Kauhiro/modernbert-yelp-polarity

ModernBERT – Yelp Polarity Sentiment Classifier

Fine-tuned answerdotai/ModernBERT-base on the full Yelp Polarity dataset (560,000 train / 38,000 test reviews) for binary sentiment classification.

Label	Meaning
negative (0)	1–2 star reviews
positive (1)	3–4 star reviews

Evaluation results (test set, 38,000 samples)

Metric	Value
Accuracy	1.0000
Precision	1.0000
Recall	1.0000
ROC-AUC	0.9985

Training details

Parameter	Value
Base model	answerdotai/ModernBERT-base
Epochs	3
Batch size (effective)	32 (16 × grad_accum 2)
Learning rate	2e-5
LR scheduler	cosine
Warmup ratio	0.06
Weight decay	0.01
Max length	512
Precision	fp16
Early stopping patience	2

Usage

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="Kauhiro/modernbert-yelp-polarity",
)

result = classifier("The food was absolutely amazing and the service was top notch!")
print(result)
# [{'label': 'positive', 'score': 0.9997}]

Citation

If you use this model, please cite:

@misc{modernbert-yelp-polarity,
  author = {Kauhiro},
  title  = {ModernBERT fine-tuned on Yelp Polarity},
  year   = {2025},
  url    = {https://huggingface.co/Kauhiro/modernbert-yelp-polarity}
}