图 11. 扩展到新的反事实概念 “更重”
参考文献
[1] The Neuro-Symbolic Concept Learner: Interpreting Scenes, Words, and Sentences From Natural Supervision. Jiayuan Mao, Chuang Gan, Pushmeet Kohli, Joshua B. Tenenbaum, and Jiajun Wu. ICLR 2019.
[2] CLEVRER: CoLlision Events for Video REpresentation and Reasoning. Kexin Yi, Chuang Gan, Yunzhu Li, Pushmeet Kohli, Jiajun Wu, Antonio Torralba, and Joshua B. Tenenbaum. ICLR 2020.
[3] Object-based attention for spatio-temporal reasoning: Outperforming neuro-symbolic models with flexible distributed architectures. David Ding, Hill Felix, Santoro Adam, and Botvinick Matt. arXiv 2020.
[4] Grounding Physical Concepts of Objects and Events Through Dynamic Visual Reasoning. Zhenfang Chen, Jiayuan Mao, Jiajun Wu, Kwan-Yee K. Wong, Joshua B. Tenenbaum, and Chuang Gan. ICLR 2021.
【精度超越Transformer，MIT、港大提出基于物理模型的Neuro-Symbolic视觉推理框架】本文来自微信公众号 “机器之心”（ID：almosthuman2014）， 作者：Synced， 36氪经授权发布 。

• 得物垂涎“范冰冰”，小红书失去“杨超越”
• 下一任「影史第一」拿什么超越《长津湖》？
• 骁龙8 Gen 1跑分曝光，超越天玑9000
• 海马云战略副总裁许琳：元宇宙所要求的高精度，沉浸感和超低延时需要全新的云计算基础设施| 2021WISE元宇宙·机器人峰会
• 杨超越投资成立文化传媒公司，注册资本900万
• 36氪首发 | 「方位角」完成近亿元天使轮融资，打造室内外全域高精度定位导航授时系统
• 凶猛！搭载最新AMD显卡的黑苹果来了：跑分超越老款iMac
• 平台人脸识别未成年人？何延哲：考虑降低精度仅识别年龄特征
• 无上超越魂 无上超越
• 乔丹的经典语录有哪些呢 为什么乔丹被认为无法超越