DimMem: Dimensional Structuring for Efficient Long-Term Agent Memory

Unknown Date

AI Summary

This work addresses the trade-off between fidelity and efficiency in long-term memory for large language model agents: maintaining full dialogue history is computationally expensive, while flat summarization often discards critical structural information. To resolve this, the authors propose DimMem, a novel framework that introduces explicit dimensional structured memory representation, modeling memories as self-contained atomic units encompassing time, location, cause, purpose, and keywords. This approach enables dimension-aware retrieval, updating, and context-selective recall without storing raw dialogues, achieving both high precision and efficiency. Experimental results show that DimMem, fine-tuned with a lightweight extractor such as Qwen3-4B, attains 81.43% accuracy on LoCoMo-10 and 78.20% on LongMemEval-S, outperforming existing lightweight memory systems while reducing per-query token cost by 24%.

Abstract

Large language model (LLM) agents require long-term memory to leverage information from past interactions. However, existing memory systems often face a fidelity--efficiency trade-off: raw dialogue histories are expensive, while flat facts or summaries may discard the structure needed for precise recall. We propose \textbf{DimMem}, a lightweight dimensional memory framework that represents each memory as an atomic, typed, and self-contained unit with explicit fields such as time, location, reason, purpose, and keywords. This representation exposes the structure needed for dimension-aware retrieval, memory update, and selective assistant-context recall without storing full histories in the model context. Across LoCoMo-10 and LongMemEval-S, DimMem achieves \textbf{81.43\%} and \textbf{78.20\%} overall accuracy, respectively, outperforming existing lightweight memory systems while reducing LoCoMo per-query token cost by \textbf{24\%}. We further show that dimensional memory extraction is learnable by compact models: after fine-tuning on the DimMem schema, a Qwen3-4B extractor surpasses LightMem with GPT-4.1-mini on both benchmarks and reaches performance comparable to, or better than, much larger extractors in key settings. These results suggest that explicit dimensional structuring is an effective and efficient foundation for long-term memory in LLM agents. Code is available at https://github.com/ChowRunFa/DimMem.

Problem

• long-term memory
• fidelity-efficiency trade-off
• memory structuring
• LLM agents
• dimensional representation

Innovation

• dimensional memory
• long-term agent memory
• structured memory representation
• efficient retrieval
• LLM agents