CUAD (NeurIPS 2021)

A repository of legal NLP research papers.

[ACL 2025] Codebase for CoCoLex

Correct way to measure BPE token cost for non-English text + synonym substitution to reduce LLM API costs. Worked example: Portuguese legal documents.

Reproducible Legal NER pipeline using RoBERTa, LegalBERT, linguistic features, and back-translation. Implements data preparation, spaCy training configs, and evaluation on the InLegalNER corpus with 14 fine-grained entity types.

A blind benchmark for legal citation verification — 4-label classification over IL + federal primary law

MSc group project evaluating instruction-tuned LLMs for legal clause classification using prompt-based inference and performance analysis.

Compressing LegalBERT via knowledge distillation for legal-domain NER (InLegalNER)

A diachronic analysis of legal metalanguage used by the U.S. Supreme Court from 1986 to 2019.

Legal Question Answering System using Knowledge Graphs and Neural Retrieval (UET Thesis)

Use LLMs to identify premises and claims within the ECHR dataset.

🎯 AI-Powered Multi-Document Legal Research Assistant | Advanced RAG System for Legal Document Analysis with Conflict Detection & Citation Management

SummarizeLaw uses InLegalBERT to generate concise summaries of Indian court judgments and evaluates output quality against reference summaries.

AI-powered system for predicting appeal case outcomes using LegalBERT and generating legal briefs with GPT-4o-mini

Article reproducibility LexIris-pt and LexBert-pt: Specialized Sentence Embeddings for Legal Similarity in Brazilian Portuguese.

(ACL 2026 Main) Mitigating Legal Hallucinations via Symbolic Constraints and Analogical Precedents

Sistema híbrido de NLP para recuperación de información y question answering sobre normativa SENASA (TF-IDF, embeddings, RAG).

Legal Argument Mining on Italian tax-court decisions (AMELIA dataset) TF-IDF+LR baseline vs BERTino fine-tuning. Macro-F1: 0.922. MD2SL Text Mining & NLP exam project, UniFi.

Multi-label legal text classifier fine-tuned on ECtHR cases using DistilBERT.

Benchmarking QLoRA, DoRA and IA3 vs RAG on Llama-3.2-3B for legal contract risk analysis — 4-layer evaluation framework with statistical significance testing