Sudoku-Bench: Evaluating creative reasoning with Sudoku variants

TL;DR

Sudoku-Bench is a new benchmark designed to evaluate creative and multi-step logical reasoning in large language models using challenging Sudoku variants that require novel problem-solving strategies.

Contribution

The paper introduces Sudoku-Bench, a curated set of unconventional Sudoku puzzles that effectively assess creative reasoning and provide tools for broad research application.

Findings

State-of-the-art LLMs solve less than 15% of puzzles unaided.

Sudoku variants challenge memorization, requiring logical breakthroughs.

The benchmark facilitates consistent evaluation of reasoning abilities.

Abstract

Existing reasoning benchmarks for large language models (LLMs) frequently fail to capture authentic creativity, often rewarding memorization of previously observed patterns. We address this shortcoming with Sudoku-Bench, a curated benchmark of challenging and unconventional Sudoku variants specifically selected to evaluate creative, multi-step logical reasoning. Sudoku variants form an unusually effective domain for reasoning research: each puzzle introduces unique or subtly interacting constraints, making memorization infeasible and requiring solvers to identify novel logical breakthroughs (``break-ins''). Despite their diversity, Sudoku variants maintain a common and compact structure, enabling clear and consistent evaluation. Sudoku-Bench includes a carefully chosen puzzle set, a standardized text-based puzzle representation, and flexible tools compatible with thousands of publicly available puzzles -- making it easy to extend into a general research environment. Baseline experiments show that state-of-the-art LLMs solve fewer than 15\% of puzzles unaided, highlighting significant opportunities to advance long-horizon, strategic reasoning capabilities.