Audio codec models as a recipe for action tokenizers
A post calls audio codec models “good action tokenizer recipes,” with its author planning to present the work at CoRL in November 2026.
TLDR
Audio codec models are surprisingly good recipes for action tokenizers, according to a September 26 post. The author says they will present the work at CoRL in November 2026.
Combined views
6.1K
2 Sources, first seen 7h ago
49 likes