Rephrasing training data with an LLM that has read the MeZO paper gives small and inconsistent accuracy gains for zeroth-order LLM fine-tuning, with no error bars reported.
Boolq: Exploring the surprising difficulty of natural yes/no questions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
dataset 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
dataset 1polarities
use dataset 1representative citing papers
citing papers explorer
-
OAT-Rephrase: Optimization-Aware Training Data Rephrasing for Zeroth-Order LLM Fine-Tuning
Rephrasing training data with an LLM that has read the MeZO paper gives small and inconsistent accuracy gains for zeroth-order LLM fine-tuning, with no error bars reported.