Optimization Analysis for LoRA-Soups Weighting
- A computational intelligence study comparing optimization strategies for merged LoRA weighting.
LoRAOptimizationMergeBLEUROUGE
- Status
- Research
- Date
- Dec 2025
- Role
- Optimization experimenter
About
This final project compares Grey Wolf Optimization, CMA-ES, and Genetic Algorithm strategies for finding effective LoRA-Soups weights in large language model adaptation. It evaluates outputs with BLEU and ROUGE across bilingual and instruction datasets.
Why it matters
Combines adapter research with classical optimization search.