe.com/shares/e273bdd4ed3f874199c1a788b768bb70
https://lightroom.adobe.com/shares/766da753f660114996517559ef2a1470
https://lightroom.adobe.com/shares/f54e0c38cbd8ad4d9fd89014522254c9
https://lightroom.adobe.com/shares/d042a84c2d001849a4f3e9d894f07383
https://lightroom.adobe.com/shares/e59ec90336b8b44497d06ec9e67197ce
https://lightroom.adobe.com/shares/e8822f05f2f6fb4ca030ed1a91ab8555
https://lightroom.adobe.com/shares/fa4135359756304fa52b9d30a0d44d2e
https://lightroom.adobe.com/shares/152a2d391b6d2a4998cfd489e4f22136
https://lightroom.adobe.com/shares/526975395975814f9f6b876e50fb0a7b
https://lightroom.adobe.com/shares/d3dcda65b9535441a32308990c21181d
https://lightroom.adobe.com/shares/3e78f0f0a0edf0488c4f0bf5e993c22b
https://lightroom.adobe.com/shares/12365636b90382499d31f539e270e0c1
https://lightroom.adobe.com/shares/228928e0eb11cc4fb54fc971a12d962e
https://lightroom.adobe.com/shares/cd4a8febde1c3e4bbe2dc9114679ef74
https://lightroom.adobe.com/shares/fa043b4cfe1a174b96cba2e0c8074507
https://lightroom.adobe.com/shares/0171d12b17aec14c9c78b19b04972b35
https://lightroom.adobe.com/shares/e9f2a252fa1a314f8b1ee0e40a196112
https://lightroom.adobe.com/shares/1888a829ea12be4aba1514fbe64bf62d
https://lightroom.adobe.com/shares/0092bdf1b7676b46a93fe3caeaaffeb0
https://lightroom.adobe.com/shares/62711b176f71d943b578046cd1ca37c2
https://lightroom.adobe.com/shares/3f548a98ab6c5a459db05882226b5475
https://lightroom.adobe.com/shares/fb90d442c88ba24aa7e4d9e8a8d682eb
https://lightroom.adobe.com/shares/e409ee1be3dfbc44879611914ee60201
https://lightroom.adobe.com/shares/976fb3fcd512474c83d530b463d01efd
https://lightroom.adobe.com/shares/b6139289e44c424285c6ad43d1c1e821
https://lightroom.adobe.com/shares/ca7bb6f57b054644848e86084bd7513b
https://lightroom.adobe.com/shares/697f11ddaa0e474c84ee8768f00982ba
https://lightroom.adobe.com/shares/f8f75224651f264c8fbcb1a1970b135f
https://lightroom.adobe.com/shares/5d234ea818d60a4e8d16f377b0d3495f
https://lightroom.adobe.com/shares/b6270c1d4c701b489fd876b2c155dab8
https://lightroom.adobe.com/shares/02a75378864649418d4d6ea9a226cb20
https://lightroom.adobe.com/shares/6d5dcbb3da12ae488ef421308140b132
https://lightroom.adobe.com/shares/cbc9dc8441fc33499ab8242ff3f73ad8
https://lightroom.adobe.com/shares/83ccc063ebffc446b541af29161853fb
https://lightroom.adobe.com/shares/54d3ed469dc574499e09d32a87ab8635
https://lightroom.adobe.com/shares/463873d498e4e04dbde444cda719c783
https://lightroom.adobe.com/shares/4862b20869afae418ae1445149cde29b
小手一抖，沙发到手！
页:
[1]