Implicit Reasoning in Transformers is Reasoning through Shortcuts

Abstract: Test-time compute is emerging as a new paradigm for enhancing LLMs' complex multi-step reasoning capabilities, as demonstrated by the success of OpenAI's o1 and o3, as well as DeepSeek's R1. Compared to explicit reasoning in test-time compute, implicit reasoning is more inference-efficient, requiring fewer generated tokens. However, why does the advanced reasoning capability fail to emerge in the implicit reasoning style? In this work, we train GPT-2 from scratch on a curated multi-step mathematical reasoning dataset and conduct analytical experiments to investigate how LLMs perform implicit reasoning in multi-step tasks. Our findings reveal: 1) LLMs can perform step-by-step reasoning and achieve high accuracy in both in-domain and out-of-domain tests via implicit reasoning. However, this capability only emerges when trained on fixed-pattern data. 2) Conversely, implicit reasoning abilities emerging from training on unfixed-pattern data tend to overfit a specific pattern and fail to generalize further. Notably, this limitation is also observed in state-of-the-art LLMs. These findings suggest that LLMs acquire implicit reasoning through shortcut learning, enabling strong performance on tasks with similar patterns while lacking generalization.