A Jev-style model fine-tuned on Qwen3.5 4B

A developer fine-tuned the Qwen3.5 4B model using LoRA, incorporating both public datasets and synthetic data to create a Jev-style model. This experimental model, while not yet matching Jev, shows significant improvement over the base model, with its performance on typed-decisions increasing from 0.596 to 0.709. The project involved a fun experiment to enhance the model's capabilities.