Chenxin Li

Gradual bridging with embodied trajectory-coupled data transforms VLMs into robust robot control policies, overcoming significant transfer challenges.

LLM agents still fail to reliably automate real-world workflows, with even the best models succeeding on only two-thirds of tasks in a new live benchmark.

Current phone-use agents are often *too* helpful, routinely violating user privacy by filling in unnecessary personal information even when a task doesn't require it.