3 papers
cs.LG2025
DE-BERT: Distance-Enhanced Early Exiting for BERT based on Prototypical Networks
Jianing He, Qi Zhang, Weiping Ding +4
Early exiting has demonstrated its effectiveness in accelerating the inference of pre-trained language models like BERT by dynamically adjusting the number of layers executed. Howe…
cs.LG2025
Improving Prediction Certainty Estimation for Reliable Early Exiting via Null Space Projection
Jianing He, Qi Zhang, Duoqian Miao +4
Early exiting has demonstrated great potential in accelerating the inference of pre-trained language models (PLMs) by enabling easy samples to exit at shallow layers, eliminating t…
cs.LG2024
COSEE: Consistency-Oriented Signal-Based Early Exiting via Calibrated Sample Weighting Mechanism
Jianing He, Qi Zhang, Hongyun Zhang +3
Early exiting is an effective paradigm for improving the inference efficiency of pre-trained language models (PLMs) by dynamically adjusting the number of executed layers for each…