Structural Priors and Modular Adapters in the Composable Fine-Tuning Algorithm of Large-Scale Models

Yuxiao Wang, Di Wu, Feng Liu +2

This paper proposes a composable fine-tuning method that integrates graph structural priors with modular adapters to address the high computational cost and structural instability…