notesum.ai

Published at November 4

Can Large Language Models generalize analogy solving like people can?

cs.AI

cs.CL

cs.HC

Released Date: November 4, 2024

Authors: Claire E. Stevenson¹, Alexandra Pafford¹, Han L. J. van der Maas¹, Melanie Mitchell²

Aff.: ¹University of Amsterdam, the Netherlands; ²Sante Fe Institute, USA

Arxiv: http://arxiv.org/abs/2411.02348v1

Refer to caption

Participant Group	n	Latin		Greek		Symbol
		Mean	SD	Mean	SD	Mean	SD
Adults	62	0.88	0.16	0.91	0.13	0.89	0.23
Children	41	0.62	0.22	0.66	0.23	0.67	0.30
Claude-3.5	54	0.68	0.18	0.62	0.21	0.46	0.24
Gemma-2 27B	54	0.60	0.24	0.39	0.20	0.14	0.15
GPT-4o	54	0.85	0.18	0.63	0.21	0.48	0.18
Llama-3.1 405B	54	0.79	0.16	0.74	0.19	0.27	0.20