Open Models Almanac
Methodspost-trainingsupervised fine-tuning

general family · filed under post-training

Imitation learning

Imitation learning trains a policy once, offline, rather than continually updating it via reinforcement learning.1 It learns high-performing policies from a fixed dataset of expert task demonstrations, without collecting new online interaction data.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. unlike other works that collect online interaction data to continually update specialist policies via reinforcement learning [19], here we focus on the simpler imitation learning paradigm — Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success
  2. we develop high-performing policies just by training once offline on a fixed dataset of expert task demonstrations — Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success

Also called Imitation learning for VLA training, imitation learning paradigm.

sources
2
models
2
labs adopt it
2
strongest
used

Documented in

Hugging Face
OpenVLA project (Stanford et al.)

Evidence

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

How sources treat it
used 2
↑