Hadas Orgad
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Is Sycophancy a Form of In-Context Updating?
According to the abstract, Shapira, Cheng and Orgad argue that factual sycophancy in language models is better seen as a failure of in-context updating than as a bid for user approval, and offer behavioral and mechanistic evidence.