Atomic Thinking of LLMs: Decoupling and Exploring Mathematical Reasoning Abilities

Philip S Yu (UIC) · Yinghui Li (Tsinghua University, Tsinghua University) · Meishan Zhang (Harbin Institute of Technology (Shenzhen), China) · Xiaoyu Tan (Tencent Youtu Lab) · Yangning Li (Tsinghua University) · Jiayi Kuang (SUN YAT-SEN UNIVERSITY) · Haojing Huang (Tsinghua University, Tsinghua University) · Xinnian Liang (ByteDance Inc.) · Zhikun Xu (Arizona State University) · Chao Qu (Ant Financial Services Group) · Ying Shen (SUN YAT-SEN UNIVERSITY, Tsinghua University)
atomic capabilitiesatomic thinkingcognitive groundingconceptual understandingcounterexample-driven reasoningevaluation datasetsfield-specific abilitiesformal math languagelogical abilitiesmathematical intelligencemathematical reasoningmodel cognitionmulti-step reasoningtraining datasetstransferable strategies

Large Language Models (LLMs) have demonstrated outstanding performance in mathematical reasoning capabilities. However, we argue that current large-scale reasoning models primarily rely on scaling up training datasets with diverse mathematical problems and long thinking chains, which raises questions about whether LLMs genuinely acquire mathematical concepts and reasoning principles or merely remember the training data. In contrast, humans tend to break down complex problems into multiple fundamental atomic capabilities. Inspired by this, we propose a new paradigm for evaluating mathematical atomic capabilities. Our work categorizes atomic abilities into two dimensions: (1) field-specific abilities across four major mathematical fields, algebra, geometry, analysis, and topology, and (2) logical abilities at different levels, including conceptual understanding, forward multi-step reasoning with formal math language, and counterexample-driven backward reasoning. We propose corresponding training and evaluation datasets for each atomic capability unit, and conduct extensive experiments about how different atomic capabilities influence others, to explore the strategies to elicit the required specific atomic capability. Evaluation and experimental results on advanced models show many interesting discoveries and inspirations about the different performances of models on various atomic capabilities and the interactions between atomic capabilities. Our findings highlight the importance of decoupling mathematical intelligence into atomic components, providing new insights into model cognition and guiding the development of training strategies toward a more efficient, transferable, and cognitively grounded paradigm of "atomic thinking".