Clark and colleagues: ELECTRA replaced-token detection

What changed

Proposed replaced-token detection as a sample-efficient discriminative pre-training objective for language representations.

Scientific scope

Proposed replaced-token detection as a sample-efficient discriminative pre-training objective for language representations. Scope limitation: ELECTRA changes the pre-training task; it is not the original BERT architecture or a general proof of compute superiority.

Evidence

The retained primary-source abstract, bibliographic record and stated review scope support this provisional assessment. No independent experimental replication by WRS is claimed. ELECTRA changes the pre-training task; it is not the original BERT architecture or a general proof of compute superiority.

Contributors

• The named authors listed in the linked research record · SCIENCE: Complete source byline retained in the research record; human research direction, method development and evaluation. Individual labor shares are not measured.

• Clark and colleagues: ELECTRA replaced-token detection — original paper
• Clark and colleagues: ELECTRA replaced-token detection — official company source

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction