C-VARC: A Large-Scale Chinese Value Rule Corpus for Value Alignment of Large Language Models

Name: C-VARC: A Large-Scale Chinese Value Rule Corpus for Value Alignment of Large Language Models
Uploaded: 2026-10-17T07:38:32.761276+00:00
Duration: 442 s
Description: The research created C-VARC, a large corpus of over 250,000 human-annotated Chinese value rules based on national and societal dimensions. This resource is used to align Large Language Models with Chinese cultural values. C-VARC shows that rule-driven guidance produces better semantic diversity and higher alignment rates than using abstract values alone.

The research created C-VARC, a large corpus of over 250,000 human-annotated Chinese value rules based on national and societal dimensions. This resource is used to align Large Language Models with Chinese cultural values. C-VARC shows that rule-driven guidance produces better semantic diversity and higher alignment rates than using abstract values alone.

Open the full episode