Rishanth Rajendhran

POLARIS trains small models for better long-form story writing

Researchers have developed POLARIS, a new training method designed to improve the long-form creative writing capabilities of smaller open-weight language models. This method utilizes a frontier LLM as a judge with a str…