List your company

Arithmetic Reasoning on ASDiv (5-fold cross val)

73.9AccuracyPaLM 540B (CoT 8-shot)
38.269647.519856.7766.0202Dec 16, 2022

Evaluation results

MethodLinks
PaLM 540B (CoT 8-shot)Model=PaLM 540B, Finetuning Strategy=None (Zero-shot/Few-shot), Prompting Strategy=8-shot Chain-of-thought2022.1273.972.6
T5 XXL (CoT Finetuned)Model=T5 XXL, Finetuning Strategy=CoT knowledge distillation, Prompting Strategy=Chain-of-thought2022.1242.1260.73
T5 XXL (Baseline)Model=T5 XXL, Finetuning Strategy=Finetuned on original target labels, Prompting Strategy=Direct2022.1239.64—