Recent advances in Large Language Models (LLMs) have enabled conversational AI agents to engage in extended multi-turn interactions spanning weeks or months. However, existing memory systems struggle to reason over temporally grounded facts and preferences that evolve across months of interaction and lack effective retrieval strategies for multi-hop, time-sensitive queries over long dialogue histories. We introduce Chronos, a novel temporal-aware memory framework that decomposes raw dialogue into subject-verb-object event tuples with resolved datetime ranges and entity aliases, indexing them in a structured event calendar alongside a turn calendar that preserves full conversational context. At query time, Chronos applies dynamic prompting to generate tailored retrieval guidance for each question, directing the agent on what to retrieve, how to filter across time ranges, and how to approach multi-hop reasoning through an iterative tool-calling loop over both calendars. We evaluate Chronos with 8 LLMs, both open-source and closed-source, on the LongMemEvalS benchmark comprising 500 questions spanning six categories of dialogue history tasks. Chronos Low achieves 92.60% and Chronos High scores 95.60% accuracy, setting a new state of the art with an improvement of 7.67% over the best prior system. Ablation results reveal the events calendar accounts for a 58.9% gain on the baseline while all other components yield improvements between 15.5% and 22.3%. Notably, Chronos Low alone surpasses prior approaches evaluated under their strongest model configurations.
Figures & tables
Figure 1: An overview of the Chronos Architecture. Event Extraction, Dual Indexing, and Query Processing result in a generated answer.
Method
Overall
Category-Level Accuracy (%)
KU
MS
SSA
SSP
SSU
TR
Chronos Low (Ours)
92.60
96.15
91.73
100.00
80.00
94.29
90.23
Honcho †
90.40
94.87
84.96
96.43
90.00
94.29
88.72
EmergenceMem Internal
86.00
83.33
81.20
100.00
60.00
98.57
85.71
Mastra
84.80
85.90
79.70
82.14
73.33
98.57
85.71
Supermemory
81.60
88.50
71.40
96.40
70.00
97.10
76.70
Table 1: Comparison of Chronos Low with state-of-the-art conversational memory systems on LongMemEval.
Method
Overall
Category-Level Accuracy (%)
KU
MS
SSA
SSP
SSU
TR
Chronos High (Ours)
95.60
100.00
88.72
100.00
100.00
98.57
95.50
Honcho †
92.60
-
-
-
-
-
-
Mastra
92.80
94.90
87.20
96.40
90.00
97.10
94.00
Supermemory
85.20
89.70
76.70
98.20
70.00
98.60
82.00
Hindsight ‡
91.40
94.90
87.20
96.40
80.00
97.10
91.00
Table 2: Comparison of systems with more advanced LLMs on LongMemEval.
Figure 2: Overall Benchmark Accuracy on both High and Low Configurations. Note: High configurations refer to evaluations with advanced frontier models, such as Opus 4.6 and Gemini 3 Pro. Standard configurations refer to the traditional evaluated model, GPT-4o, or similar models.
Figure 3: Category-wise Error rate comparison across Chronos variations. Chronos High demonstrates marked improvement on nearly every category.
Model
Ablation
Overall
KU
MS
SSA
SSP
SSU
TR
Chronos High
Chronos
94.8
100.0
90.3
100.0
100.0
93.8
93.5
No Initial Retrieval
91.4
100.0
80.6
100.0
85.7
93.8
93.5
No Dynamic Prompting
94.8
100.0
87.1
100.0
100.0
93.8
96.8
No Rerank
92.2
100.0
87.1
100.0
85.7
87.5
93.5
No Date Filter
92.2
94.4
83.9
100.0
85.7
93.8
96.8
Grep Only (no vector)
87.1
94.4
83.9
61.5
100.0
93.8
90.3
Table 3: Ablation results grouped by model configuration on a 116-question sample of the benchmark.
Large language models still struggle with reliable long-term conversational memory: simply enlarging context windows or applying naive retrieval often introduces noise and destabilizes responses. We present APEX-MEM, a conversational memory system that combines three key innovations: (1) a property graph which uses domain-agnostic ontology to structure conversations as temporally grounded events in an entity-centric framework, (2) append-only storage that preserves the full temporal evolution of information, and (3) a multi-tool retrieval agent that understands and resolves conflicting or evolving information at query time, producing a compact and contextually relevant memory summary. This retrieval-time resolution preserves the full interaction history while suppressing irrelevant details. APEX-MEM achieves 88.88% accuracy on LOCOMO's Question Answering task and 86.2% on LongMemEval, outperforming state-of-the-art session-aware approaches and demonstrating that structured property graphs enable more temporally coherent long-term conversational reasoning.
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
Long-term memory has been widely adopted in LLM-based agents to support persistent interaction and personalized assistance. Long-term multi-party interactions, however, accumulate substantially more detailed information, with decisions repeatedly updated and states evolving across participants and interaction stages. Temporal structure is therefore crucial for tracing such memories: relevant information often appears densely within short intervals, while transitions among different states are anchored by a strict temporal order. Existing temporally aware memory frameworks primarily store time as metadata or use temporal signals to guide retrieval over preconstructed memory structures, without using temporal structure to both recover locally concentrated evidence and reconstruct its complete event progression. To address these limitations, we propose T-Mem, a temporal-structure-based memory framework that preserves interactions and organizes query-relevant evidence through two temporal mechanisms. Temporal Evidence Discovery exploits the local concentration of related messages to recover overlooked evidence, while Temporal Memory Structuring models state transitions and supporting relations over time and compresses low-information spans within the resulting forest. Evidence Forest Linearization then serializes the structured evidence for answer generation. Experiments on two long-term multi-party memory benchmarks demonstrate that T-Mem consistently outperforms competitive memory frameworks across LLMs, parameter settings, domains, and query types.
Junyou Guo, Chongyang Tao, Yan He +2
School of Computer Science and Engineering, Beihang University, Beijing, China · Alibaba Group