LLM-based memory systems increasingly maintain facts that evolve over time, where a recurring failure is conflict resolution: when a fact has multiple contradictory values, which should the agent return? MemoryAgentBench (MAB; Hu et al., 2026) makes this explicit in its FactConsolidation task: facts are numbered, the counterfactual has the higher serial, and agents are told newer facts have larger serials. Yet every published system underperforms: HippoRAG-v2 reaches 54% on single-hop (FC-SH), BM25 48%, Mem0 18%, and the temporal KG Zep/Graphiti just 7%. Multi-hop is near-unsolved (at most 7% across 22 systems). We argue the bottleneck is the assembly step: baselines leave conflict resolution to LLM-mediated retrieval or generation rather than version-aware aggregation. A matched-setup comparison (same backbone, retrieval, chunking, TOP_K) shows that replacing the LLM-judgment answer pipeline with candidate-extraction plus Python max(serial) yields +10.8 points on FC-SH (gpt-4o-mini), widening from +8 at 6K to +21 at 262K. This is a whole-pipeline effect (resolver, prompt, format, and temperature vary jointly); isolating the resolver is future work. The recipe reaches 78.0% on FC-SH (gpt-4o-mini), 94.8% (gpt-4o), and 30.2% on FC-MH (gpt-4o-mini, rising to 51.5% with gpt-4o) via a per-hop deterministic extension of Self-Ask. At matched-262K, it beats HippoRAG-v2 by +28 points and the best published FC-MH result by +20. The implication is corrective for the subfield: the bottleneck on conflict resolution is assembly (post-retrieval aggregation), not storage. A LongMemEval knowledge-update check shows the mechanism ports from max(serial) to max(timestamp) but only ties LLM judgment (57.8% vs 64.4%, n=45): deterministic aggregation is the right primitive for current-value conflicts and must be composed with question-type-aware handling for broader memory QA.

翻译：基于大模型的记忆系统需要持续维护随时间演变的事实，其中反复出现的故障点是冲突解决：当某个事实存在多个矛盾值时，智能体应返回哪一个？MemoryAgentBench（MAB；Hu等，2026）在其FactConsolidation任务中明确界定了此问题：事实按编号排列，反事实对应更高序号，且智能体被告知新事实的序号更大。然而，所有已发表系统表现均不理想：HippoRAG-v2在单跳任务（FC-SH）上达到54%，BM25为48%，Mem0为18%，时序知识图谱Zep/Graphiti仅达7%。多跳任务近乎未解决（22个系统中最高仅7%）。我们认为瓶颈在于聚合步骤：基线方法将冲突解决交由大模型驱动的检索或生成处理，而非采用版本感知式聚合。在匹配设置对比（相同主干模型、检索方式、分块策略、TOP_K参数）下，将基于大模型判断的答案管线替换为候选提取加Python max(序列号)操作，在FC-SH任务（gpt-4o-mini）上提升10.8个百分点，且增益随文档量从6K增至262K而从+8%扩大至+21%。此为全管线效应（解析器、提示词、格式和温度参数协同变化）；单独隔离解析器的影响留待未来研究。该方案通过基于Self-Ask的每跳确定性扩展，在FC-SH任务达到78.0%（gpt-4o-mini）与94.8%（gpt-4o），在FC-MH任务达到30.2%（gpt-4o-mini，使用gpt-4o时提升至51.5%）。在匹配262K文档量时，该方案分别超越HippoRAG-v2达28个百分点，超越已发表的FC-MH最佳结果达20个百分点。该结论对子领域具有修正意义：冲突解决的瓶颈在于聚合（检索后聚合），而非存储。LongMemEval知识更新检验表明，该机制可从max(序列号)移植至max(时间戳)，但仅能与大模型判断持平（57.8% vs 64.4%，n=45）：确定性聚合是当前值冲突的正确基元，需与问题类型感知处理结合以应对更广泛的记忆问答场景。