MolmoAct 2
Allen Institute for AInew depth tokens are computed only for novel visual signals
specific method · filed under inference & serving
The model adjusts reasoning depth or decides when more reasoning is useful; one report specifies computing depth tokens only for novel visual signals.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See reasoning control for the methods filed beside it.
Also called adaptive depth reasoning.
2 quotes from 2 sources, one card per model, each opening with its strongest.
new depth tokens are computed only for novel visual signals