Comparing the Performance of LLMs: A Deep Dive into Roberta, Llama 2, and Mistral for Disaster Tweets Analysis with Lora

A comparative analysis of the performance of RoBERTa, LLaMA 2, and Mistral models on disaster tweet classification tasks was conducted, utilizing LoRA (Low-Rank Adaptation) for fine-tuning. The study evaluated model accuracy, with Mistral achieving the highest F1 score of 0.87, followed by LLaMA 2 at 0.83, and RoBERTa at 0.80. This analysis provides insights into the trade-offs between model architectures and fine-tuning techniques, critical for practitioners seeking to optimize LLMs for specific applications in disaster response scenarios.