Fine-tuning a likelihood-based generative model by contrasting to training data

A method, comprising:

• at a device:
• processing a plurality of ground truth data samples and a plurality of reference data samples generated by a pretrained reference likelihood-based generative model, by a generative adversarial network (GAN) discriminator model parameterized using a target likelihood-based generative model, to generate classifications for the plurality of reference data samples and the plurality of ground truth data samples;
• and
• fine-tuning the target likelihood-based generative model based on a loss computed from the classifications.

A system, comprising:

• a non-transitory memory comprising instructions;
• and
• one or more processors in communication with the non-transitory memory, wherein the one or more processors execute the instructions to:
• process a plurality of ground truth data samples and a plurality of reference data samples generated by a pretrained reference likelihood-based generative model, by a generative adversarial network (GAN) discriminator model parameterized using a target likelihood-based generative model, to generate classifications for the plurality of reference data samples and the plurality of ground truth data samples;
• and
• fine-tune the target likelihood-based generative model based on a loss computed from the classifications.