eMoT: evolving Memory-of-Thought via Symbolic Anchoring and Memory Corrosion
2026-06-01 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors identify that big language models often make mistakes in complex reasoning tasks because they treat problem-solving like a one-time effort without remembering what worked before. To fix this, they created eMoT, a system that keeps track of reasoning steps as evolving memories and uses Python code for exact calculations. Their approach gradually forgets less useful strategies, checks for consistency between logic and math, and improves accuracy on multiple math problems. Notably, eMoT reached 100% accuracy on the Game of 24 and outperformed bigger models by focusing on better reasoning control rather than size alone.
Large Language ModelsMulti-step ReasoningChain-of-ThoughtMemory MechanismSymbolic ComputationPython AnchoringNumerical ComputationConsistency RefinementMathematical ReasoningGame of 24
Authors
Xiang Li, Jiwei Wei, Ke Liu, Yitong Qin, Jinyu Guo, Malu Zhang, Peng Wang, Yang Yang
Abstract
While Large Language Models (LLMs) achieve impressive performance on multi-step reasoning tasks, their reliability is persistently hindered by critical limitations such as unconstrained hallucinations and poor numerical computation. Fundamentally, these issues arise because standard models treat reasoning as a transient, one-off generation process rather than retaining and refining successful procedural logic. To address these challenges, we propose eMoT (evolving Memory-of-Thought), a unified framework that stabilizes multi-step reasoning by treating reasoning trajectories as dynamic, evolving memories rather than static templates. The framework primarily consists of three interconnected modules: (i) a memory corrosion mechanism that reinforces high-utility reasoning structures while gradually decaying less frequent ones; (ii) a symbolic anchoring engine that utilizes Python for deterministic computation, much like a human uses a calculator; and (iii) a consistency-driven refinement process that aligns neural inference with symbolic outcomes, reducing the accumulation of logical discrepancies. Across multiple reasoning benchmarks, eMoT improves accuracy and solution consistency over standard Chain-of-Thought and structured reasoning baselines.On the traditional task Game of 24, eMoT achieves 100% accuracy, surpassing the baseline by up to 17.6%. Evaluations on mathematical task GSM8K, ASDiv, SVAMP, and MGSM further show consistent gains in multi-step mathematical reasoning. In our evaluation, we achieve superior performance despite utilizing a lightweight backbone model with constrained baseline capabilities. Compared to alternative methods that rely on massively scaled models, our results demonstrate that the performance gains are fundamentally driven by the eMoT framework's reasoning control rather than sheer model size.