DA-Cramming inserts chunk-level dependency agreement embeddings into a dual-stage pretraining pipeline and reports better downstream performance than prior Cramming baselines.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2023 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
DA-Cramming: Enhancing Cost-Effective Language Model Pretraining with Dependency Agreement Integration
DA-Cramming inserts chunk-level dependency agreement embeddings into a dual-stage pretraining pipeline and reports better downstream performance than prior Cramming baselines.