Teaching small models to review contracts
We investigated whether a small language model can match state-of-the-art contract review performance. We built a training pipeline, tested distillation, and implemented a hand-off system to reach the target accuracy of the best published models.
Read