Feature-wise Linear Modulation (FiLM)

First seen somewhere, but more recently in OpenVLA-OFT.

FiLM learns to adaptively influence the output of a neural network by applying anAffine Transformation, or FiLM, to the network’s intermediate features, based on some input.

What is conditioning?

Conditioning, in a general sense, involves adjusting a model’s behavior based on additional information. In probabilistic terms, it’s akin to computing the probability of an event given some known context.

2000+ citations

Links:

How does FILM conditioning work?