Code-Driven Inductive Synthesis: Enhancing Reasoning Abilities of Large Language Models with Sequences
This work addresses the problem of enhancing inductive reasoning abilities in large language models for AI researchers and developers, representing an incremental advancement in reasoning capabilities.
The paper tackles the lack of high-quality process supervision data for inductive reasoning in large language models by using number sequences packaged as algorithmic problems to generate synthetic training data, resulting in models that improve on code and comprehensive reasoning benchmarks.
Large language models make remarkable progress in reasoning capabilities. Existing works focus mainly on deductive reasoning tasks (e.g., code and math), while another type of reasoning mode that better aligns with human learning, inductive reasoning, is not well studied. We attribute the reason to the fact that obtaining high-quality process supervision data is challenging for inductive reasoning. Towards this end, we novelly employ number sequences as the source of inductive reasoning data. We package sequences into algorithmic problems to find the general term of each sequence through a code solution. In this way, we can verify whether the code solution holds for any term in the current sequence, and inject case-based supervision signals by using code unit tests. We build a sequence synthetic data pipeline and form a training dataset CodeSeq. Experimental results show that the models tuned with CodeSeq improve on both code and comprehensive reasoning benchmarks.