DistilBERT MRPC Paraphrase Classifier
Fine-tuned distilbert-base-uncased model for paraphrase identification using the GLUE MRPC dataset.
Model Description
This model predicts whether two sentences have the same meaning.
Labels
0 → NOT_PARAPHRASE1 → PARAPHRASE
Dataset
- Dataset: GLUE MRPC
- Task: Sentence pair classification
Training Details
- Base model: distilbert-base-uncased
- Epochs: 1
- Batch size: 8
- Maximum sequence length: 128
Evaluation
Accuracy: 85.05%
Usage
Use the model with Hugging Face Transformers:
from transformers import pipeline
classifier = pipeline(
"text-classification",
model="shadowdemonosprey/mrpc-distilbert"
)
result = classifier(
{
"text": "The company released a new phone.",
"text_pair": "A new phone was released by the company."
}
)
print(result)
Example Output
PARAPHRASE
Confidence: 0.9643
Limitations
This model was trained on the MRPC dataset and may not generalize perfectly to all sentence similarity tasks.
