Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Light
Dark
The False Promise of Imitating Proprietary Language Models
2024-01-16
· ICLR 2024 spotlight ·
anchor
Findings
IC-899
GPT-4, when used as a blind pairwise evaluator, exhibits the same style-over-factuality preference as human crowdworkers