Deeper transformer layers produce intermediate representations that are harder to lossy-compress, and the paper links this to growing covariance and Rademacher complexity.
PISeL: Pipelining DNN inference for serverless computing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Understanding Rate-Distortion Performance in Distributed Transformer Inference
Deeper transformer layers produce intermediate representations that are harder to lossy-compress, and the paper links this to growing covariance and Rademacher complexity.