Best model from a 3-model comparison for the DLGenAI Smart MCQ Solver project (T22026). Predicts the correct answer out of 5 options (A-E) for a multiple choice question.

Validation MAP@3: 1.0 Validation Accuracy: 1.0

See all_model_results.csv in this repo for full comparison across all 3 models tested (scratch TF-IDF+LogReg, pretrained e5-small-v2+MLP, and fine-tuned DeBERTa-v3-small).