Training language models to predict number ranges, not single points

Original: Distribution-Aware Reward: Reinforcement Learning over Predictive Distributions for LLM Regression

Source: arxiv.org ↗