lexglue-ledgar-distilbert

🧠 LexGLUE–LEDGAR DistilBERT Legal Clause Classifier

Author: Salman Abbasi
Affiliation: Beijing Institute of Technology, China
Research Area: NLP × Agentic AI × Prompt Engineering
Date: October 2025

---

📝 Model Overview

This model is a DistilBERT-based legal clause classifier fine-tuned on the LexGLUE–LEDGAR dataset.
It predicts the type of contract clause (e.g., Confidentiality, Termination, Liability, Payment, etc.) given a legal paragraph or clause.

• Base Model: distilbert-base-uncased
• Dataset: coastalcph/lex_glue (ledgar configuration)
• Task Type: Multi-class text classification
• Language: English
• Domain: Legal / Contractual Texts

This model achieves strong performance despite its small size, making it efficient for deployment in legal NLP systems and agentic AI applications.

---

🚀 Model Performance

---

📚 Dataset — LexGLUE LEDGAR

The LEDGAR dataset contains contract clauses annotated with one of 100+ provision types.
It is part of the LexGLUE benchmark, which evaluates legal text understanding models.

Each sample includes:

{
  "text": "This clause sets out the obligations of each party regarding confidentiality.",
  "label": "Confidentiality"
}