As the title suggests, I am seeking a model, preferably around 7 billion parameters, to generate more human-like text, specifically at the level of a student. I am working on a project where I aim to produce flawless generations that challenge discriminator models in distinguishing between human-written and generated texts. While prompt engineering aids to a certain extent, high-quality models tend to lean towards more formal and grammatically correct writing. Do you have any personal recommendations or preferences for a model of this nature?