Cohere's Kris Cao argues the classic imitation learning algorithm DAgger has been rebranded as "on-policy distillation

— Both concepts rely on iteratively aggregating expert feedback data.

Original post

OPKC#991Kris Cao|@KROSCOO

When did DAgger get renamed on-policy distillation?

4:57 AM · May 28, 2026 View on X

REPLYLB#567Leo Boytsov|@SRCHVRS

@kroscoo LOL, I didn't make the connection, but this seems to be a correct one. 😅

KCKris Cao@kroscoo

When did DAgger get renamed on-policy distillation?

11:57 AM · May 28, 2026 · 1.7K Views

11:43 PM · May 28, 2026 · 531 Views