Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

human preference alignment

Human preference alignment is the process of training or adjusting an AI model so its responses better reflect what people prefer, often using human judgments or comparisons of possible responses as feedback. In language models, this feedback can guide fine-tuning so the model is more likely to produce preferred outputs.

1 item

ChiMed-GPT: A Chinese Medical Large Language Model with Full Training Regime and Better Alignment to Human Preferences

ChiMed-GPT: A Chinese Medical Large Language Model with Full Training Regime and Better Alignment to Human Preferences

Yuanhe Tian, Ruyi Gan, Yan Song, Jiaxing Zhang, Yongdong Zhang

Why you should read this

Develops ChiMed-GPT, an open-source Chinese medical large language model trained through continual pre-training, supervised fine-tuning, and rejection-sampling-based reinforcement learning from human feedback to generate accurate, safe clinical responses with an extended context window.

Added

2026-10-02