Modelpedia
Work in progress
About
Findings
Models
Concepts
Methods
Datasets
Sources
Contribute
Light
Dark
Instruction Induction
Report a problem with this record
anchor
Findings
IC-991
LLaMA-2, Falcon-7B, and GPT-3.5-Turbo exhibit large performance spread (up to 76 accuracy points) across semantically equivalent prompt formats, and model comparison rankings are frequently reversed by format choice
[eval]