textttReMind: Understanding Deductive Code Reasoning in LLMs

Large Language Models (LLMs) have achieved remarkable progress in code-related tasks. Despite their advancement, empirical evidence reveals that they still struggle with deductive code reasoning, the ability to reason about the program execution process. While prior studies have recognized this limitation, the underlying causes remain largely underexplored. In this paper, we begin by presenting a comprehensive empirical study that reveals three key challenges undermining deductive code reasoning: (1) an intrinsic gap between generation and reasoning abilities, (2) a consistent bias towards code sources, and (3) weak zero-shot generalization on complex benchmarks. In light of these challenges, we propose textttReMind, a multi-agent framework composed of textttMutator, textttExecutor, and textttInspector. The textttMutator generates code variants to mitigate bias towards code sources, the textttExecutor traces variable states step-by-step to expose inconsistency, and the textttInspector identifies problematic reasoning steps and provides control-flow refinement to bridge the intrinsic reasoning gap. Through their coordinated collaboration, textttReMind systematically identifies and refines reasoning flaws, achieving outstanding performance and enabling robust zero-shot generalization. Extensive experiments on two benchmarks with five LLMs demonstrate the superior advantages of textttReMind compared to baseline approaches in deductive code reasoning.