COSEE trains each early-exit classifier to emphasize samples that would exit at it under simulated thresholds, improving GLUE accuracy at matched speed-up over prior methods.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
COSEE: Consistency-Oriented Signal-Based Early Exiting via Calibrated Sample Weighting Mechanism
COSEE trains each early-exit classifier to emphasize samples that would exit at it under simulated thresholds, improving GLUE accuracy at matched speed-up over prior methods.