Open access
Aug 2026
GME-Init: Gamma-Moment Equalization for LoRA Initialization in Parameter-Efficient Fine-Tuning
Gamma-Moment Equalization Initialization (GME-Init) is proposed, a data-aware asymmetric LoRA initialization method based on output-moment calibration that consistently improves standard LoRA and selected LoRA-style methods across the evaluated text understanding and multimodal tasks.
Yuhui Lin, Chao-Peng Li, Zhi-Wei Shen et al.
· Entropy · 0 citations