ActMem: Bridging the Gap Between Memory Retrieval and Reasoning in LLM Agents

2026-02-04

arXiv.org

AI Summary

This work addresses the limitation of existing large language model agents, whose memory mechanisms typically treat memory as passive recording and thus struggle to support complex reasoning and decision-making. To overcome this, the authors propose ActMem, a novel framework that deeply integrates causal reasoning into the memory system. ActMem constructs a causal semantic graph to transform unstructured dialogue history into structured representations and leverages counterfactual reasoning and commonsense completion to infer implicit constraints, thereby reconciling conflicts between historical states and current intents. The contributions include an actionable memory architecture and ActMemEval, the first memory evaluation benchmark specifically designed for logical reasoning. Experimental results demonstrate that ActMem significantly outperforms existing approaches on memory-dependent complex tasks, substantially enhancing the consistency and reliability of agent behavior.

Abstract

Effective memory management is essential for large language model (LLM) agents handling long-term interactions. Current memory frameworks typically treat agents as passive"recorders"and retrieve information without understanding its deeper implications. They may fail in scenarios requiring conflict detection and complex decision-making. To bridge this critical gap, we propose a novel actionable memory framework called ActMem that integrates memory retrieval with active causal reasoning. ActMem transforms unstructured dialogue history into a structured causal and semantic graph. By leveraging counterfactual reasoning and commonsense completion, it enables agents to deduce implicit constraints and resolve potential conflicts between past states and current intentions. Furthermore, we introduce a comprehensive dataset ActMemEval to evaluate agent reasoning capabilities in logic-driven scenarios, moving beyond the fact-retrieval focus of existing memory benchmarks. Experiments demonstrate that ActMem significantly outperforms state-of-the-art baselines in handling complex, memory-dependent tasks, paving the way for more consistent and reliable intelligent assistants.

Problem

• memory retrieval
• reasoning
• LLM agents
• complex decision-making
• causal reasoning

Innovation

• actionable memory
• causal reasoning
• counterfactual reasoning
• structured memory graph
• LLM agents