Enhancing Reasoning Capabilities of LLMs via Principled Synthetic Logic Corpus
This addresses reasoning limitations in LLMs, which is a critical bottleneck for AI applications, though it is incremental as it builds on existing training methods.
The paper tackles the problem of LLMs struggling with reasoning by proposing Additional Logic Training (ALT) using a program-generated synthetic logic corpus, resulting in gains of up to 30 points on logical reasoning benchmarks and up to 10 points on math and coding benchmarks.
Large language models (LLMs) are capable of solving a wide range of tasks, yet they have struggled with reasoning. To address this, we propose $\textbf{Additional Logic Training (ALT)}$, which aims to enhance LLMs' reasoning capabilities by program-generated logical reasoning samples. We first establish principles for designing high-quality samples by integrating symbolic logic theory and previous empirical insights. Then, based on these principles, we construct a synthetic corpus named $\textbf{Formal Logic Deduction Diverse}$ ($\textbf{FLD}$$_{\times 2}$), comprising numerous samples of multi-step deduction with unknown facts, diverse reasoning rules, diverse linguistic expressions, and challenging distractors. Finally, we empirically show that ALT on FLD$_{\times2}$ substantially enhances the reasoning capabilities of state-of-the-art LLMs, including LLaMA-3.1-70B. Improvements include gains of up to 30 points on logical reasoning benchmarks, up to 10 points on math and coding benchmarks, and 5 points on the benchmark suite BBH.