DeL-TaiseiOzaki/

JDERW

	JDERW: A Benchmark for Evaluating World Models in Large Language Models (LLMs)




	
	Overview

JDERW (Japanese Dataset for Evaluating Reasoning with World Models) is a benchmark dataset designed to assess the ability of Large Language Models (LLMs) to understand and reason about real-world phenomena and common sense. It includes 103 questions categorized into six reasoning types:

Causal Reasoning (e.g., Why does it snow in winter?) Temporal Reasoning (e.g., What happens when… See the full description on the dataset page: https://huggingface.co/datasets/DeL-TaiseiOzaki/JDERW.