Introducing CACTI — a framework for scalable multi-task multi-scene imitation learning.

Our agents show generalization to unseen distractors & are extremely sample efficient, capable of 10 unique real-world tasks with just 1k demonstrations. Paper ➡️ https://t.co/5zB4eMQvWV pic.twitter.com/nqSVcUP1yr

— AI at Meta (@AIatMeta) 28 février 2023

Introducing CACTI — a framework for scalable multi-task multi-scene imitation learning. Our agents show generalization to unseen distractors & are extremely sample efficient, capable of 10 unique real-world tasks with just 1k demonstrations. Paper https://
bit.ly/3SD5roy

→ View original post on X — @aiatmeta,