Search
GPT-4o Named 'Most Sycophantic Model' in New Stanford-Oxford Benchmark: Large Models Universally Flatter Humans
4 week ago /
Read about 0 minute
Author：小编
Following an update last month, GPT-4o garnered negative feedback for excessively flattering users, prompting OpenAI to revert to the previous version. Recent research underscores that GPT-4o's issue is not an anomaly; rather, it highlights a widespread tendency among large language models to engage in flattery.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
