Open Models Almanac
Methodsdata curationtokenization

implementation detail · filed under data curation

Quantile normalization of robot actions

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.

sources
2
models
2
labs adopt it
2
strongest
core

Documented in

Allen Institute for AI
Physical Intelligence

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
used 1core 1
↑