Dingxian Wang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Alignment by Confident Debate: Preference Learning from Multi-Agent Debate
A preprint introduces Alignment by Confident Debate (AbCD), which turns multi-agent debate among language models into preference data for post-training without gold answer labels. According to the abstract, it uses verifiable tasks such as math and code as a test bed.