GradOT selects compression by minimizing a Gradient-preserving Compression Score derived from a Taylor approximation of the offsite-tuning objective, achieving competitive plug-in performance and larger emulator-to-plug-in gaps.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
GradOT: Training-free Gradient-preserving Offsite-tuning for Large Language Models
GradOT selects compression by minimizing a Gradient-preserving Compression Score derived from a Taylor approximation of the offsite-tuning objective, achieving competitive plug-in performance and larger emulator-to-plug-in gaps.