deathsamsul/finetune-vs-rag-benchmark

A reproducible benchmark comparing LLM fine-tuning (LoRA/QLoRA), RAG, and Fine-tuning + RAG across answer quality, faithfulness, latency, robustness, and failure modes.