MolmoAct 2
Allen Institute for AIwe normalize using dataset quantiles and discretize into 256 uniform-width bins between the first and ninety-ninth percentiles
implementation detail · filed under data curation
Action dimensions are normalized using dataset quantiles, with the first and 99th percentiles mapped to a specified range; one cited implementation then discretizes into 256 uniform-width bins.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See tokenization for the methods filed beside it.
Also called quantile normalization.
2 quotes from 2 sources, one card per model, each opening with its strongest.
we normalize using dataset quantiles and discretize into 256 uniform-width bins between the first and ninety-ninth percentiles
We first normalize the input actions, such that the 1st and 99th quantile of values in the training dataset for each action dimension maps to the range