We find that GPT-NeoX-20B is a particularly powerful few-shot reasoner and | |
gains far more in performance when evaluated five-shot than similarly sized GPT-3 and FairSeq models. |
We find that GPT-NeoX-20B is a particularly powerful few-shot reasoner and | |
gains far more in performance when evaluated five-shot than similarly sized GPT-3 and FairSeq models. |