AI TrackerFrontier desk · AEST
Back to news
normalaitech

From Natural Language to Exact Cover: A Neuro-Symbolic Approach to Zebra Puzzles

arXiv cs.AI · 3 Nov 2026, 04:58 AEST

A new neuro-symbolic framework enhances the performance of Large Language Models on Zebra puzzles. This approach combines neural semantic parsing with deterministic exact cover reasoning to improve logical reasoning capabilities.

Key facts

Read aloud

A new neuro-symbolic framework enhances the performance of Large Language Models on Zebra puzzles. This approach combines neural semantic parsing with deterministic exact cover reasoning to improve logical reasoning capabilities. Chain-of-Thought CoT generation has substantially improved the performance of Large Language Models L L M s on complex reasoning tasks, including code generation, data analysis, and exam-style question answering. Despite these advances, purely neural L L M s continue to struggle with elementary logical reasoning problems and lack the determinism, soundness, and reliability characteristic of symbolic reasoning systems. In this work, we present a synergistic neuro-symbolic framework that bridges this gap by combining neural semantic parsing with deterministic exact cover reasoning. We conduct a comprehensive empirical evaluation across multiple parameter scales, post-training paradigms, and L L M families.

Sources in this cluster