---
license: apache-2.0
---
# 🧠 Smart MCQ Solver

A Deep Learning based Multiple Choice Question Solver built using **RoBERTa-Base** and Hugging Face Transformers.

---

## Project Overview

This project predicts the correct answer among five options for a multiple-choice question.

The model is fine-tuned using the Hugging Face `AutoModelForMultipleChoice` architecture.

---

## Features

- Fine-tuned RoBERTa-base
- Multiple Choice Question Answering
- Gradio Web Interface
- Hugging Face Transformers
- PyTorch Implementation

---

## Model

RoBERTa-base

Task:

Multiple Choice Question Answering

---

## Performance

| Metric | Score |
|---------|-------|
| MAP@3 | **0.9893** |
| F1 Score | **0.99** |

Cross Validation

| Fold | MAP@3 |
|------|--------|
|1|0.9871|
|2|0.9854|
|3|0.9963|
|4|0.9933|
|5|0.9846|

Mean MAP@3

0.9893

---

## Installation

```bash
git clone https://github.com/YOUR_USERNAME/smart-mcq-solver.git

cd smart-mcq-solver

pip install -r requirements.txt
```

---

## Run

```bash
python app.py
```

---

## Screenshots

### Home Page

![Loading](image.png)

### Prediction

![Loading](image-1.png)

---

## Tech Stack

- Python
- PyTorch
- Transformers
- Hugging Face
- Gradio

---

## Author

Rohit Kumar

IIT Madras BS Degree