Darren LI

无模型方法专注于直接从与环境的交互中找出value function。此类中的算法严重依赖reward 来学习值函数。因此,具有学习reward的功能很重要。 此外,它们通常更易于实现和调整超参数。

从单个图像或语言描述的任意位置开始,合成一系列不同的3D场景

Instead of being a passive user of pre-built applications, I can start assembling my own personalized way of using my computer.

Language is a richer form of supervision than classical closed-set labels.

Only 17.9% of DAO treasury funds came from centralized services, while the remaining 82.1% originated at decentralized services. This suggests that most DAO contributors also engage with DeFi platforms and likely self-host their cryptocurrency.

The modern browser extension API has done a good job balancing extensibility with security, and yet we’re still grappling with the consequences of browser extensions invading people’s privacy. Giving extensions more power would raise the stakes further. Still, we shouldn’t give up in the name of security—we should fight for extensibility as a value... See more

We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms.