A Comprehensive Analysis of Evaluation Datasets

This application presents a detailed analysis and compilation of evaluation datasets for the "Framework for Complex-Reasoning Taxonomy." The primary objective is to systematically explore the "Evaluation Datasets" for each category by synthesizing empirical findings from cognitive science, education, and artificial intelligence.

A significant theme that emerges is the parallel and increasingly convergent evolution of datasets for human and artificial intelligence. Historically, assessment has followed two distinct tracks: one rooted in cognitive psychology (e.g., Wason Selection Task) and another driven by Large Language Models (LLMs) using large-scale datasets (e.g., GSM8K). This explorer illuminates the dynamic interplay between these two traditions.

Chart: High-level overview of the primary dataset categories discussed throughout this report.