#readability

The Decoder4d ago

NYU and UT Austin study finds six patterns that make AI text hard to read

Across 14 models, post-training and verifiable-reward training pushed text further from human writing than base models.