Optimizing LoRA target module selection for efficient fine tuning

Ablation study clarifies trade-offs between accuracy and efficiency when using low-rank adaptation (LoRA) to fine-tune AI models.