ELECTRA

E1339923 UNEXPLORED

ELECTRA is a pre-training method for transformer-based language models that trains discriminators to distinguish real input tokens from replaced ones, offering greater sample efficiency than traditional masked language modeling.

All labels observed (1)

Label Occurrences
ELECTRA canonical 2

How this entity was disambiguated

Referenced by (2)

Full triples — surface form annotated when it differs from this entity's canonical label.

BERT influenced ELECTRA