Open Models Almanac
Methodsmodel architecturemultimodal architecture

single source · model architecture

Masking action tokens from expert conditioning

Masks the target action-token span so the action expert cannot use it as VLM conditioning.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See multimodal architecture for the methods filed beside it.

Filed under model architecture › multimodal architecture