Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning
This work addresses the problem of evaluating efficient fine-tuning methods for LLMs, showing that reported improvements may be incremental and dependent on hyperparameter tuning.
The paper systematically re-evaluated LoRA variants for LLM fine-tuning through extensive hyperparameter searches, finding that once learning rates are properly tuned, all methods achieve similar peak performance within 1-2%, suggesting vanilla LoRA remains competitive.
Low-Rank Adaptation (LoRA) is the prevailing approach for efficient large language model (LLM) fine-tuning. Building on this paradigm, recent studies have proposed alternative initialization strategies and architectural modifications, reporting substantial improvements over vanilla LoRA. However, these gains are often demonstrated under fixed or narrowly tuned hyperparameter settings, despite the known sensitivity of neural networks to training configurations. In this work, we systematically re-evaluate four representative LoRA variants alongside vanilla LoRA through extensive hyperparameter searches. Across mathematical and code generation tasks on diverse model scales, we find that different LoRA methods favor distinct learning rate ranges. Crucially, once learning rates are properly tuned, all methods achieve similar peak performance (within 1-2%), with only subtle rank-dependent behaviors. These results suggest that vanilla LoRA remains a competitive baseline and that improvements reported under single training configuration may not reflect consistent methodological advantages. Finally, a second-order analysis attributes the differing optimal learning rate ranges to variations in the largest Hessian eigenvalue, aligning with classical learning theories.