Organizations: State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China · School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China · Beijing Key Laboratory of Super Intelligent Security of Multi-Modal Information, Beijing, China · Alibaba Group, China · PeopleAI Inc., Beijing 100190, China · Gaoling School of Artificial Intelligence, Renmin University of China, Beijing 100872, China · School of Information Science and Technology, ShanghaiTech University, Shanghai 201210, China
For persistent and personalized conversational agents, memory systems can enable them to remember, update, and reason over long histories by storing past interactions and retrieving relevant information. Existing memory systems typically follow two paradigms: flat-structured memory and graph-based memory. The former is lightweight but leaves event relations and state updates implicit, while the latter explicitly models memory structure but incurs additional construction cost and introduces irrelevant relations over long histories. To address these limitations, we propose QGMem, a novel memory construction and activation framework motivated by human memory, in which experience is organized into events and query-relevant events are modeled by graph as working memory. QGMem converts long dialogue histories into event-indexed atomic memory units that preserve individual experiences and consolidates related units into dynamic memory traces that retain state trajectories and current states. When a query arrives, hybrid memory retrieval gathers complementary candidate memories, and query-aware reranking activates the most relevant units as a compact working memory. To expose relational dependencies in the working memory and support conflict-aware reasoning, QGMem organizes the working memory as a local graph, which is then encoded as a graph token and provided to the LLM together with the textual working memory to improve evidence utilization during answer generation. Experiments across six benchmarks validate the framework and show consistent gains in retrieval, multi-hop evidence composition, conflict resolution, and ultra-long dialogue reasoning with compact contexts and moderate inference cost.
Figures & tables
Fig. 1: Comparison of flat-structured memory, graph-based memory, and QGMem .
Fig. 2: Comparison of retrieval quality, LLM-judged answer score, and retrieved-context size on LoCoMo with Qwen3-8B. Colors distinguish the retrieval-augmented baseline, flat-structured memory, graph-based memory, and QGMem . The horizontal axis denotes normalized discounted cumulative gain at rank 10 (nDCG@10), the vertical axis denotes the LLM-judged score, and bubble area denotes the average number of retrieved-context tokens. Higher axis values and smaller bubble areas indicate better overall performance.
Fig. 3: Overall framework of QGMem .
Method
Accurate Retrieval
Conflict Resolution
Ultra-Long Scalability
LoCoMo
LongMemEval-S
Fact-SH
Fact-MH
BEAM-100K
BEAM-500K
N =1540
N =500
N =400
N =400
N =400
N =700
(a) LLM-judged Score (%)
Retrieval-Augmented Baseline
RAG
57.93±0.06
34.74±0.10
37.82±0.24
3.93±0.17
29.04
28.00
Flat-Structured Memory Baselines
TABLE I: Main results with Qwen3-8B across six long-term memory benchmarks. Boldface and underlining indicate the best and second-best performance respectively.
Figure 5Figure 6
\fnum@figure : A case study from LoCoMo. Mem0, LightMem, and EMem-G respectively illustrate ranking, evidence-utilization, and relation-selection failures, whereas QGMem promotes the answer-bearing comparison to Rank 1 and produces the grounded answer.
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
Hyperparameter
Value
Backbone LLM
Qwen3-8B, Llama3.1-8B-Instruct
LoRA
Disabled
Trainable modules
Graph encoder, projector
Graph encoder
4 TransformerConv layers
Hidden size
768
Attention heads
4
Appendix
\fnum@table : Training hyperparameters for the memory graph token.
Setting
Avg. Score
Full activation pipeline
82.77
w/o Hybrid memory retrieval
81.93
w/o Event-attribute retrieval
71.67
w/o reranking
73.82
Appendix
\fnum@table : Ablation of retrieval steps on LoCoMo with Qwen3-8B.
Agents for long term reasoning require a memory that can be efficiently and effectively updated over time, as new facts and external feedback continue to arrive. Recently, graph memory has been adopted to offer structural organization for multi-hop retrieval and reasoning. However, existing methods store all memories in a flat graph, and accumulated historical memories can introduce irrelevant contexts and increase the cost of evidence selection during retrieval. Moreover, they typically update memory units independently, requiring repeated unit-wise rewrite to cover related changes. To address these issues, we propose HiGram, an evolving hierarchical graph memory framework with path-level localization and rewriting. Specifically, we first propose a hierarchical graph memory, which organizes the memory into coarse-to-fine architecture composed of upper-level nodes and MemoryUnits, thereby reducing the amount of irrelevant information during retrieval. We further propose MicroGraph-based path-level localization, which leverages query and update conditioned MicroGraphs to identify support subgraph and evidence path before rewrite. Finally, we propose a coordinated rewriting method that jointly revises intra-unit memory and inter-unit dependencies, enable valid dependency structures updating in the localized evidence path. Experiments on benchmarks for long-term conversational question answering and conflict-aware memory evaluation demonstrate that our method demonstrate substantial improvements over baselines in answer quality and token efficiency. Besides, our method improves answer accuracy and query-valid evidence selection under dynamic, static, and conditional conflicts.
Xiawei Yue, Boran Wang, Xiaoqing Zhang +2
Nankai University, Tianjin, China · Zhongguancun Academy, Beijing, China · Beihang University, Beijing, China
Long-term conversational large language model (LLM) agents require memory systems that can recover relevant evidence from historical interactions without overwhelming the answer stage with irrelevant context. However, existing memory systems, including hierarchical ones, still often rely solely on vector similarity for retrieval. It tends to produce bloated evidence sets: adding many superficially similar dialogue turns yields little additional recall, but lowers retrieval precision, increases answer-stage context cost, and makes retrieved memories harder to inspect and manage. To address this, we propose HiGMem (Hierarchical and LLM-Guided Memory System), a two-level event-turn memory system that allows LLMs to use event summaries as semantic anchors to predict which related turns are worth reading. This allows the model to inspect high-level event summaries first and then focus on a smaller set of potentially useful turns, providing a concise and reliable evidence set through reasoning, while avoiding the retrieval overhead that would be excessively high compared to vector retrieval. On the LoCoMo10 benchmark, HiGMem achieves the best F1 on four of five question categories and improves adversarial F1 from 0.54 to 0.78 over A-Mem, while retrieving an order of magnitude fewer turns. Code is publicly available at https://github.com/ZeroLoss-Lab/HiGMem.
Shuqi Cao, Jingyi He, Fei Tan
East China Normal University, Shanghai, China · Shanghai Jiao Tong University, Shanghai, China
Large Language Models (LLMs) serving as long-term dialogue agents require memory systems that support reliable reasoning over extended interactions. However, existing Retrieval-Augmented Generation (RAG) frameworks typically treat memory as passive storage, making it difficult to distinguish source-attributed beliefs from unattributed event/fact records and to connect evidence dispersed across sessions. We introduce CogMem, a cognitive memory architecture based on the PEC2F (Person-Event-Concept-Claim-Fact) graph schema. Dedicated Claim nodes preserve the source and target of subjective statements, while Fact and Event nodes represent semantic and episodic knowledge. Dialogue turns are incrementally converted into provenance-aware graph records, consolidated into higher-level facts, and reconciled into temporally scoped Claim views when the same source provides conflicting updates. For retrieval, a rule-based controller driven by LLM intent parsing composes four deterministic graph operators---anchoring, traversal, intersection, and evidence grounding---to reconstruct query-relevant context. Experiments on LoCoMo and LongMemEval show strong performance, especially on multi-hop, temporal, and knowledge-update tasks. Ablations and a semantic-collapse probe support complementary contributions from epistemic separation, consolidation, and agentic retrieval. Code: https://github.com/Silent-Rain02/CogMem.
Zirui Liao, Zhengxian Wu, Zhuohong Chen +4
Tsinghua University Shenzhen International Graduate School, Shenzhen, China