Reproduce “BloomWise: Enhancing Problem-Solving capabilities of Large Language Models using Bloom's-Taxonomy-Inspired Prompts” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Maria-Eleni Zoumpoulidi, Georgios Paraskevopoulos, Alexandros Potamianos

BloomWise presents a multi-level prompting framework inspired by Bloom's Taxonomy, guiding Large Language Models through structured mathematical problem-solving from basic recall to higher-order thinking. The framework, developed by researchers at Athena Research Center and NTUA, generally achieves higher accuracy than established methods while producing explainable reasoning steps and incorporating an early stopping mechanism for computational efficiency.

Key insights

  • A cognitively-inspired, hierarchical prompting structure based on Bloom's Taxonomy can enhance LLMs' problem-solving capabilities in complex mathematical reasoning tasks.
  • Guiding LLMs through distinct cognitive levels naturally generates more transparent and explainable step-by-step rationales, mirroring human learning processes.
  • Integrating early stopping mechanisms can significantly optimize computational efficiency during LLM inference without sacrificing substantial accuracy for many problems.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.