Open Models Almanac
Methodspost-trainingsupervised fine-tuning

specific method · filed under post-training

Embodiment-specific fine-tuning

Embodiment-specific fine-tuning continues training a post-trained VLA checkpoint, keeping the same architecture, to adapt it to a specific robot embodiment.1 Unlike post-training, each fine-tuning run trains only on robot data, dropping the multimodal VLM mixture entirely.2

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. Embodiment-specific fine-tuning starts from the post-trained MolmoAct2-Post checkpoint and keeps the same VLM–action-expert architecture — MolmoAct2 Action Reasoning Models for Real-World Deployment
  2. each run is robot-only: we do not include the multimodal VLM mixture or a separate VLM dataloader — MolmoAct2 Action Reasoning Models for Real-World Deployment

Also called Embodiment-specific deployment fine-tuning, Embodiment-specific downstream adaptation, embodiment-specific adaptation.

sources
2
models
2
labs adopt it
2
strongest
core

Documented in

Allen Institute for AI
X-VLA authors

Evidence

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

How sources treat it
used 2core 1

X-VLA

X-VLA authorsadopted1 source

Currently, deployment still relies on embodiment-specific adaptation, typically involving the collection of a small number of demonstrations for post-training.

usedpost trainingin X-VLA
1 more from this model

typically involving the collection of a small number of demonstrations for post-training.

usedpost trainingin X-VLA
↑