krishy-d 's Collections

Prosify: Correcting Formatting Bias in LLMs

An end-to-end research project on correcting the systematic over-formatting habit of RLHF-trained language models.