LIMITS OF PRE-TRAINED LANGUAGE MODELS

Summary from the TLDR AI newsletter

A study exploring whether instruction-tuned models possess fundamentally different capabilities from base models that are prompted using in-context examples.