AI 中文总结
该研究针对LLM智能体记忆系统的状态追踪能力不足问题，构建StateMemBench基准，提出StateMem方法，可显著提升当前状态准确率，且作为轻量包装器适配现有系统。
AI 中文摘要
随着基于大语言模型（LLM）的智能体被部署用于更长、更高风险的任务，其记忆系统仍存在关键缺陷。现有记忆基准大多聚焦于回忆类任务，而我们认为，有效的记忆系统必须能追踪世界的演化状态：随着事实、约束和决策在长期交互中被修正，答案必须反映当前状态，而非已被取代的状态。我们将这一能力定义为状态追踪，并在StateMemBench中实现，该基准包含234个多会话场景，覆盖两种对话长度 regime（ regime 保留原专业术语）。其闭集评分标准会判断答案是否反映当前状态、已被取代的状态，或其他情况，从结构上区分状态追踪失败与其他错误。我们的分析显示，该任务对现有记忆系统、检索增强基线及长上下文基线均具挑战性。随后，我们提出StateMem，一种以状态为先的记忆方法，明确追踪取代关系与关联依赖，结果显示，在DeepSeek-V4-Flash上，它将当前状态准确率较最强同骨干基线提升1.8倍（0.205→0.363），在Qwen-3.5-9B上较最强记忆系统提升1.6倍（0.149→0.233），同时与长上下文基线表现相当。最后，我们证明该状态方法可作为轻量单调用包装器应用于现有记忆系统，在StateMemBench的6种记忆与检索骨干上，将当前状态准确率提升32至67个百分点；长度与成本匹配的对照实验显示，其中15至32个百分点的提升源于状态结构而非新增上下文。
英文摘要
As LLM-based agents are deployed for longer and higher-stakes tasks, their memory systems continue to have crucial gaps. While existing memory benchmarks focus largely on recall-shaped tasks, we argue an effective memory system must track the evolving state of the world; as facts, constraints, and decisions are revised over a long interaction, answers must reflect the current state and not a superseded one. We define this capability as state tracking and instantiate it in StateMemBench, a benchmark of 234 multi-session scenarios spanning two conversation-length regimes. Its closed-pool grading scores whether an answer reflects the current state, the superseded state, or fails otherwise, separating state-tracking failures from other errors by construction. Our analysis shows that this task is challenging for existing memory systems, retrieval-augmented baselines, and long-context baselines. We then present StateMem, a state-first memory method that explicitly tracks supersession and relational dependencies, and show it improves current-state accuracy over the strongest same-backbone baseline by 1.8x (0.205 -> 0.363) on DeepSeek-V4-Flash and over the strongest memory system by 1.6x (0.149 -> 0.233) on Qwen-3.5-9B, while remaining competitive with the long-context baselines. Finally, we show the same state approach can be applied as a lightweight single-call wrapper over existing memory systems, lifting current-state accuracy by +32 to +67 points on StateMemBench across six memory and retrieval backends. A length- and cost-matched control attributes +15 to +32 of those points to state structure rather than added context.
