/LLM-FineTuning-for-Sentiment-Classification

Fine-Tuning Microsoft Phi-2 for financial news sentiment analysis improved accuracy from 0.349 to 0.872 and reduced training loss by 38%. Dataset: 4846 entries with "Neutral," "Positive," and "Negative" labels.

Primary LanguageJupyter NotebookMIT LicenseMIT

Watchers