Closed gabrer closed 2 years ago
Yes. The DMBERT model here is the vanilla baseline without adversarial training (corresponding to DMBERT
notation in the original paper). The DMBERT with adversarial training (corresponding to DMBERT+ADV
or DMBERT+Boot
) is not included in MAVEN experiments since here are in the fully supervised setting rather than the weakly supervised setting.
Just a quick question, as I cannot find the adversarial training in the DMBERT implementation.
It seems that it's just doing a sentence encoding and classification, without any generator/discriminator mechanism.
Is this implementation different from the one described in the original paper?
Thanks!