Thinking in Many Modes: How Composite Reasoning Elevates Large Language Model Performance with Limited Data

Zishan Ahmad, Saisubramaniam Gopalakrishnan

Large Language Models (LLMs), despite their remarkable capabilities, rely on singular, pre-dominant reasoning paradigms, hindering their performance on intricate problems that dema…