MIT Uses LLMs to Teach Robots to Understand Vague Commands

MIT_CSAIL · x · 2026-07-04

MIT introduced "Masked IRL," leveraging LLMs to help robots perform tasks like household chores despite ambiguous instructions by focusing on crucial details. One model clarifies the user's vague prompts, while the other filters out irrelevant information.

Original post →

More from Research

Research channel →