InductionBench: LLMs Fail in the Simplest Complexity Class
Wenyue Hua, Tyler Wong, Sun Fei, Liangming Pan, Adam Jardine, William Yang Wang
TL;DR
InductionBench formalizes inductive reasoning evaluation by assessing LLMs on string-to-string transformations drawn from the subregular hierarchy (ISL, L-OSL, R-OSL) with controlled parameters $k$ and $|\,\Sigma|$. The benchmark emphasizes minimal, non-redundant rule representations and uses characteristic samples to test whether models infer the underlying function rather than memorize data. Across zero-shot CoT prompts, strong LLMs struggle as problem complexity grows, with performance dominated by the context window size and the number of rules rather than alphabet size, and few-shot prompts offering limited gains under high complexity. To address potential shortcuts, the authors propose an IOSL-based exploration leaderboard and provide a rigorous, algorithmically grounded framework for future work in inductive reasoning, highlighting a meaningful gap between current LLM capabilities and robust inductive generalization.
Abstract
Large language models (LLMs) have shown remarkable improvements in reasoning and many existing benchmarks have been addressed by models such as o1 and o3 either fully or partially. However, a majority of these benchmarks emphasize deductive reasoning, including mathematical and coding tasks in which rules such as mathematical axioms or programming syntax are clearly defined, based on which LLMs can plan and apply these rules to arrive at a solution. In contrast, inductive reasoning, where one infers the underlying rules from observed data, remains less explored. Such inductive processes lie at the heart of scientific discovery, as they enable researchers to extract general principles from empirical observations. To assess whether LLMs possess this capacity, we introduce InductionBench, a new benchmark designed to evaluate the inductive reasoning ability of LLMs. Our experimental findings reveal that even the most advanced models available struggle to master the simplest complexity classes within the subregular hierarchy of functions, highlighting a notable deficiency in current LLMs' inductive reasoning capabilities. Coda and data are available https://github.com/Wenyueh/inductive_reasoning_benchmark.