上海征集“先进生物制造”重点专项攻关需求
上海征集“先进生物制造”重点专项攻关需求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/80e8b1298b1c8549b4771a9e9cb622ff
https://lightroom.adobe.com/shares/a63f8b425cfea04b8c26659898c6abf9
https://lightroom.adobe.com/shares/18d379eebf502440800ff55a2c55168d
https://lightroom.adobe.com/shares/3f4a2d801d60c649a824fbb6a92147da
https://lightroom.adobe.com/shares/e96c9bdd2e733f40907cfa9c7be3bbb3
https://lightroom.adobe.com/shares/57b94f2fed02ea49ac9efa52939d827c
https://lightroom.adobe.com/shares/278c459246b7f847a965107cb8ba6e11
https://lightroom.adobe.com/shares/35afbbff7333854194fcb56123185271
https://lightroom.adobe.com/shares/e92703a5d8bd3440aa404dc3b1099548
https://lightroom.adobe.com/shares/ab037844925c1b46960cab0fe6405393
https://lightroom.adobe.com/shares/cafe05f813369a4ba1797f7185f4e987
https://lightroom.adobe.com/shares/15aab4e5cf55e240b1b14df740f07b50
https://lightroom.adobe.com/shares/8872bd9864f2c4479d3f7a6e9b1316ea
https://lightroom.adobe.com/shares/5e79da90b3793345b70f9828780ab647
https://lightroom.adobe.com/shares/e2045d1691ff414aa62844c6136dfeb8
https://lightroom.adobe.com/shares/a3f0e19d41877247878c65e0080cfeba
https://lightroom.adobe.com/shares/94f438ef1819124a98c5b18ef7c73886
https://lightroom.adobe.com/shares/d5b26e379a450d40a695075fd2f009ca
https://lightroom.adobe.com/shares/961026bd60f64140bf84fbad60fa0a05
https://lightroom.adobe.com/shares/295156f86d5df04aa9e484170c4a3417
https://lightroom.adobe.com/shares/4cbee901025ad248af0919f5bd627b9b
https://lightroom.adobe.com/shares/783eae642579ac418515bba128b1ba19
https://lightroom.adobe.com/shares/26ee63e05bf8a74ca61c630463a414dd
https://lightroom.adobe.com/shares/3585846052e5c34aa21d64a14d0322e4
https://lightroom.adobe.com/shares/9916299bd9e52146aeacf042d319a99e
https://lightroom.adobe.com/shares/5497aa6004f7ff498a79c1404c0a8bdc
https://lightroom.adobe.com/shares/dd16119aafdc054fb67a10e0e34bcc57
https://lightroom.adobe.com/shares/b48abda4a873a548a480355ed3d371a2
https://lightroom.adobe.com/shares/fa7d2ebd47523842b86fee28f1ae9d80
https://lightroom.adobe.com/shares/937014a2be082c4d9af50a24e20c0577
https://lightroom.adobe.com/shares/8bb59150222bfb4d85f7b75e915ebfbd
https://lightroom.adobe.com/shares/9c6c04d6c80068488bfa804e25b76d10
https://lightroom.adobe.com/shares/15a5d478ef9c584b8897a764c8ac097a
https://lightroom.adobe.com/shares/e3bb1a57b2c6c144838350f3689f7be0
https://lightroom.adobe.com/shares/1f9190027214c54599014abf866102a7
https://lightroom.adobe.com/shares/b03f4a0d0fc6ee4aa1436bc0867da9be
https://lightroom.adobe.com/shares/d8a1155e48d44048b745ece842af30d4
https://lightroom.adobe.com/shares/20993f29a3049949a9ba98d8526c9e28
https://lightroom.adobe.com/shares/8712f3d832870d42b548773c3563f86b
https://lightroom.adobe.com/shares/8e37f859a8f51648a6c22e536c186382
https://lightroom.adobe.com/shares/4e7702200f59294c88e3a86b6a1a1c58
https://lightroom.adobe.com/shares/6bac28646d36224cacf866cfbe2d2a66
https://lightroom.adobe.com/shares/ffeaa55893679d42b3bf128ba22e8375
https://lightroom.adobe.com/shares/31b3fdf1d1328444a790ed2eaa310ca2
https://lightroom.adobe.com/shares/338d1dc9bd57fc42b1d1f9db0ccc694d
https://lightroom.adobe.com/shares/e6b2cc1f8af9434b82102b4c4ed93a01
https://lightroom.adobe.com/shares/0aec61f4aea26541b344669f199864bc
https://lightroom.adobe.com/shares/96e28e14ef4a70489e7ffb460fbefc4e
https://lightroom.adobe.com/shares/f2fde0909de33a4e8aa84fc2e8b8388e
https://lightroom.adobe.com/shares/af1f60241f13264391c0325f857e4c30
沙发！沙发!
页:
[1]