1 paper · 1 filter
Juncheng Xie, Hung-yi Lee
Controlling the length of text produced by large language models (LLMs) remains challenging: models frequently overshoot or undershoot explicit length instructions because they can…