23f2001332/my-mcq-model

Smart MCQ Solver - Model 1 (SimpleMCQModel)

An end-to-end custom PyTorch deep learning architecture designed for ranking multiple-choice options (MCQs).

Model Architecture

• Type: Embedding Average-Pooling Multi-Layer Perceptron (MLP)
• Embedding Dim: 128
• Hidden Dim: 64
• Sequence Length: 128
• Vocabulary Size: 3084
• Classes: 5 options (A, B, C, D, E)

Performance Metrics

Metric	Score
Validation Accuracy	91.50%
MAP@3 Score	0.9531
Training Loss	0.3854
Epochs	5
Random Seed	42

Reproduction & Usage

from model1_hf.inference import MCQInferencePipeline

pipeline = MCQInferencePipeline.load_from_dir("./model1_hf")
result = pipeline.predict(
    prompt="What is the capital of France?",
    options=["London", "Paris", "Berlin", "Madrid", "Rome"]
)

print("Top Prediction:", result["top1_label"], "-", result["top1_option_text"])
print("Top 3 Choices:", result["top3_str"])
print("Confidence Scores:", result["confidence_scores"])