Connect
Ongoing research on compressing text embeddings without labeled data, using a novel two-stage distillation approach with generative adversarial adapters.
Learning to mimic famous musicians with generative language models.