notesum.ai

Published at November 15

Evaluating Creativity and Deception in Large Language Models: A Simulation Framework for Multi-Agent Balderdash

cs.MA

cs.AI

Released Date: November 15, 2024

Authors: Parsa Hejabi¹, Elnaz Rahmati¹, Alireza S. Ziabari¹, Preni Golazizian¹, Jesse Thomason¹, Morteza Dehghani¹

Aff.: ¹University of Southern California

Arxiv: http://arxiv.org/abs/2411.10422v1

Dataset	# Words	Avg. Frequency
All Balderdash	225	1.8e-8
Llama-Known	84	3.6e-8
Phi-Known	88	3.7e-8
Gemma-Known	35	5.7e-8
Mistral-Known	88	3.7e-8
GPT-Known	131	2.6e-8
Basic English	2865	6.3e-5