定价 1748 元：索尼官宣因生产问题延迟发售 PULSE Elevate 音箱
定价 1748 元：索尼官宣因生产问题延迟发售 PULSE Elevate 音箱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/da75ecec368acf498c9aef0c8ccdd9b3
https://lightroom.adobe.com/shares/6f2856515111644db0e5c11927829d16
https://lightroom.adobe.com/shares/14d797d042255446928899d385adc67e
https://lightroom.adobe.com/shares/c200e06e97b75143b6f5092d1215b7c3
https://lightroom.adobe.com/shares/1c9b2791ab11cb4eb26285be782640b6
https://lightroom.adobe.com/shares/aa186a9186c2df48a6f74e2354cf3b98
https://lightroom.adobe.com/shares/e8c1bf20852cf7459e64f0fff7097478
https://lightroom.adobe.com/shares/c65918d86a577d4e80acdefce8fae44c
https://lightroom.adobe.com/shares/8fafa1d3e4860446b720efdbe47c9de8
https://lightroom.adobe.com/shares/03d2824cde08ad43bffb4be3fd4fa107
https://lightroom.adobe.com/shares/7636176fa10db1488124b12b8fd5943c
https://lightroom.adobe.com/shares/766093eb76fcc44eb5739e1fcaa3f5b2
https://lightroom.adobe.com/shares/4112615df5199c4b9908fe233c136511
https://lightroom.adobe.com/shares/b8c5de29fa635f4fb6a1fc1b3360cb89
https://lightroom.adobe.com/shares/a723265a559c2742963a901c3201a846
https://lightroom.adobe.com/shares/18c98e13bc077344a38a5ba848cbcc79
https://lightroom.adobe.com/shares/2bd37ed3c277df4992b5f45883616d11
https://lightroom.adobe.com/shares/9b073cc1c7111747801ea75872bd4386
https://lightroom.adobe.com/shares/2e5ba8926643844d9fce9b8feb539e09
https://lightroom.adobe.com/shares/daf3d97a6d76074faae597156fb11b84
https://lightroom.adobe.com/shares/ce472874c77fde43867835146c2ab276
https://lightroom.adobe.com/shares/866a639ba632fe46bef0cfcec0e9a5c7
https://lightroom.adobe.com/shares/3a51d64bd4f1ae4c996fb2caae45c5ca
https://lightroom.adobe.com/shares/a67ba87506737747bfa50daed089f457
https://lightroom.adobe.com/shares/0c5ba101165cdf4ab8a897b08fc60d90
https://lightroom.adobe.com/shares/99deb5ceb54ed84f9d2bdb6315d39ca2
https://lightroom.adobe.com/shares/27854ebaea78fd4db61be18a7ba7f6ac
https://lightroom.adobe.com/shares/6109c16ba13a9440a17621689f628a37
https://lightroom.adobe.com/shares/3f0687bfc6d83a4787f53ebfe2b6b01a
https://lightroom.adobe.com/shares/5a404d608154954891572dbc4b905a10
https://lightroom.adobe.com/shares/a12f2edcbacd204d88802f0e76c72186
https://lightroom.adobe.com/shares/4e59304b70d38940a2ae39a0e99ab281
https://lightroom.adobe.com/shares/67b7ac775d96514ca21c81983a88895f
https://lightroom.adobe.com/shares/be190f761457bb4bb13b1fc8ba69d41b
https://lightroom.adobe.com/shares/79b38ccd1418a0409e10a43a224e3c8d
https://lightroom.adobe.com/shares/d390294d53cfdf46865b804e74f0c929
https://lightroom.adobe.com/shares/dd8ebf7534eb1b4799485c01c6707b31
https://lightroom.adobe.com/shares/9d40f9d3f7d4ae45b2f9ad6eabf2bd2f
https://lightroom.adobe.com/shares/313c8eb23b18404f864df7ec9661b241
https://lightroom.adobe.com/shares/40a0d898a2ff06468a4dbf611b1aaece
https://lightroom.adobe.com/shares/035ed883f60aa84db71383da537e7678
https://lightroom.adobe.com/shares/de547854bddaf84c9fda9b6c3b73038c
https://lightroom.adobe.com/shares/065bd5af9de8974b84480580fdebb226
https://lightroom.adobe.com/shares/9c7f6522afd2c34e847eef4b49bf640b
https://lightroom.adobe.com/shares/cfbb7c9bc26c4640812d5fee14d3724f
https://lightroom.adobe.com/shares/d389c3e66f91e74694115ed09bd7e996
https://lightroom.adobe.com/shares/1fd3dc3477f1a44583f49011923407a8
https://lightroom.adobe.com/shares/63e027b901fcb64590dc4c06d8d245c5
https://lightroom.adobe.com/shares/a7712a5599c5f34783a9e6efa5296edf
https://lightroom.adobe.com/shares/9f55da6a8f3ff041984b826296bff89d
我抢、我抢、我抢沙发~
页:
[1]