consistent. accurate.
Unlike generic AI platforms that rely heavily on unsupervised learning, our approach leverages:
.svg){kind=icon}
Expert-labelled data
Every training example verified by legal professionals
Contextually aware annotation
Nuanced legal implications beyond simple text
Strategic sampling methods
Balanced representation across contract types and legal domains
Lexible® isn't just another AI model with “legal” features, it's purpose-built for contract review from the ground up:
Specialized Legal Training
Trained on millions of meticulously hand-labelled concepts
Continuous Improvement
Rigorous testing 3x weekly against 750,000 verified data points
Multi-Layered Architecture
Combines general-purpose LLMs with our legal-specific model for comprehensive understanding
Expert Validation
Every training triplet verified by our in-house legal team
Accuracy (F1)
Accuracy (or the F1 score) is a measure combining recall and precision. Trade-offs exist. F1, therefore, measures how effectively our models make that judgment.
F1 scoring penalises extreme negative values of either component. Thus, if either component fails, the score falls to zero.
Precision
The precision metric provides the proportion of true positives to the amount of total positives predicted. It answers the question:
“Out of all the positive predictions we made, how many were true?”
Recall
Recall focuses on how well the model finds positives. It is also called the “true positive rate”. It answers the question:
“Out of all the data points that should be predicted as true, how many did we correctly predict as true?”
