Enhancing memory retrieval in generative agents through LLM-trained cross attention networks
Open accessThe DOI matches its Crossref record; 68 of the 70 references it cites were found in a registry or an index; 10 of the 12 sources its claims cite were checked for support.
Abstract
Introduction The surge in the capabilities of large language models (LLMs) has propelled the development of Artificial General Intelligence (AGI), highlighting generative agents as pivotal components for emulating complex AI behaviors. Given the high costs associated with individually training LLMs for each AI agent, there is a critical need for advanced memory retrieval mechanisms to maintain the unique characteristics and memories of individual AI agents. Methods In this research, we developed a text-based simulation of a generative agent world, constructing a community with multiple agents…
Overview
CoChat Summary
## TL;DR
The paper proposes ACAN, a cross-attention-based memory retrieval network for generative agents, trained with LLM feedback against a weighted memory retrieval baseline. In simulations with eight agents, ACAN improved memory quality and event attendance compared with the baseline.
## Key findings
- In a 16-hour/day simulation, the authors generated 1,280 unique training entries from 8 agents over 3 consecutive days; test simulations produced 435 data entries from 8 agents over 1 day.
- In the memory-quality evaluation, ACAN scored 5.94 ± 1.66 versus 5.05 ± 1.88 for the WMR baseline; the difference was significant (T = 7.44, P = 2.42 × 10−13).
- In the invitation/attendance experiment over 10 days with 100 invitation and attendance records, ACAN agents attended 32.6% of invitations on average versus 24.6% for WMR; the paired t-test gave T = 11.31, P = 0.00035.
- The model’s training loss reportedly fell from 1.5 to 0.12.
## Methods
- The authors built a text-based generative agent simulation world with 8 predefined agents (Farmer, Grocer, Doctor, Mayor, Chef, Hunter, Blacksmith, Carpenter) and multiple locations; interactions were simulated using ChatGPT (GPT-3.5-turbo).
- For each decision step, they stored the agent’s current state, memory bank, and retrieved memories, embedded with text-embedding-ada-002; the training set included current state q, memory bank M_q, and WMR-retrieved memories m_r′ ranked by Recency, Importance, and Relevance.
- They trained an Auxiliary Cross Attention Network (ACAN) with Adam (lr = 0.001, batch size = 16), using LLM-based scoring of ACAN-retrieved memories versus baseline-retrieved memories to define the loss; retrieval output was the top five memories ranked by attention weights.
## Limitations
- The paper states that reliance on LLMs for training and evaluation increases computational demands and cost, and that LLM feedback can slow training and limit scalability.
- The paper states that the LLM-based evaluation may not generalize across different configurations or domains, and that using LLMs instead of human assessment may affect rigor and objectivity.
- Evident from the text: the empirical validation is based on a small simulated setup with 8 agents in a text-only environment, so broader real-world generalization is not established.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures




Verification
Claims the paper makes
7 Valid · 2 Not valid · 2 Not checked · 39 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recent studies have demonstrated that large language models (LLMs) can serve as reliable evaluators for natural language generation (NLG) tasks, exhibiting strong correlation with human judgments.”
§ Enhancing memory retrieval in generative agents through LLM-trained cross attention networks > 3 Methods > 3.2 Auxiliary cross attention network for memory retrieval
What the sources say
- Supportedcites Wang et al. (2023) matched to “Is chatGPT a good nlg evaluator? A preliminary study” (2023) source
The Introduction reports that ChatGPT had high correlation with human judgment in most cases across the NLG datasets tested, supporting its potential as an evaluator. The Limitations section qualifies this result: performance depends on prompts and datasets, and the experiments covered English-language tasks only. (LLM, conf=0.89)
- Partly supportedcites Chiang and Lee (2023) matched to “Can Large Language Models Be an Alternative to Human Evaluations?” (2023) sourceExperimental — accuracy not yet measured
The source does support LLMs evaluating NLG output consistently with human experts: the Abstract (p. 1) states 'the result of LLM evaluation is consistent with the results obtained by expert human evaluation', and the Contributions (p. 1) and Discussion (p. 8) claim LLM evaluation 'produces results similar to expert human evaluation' and 'has the potential to be used to evaluate NLP systems'. However, the specific assertion of 'strong correlation with human judgments' overstates the reported figures: §3.3.1 (pp. 5-6, Table 2) gives Kendall's tau between English teachers and text-davinci-003 of 0.12-0.43, which the paper itself describes as 'weak to strong positive correlations' under its own benchmark where only |tau| >= 0.3 counts as strong — i.e. only relevance, not grammaticality, cohesiveness or likability. The paper also hedges reliability, noting LLMs 'are not always suitable to be used for LLM evaluation' (§5, p. 8) and recommending LLM evaluation be used alongside, not instead of, human evaluation (p. 10). (LLM, conf=0.78)
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“To achieve this goal, the ideal approach would be to train a dedicated LLM for each agent. However, considering the complexity of LLM training (Yang et al., 2024) and the practical demands of a large variety and number of agents, this approach is impractical.”
§ Enhancing memory retrieval in generative agents through LLM-trained cross attention networks > 1 Introduction
What the sources say
- Supportedcites Yang et al., 2024 matched to “Harnessing the power of llms in practice: a survey on chatGPT and beyond” (2024) source
The citation is attached to the complexity of LLM training, not to the citing authors’ conclusion about training one model per agent. The paper’s Cost section describes substantial training costs, energy consumption, data volumes, computing requirements, and hardware needs, supporting that attributed point. (LLM, conf=0.90)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Existing memory retrieval methods for agents primarily focus on relevance and temporal validity.”
§ Enhancing memory retrieval in generative agents through LLM-trained cross attention networks > 3 Methods > 3.2 Auxiliary cross attention network for memory retrieval
What the sources say
- Partly supportedcites Hou et al. (2024) matched to “'My agent understands me better': Integrating Dynamic Human-like Memory Recall and Consolidation in LLM-Based Agents” (2024) sourceExperimental — accuracy not yet measured
Section 2.2 describes MemoryBank retrieving memories by relevance and accounting for elapsed time in forgetting. Section 2.4 describes Generative Agents using recency and relevance, but also importance. These examples support the role of relevance and time-related factors, but the paper does not establish that existing methods primarily focus on those two factors. (LLM, conf=0.83)
- Partly supportedcites Alonso et al. (2024) matched to “Toward conversational agents with context and time sensitive long-term memory” (2024) sourceExperimental — accuracy not yet measured
The Introduction (p. 2) says existing conversational-agent retrieval models primarily use semantic vector-database search, supporting the relevance part of the claim. But the Related Works section (p. 4) says these systems do not directly test retrieval based on temporal metadata, so the paper does not support saying they primarily focus on temporal validity. (LLM, conf=0.87)
- Partly supportedcites Park et al. (2023) matched to “Generative Agents: Interactive Simulacra of Human Behavior” (2023) sourceExperimental — accuracy not yet measured
The Introduction and §4.1 (page numbers unavailable) show that this paper’s retrieval model uses relevance and recency, which supports part of the claim. But it also uses importance, and the cited passages do not establish that existing agent-retrieval methods primarily focus on the first two factors. (LLM, conf=0.79)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“recent work such as Edge et al. (2024) demonstrates that LLMs can reliably evaluate relevance and faithfulness in RAG systems, supporting their use as cost-effective alternatives to human assessments.”
§ Enhancing memory retrieval in generative agents through LLM-trained cross attention networks > 5 Discussion
What the sources say
- Partly supportedcites Edge et al. (2024) matched to “From local to global: a graph rag approach to query-focused summarization” (2024) sourceExperimental — accuracy not yet measured
Section 3.3 describes using an LLM to compare RAG answers, but its criteria are comprehensiveness, diversity, and empowerment—not relevance or faithfulness. The RAG evaluation criteria discussion explicitly calls context relevance, faithfulness, and answer relevance unsuitable for this paper’s global-sensemaking task. Section 2.4 cites prior work on LLM evaluations competitive with human evaluations; this paper does not demonstrate reliable relevance and faithfulness judgments or establish a cost-effective alternative to human assessment. The indexed passages provide no page numbers. (LLM, conf=0.88)
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Our empirical evaluations demonstrate that this approach substantially enhances the quality of memory retrieval, thereby increasing the adaptability and behavioral consistency of agents in fluctuating environments.”
- Also stated in: § 1 Introduction · boosted
“Our experiments demonstrate that ACAN substantially outperforms traditional methods in memory retrieval, resulting in enhanced adaptability of agents and more effective interactions with their environment and other agents.”
- Also stated in: § 1 Introduction · plain
“By calculating attention weights between the agent's current state and all memories in the memory bank, ACAN ranks and retrieves the most relevant memories, leading to Enhanced Memory Retrieval compared to base methods.”
§ Enhancing memory retrieval in generative agents through LLM-trained cross attention networks > Abstract > Results
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 1 Introduction · boosted
13 more claims
Create a free account to see all 50 claims and every source verdict.
- Probably not a substantial claim. 18
- It may state two separate claims as one. 18
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Online memory handling typically consists of structured subtasks (such as intent routing (Zhang et al., 2024), query construction (Hong and He, 2025) and semantic filtering (Hatalis et al., 2023)).”
From Lightweight LLM Agent Memory with Small Language Models (2026) · cites this paper as Hong and He, 2025
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Online memory handling typically consists of structured subtasks (such as intent routing (Zhang et al., 2024), query construction (Hong and He, 2025) and semantic filtering (Hatalis et al., 2023)).”
From Lightweight LLM Agent Memory with Small Language Models (2026) · cites this paper as Hong and He, 2025
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“reflection focused on trajectories may lead to a linear expansion of the memory bank, which imposes a burden for inference and may potentially result in a characteristic of following trajectories (Hong and He, 2025; Zhu et al., 2025; Fu et al., 2025).”
From From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanisms (2026) · cites this paper as Hong and He, 2025
Experimental — accuracy not yet measured
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Language Models are Few-Shot Learners
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.14165 · Cited by 23,372
Artificial intelligence: a modern approach
(1995) · Choice Reviews Online · doi:10.5860/choice.33-1577 · Cited by 21,974
Kaugeranna, R., Kaugeranna, E., & Aion, (. S. 4. 6. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.giscience.2023.43 · Cited by 13,690
Stanford Encyclopedia of Philosophy 2009: Building an Endowment with Community Support
Loy, M. (2015) · doi:10.18665/sr.22353 · Cited by 2,465
Generative Agents: Interactive Simulacra of Human Behavior
Park, J. S., O'Brien, J., Cai, C. J., Morris, M. R., Liang, P., & Bernstein, M. S. (2023) · Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3586183.3606763 · Cited by 2,187
A survey on large language model based autonomous agents
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., et al. (2024) · Frontiers of Computer Science · doi:10.1007/s11704-024-40231-1 · Cited by 1,941 · cited 1×
“Consequently, an increasing number of researchers are exploring the use of LLMs to develop Generative Agents with robust content generation capabilities ( Park et al., 2023; Liu et al., 2023; Wang et al., 2024).”
A Survey on Large Language Model based Autonomous Agents
Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., et al. (2023) · arXiv · doi:10.48550/arxiv.2308.11432 · Cited by 1,925
Expel: LLM agents are experiential learners
Zhao, A., Huang, D., Xu, Q., Lin, M., Liu, Y.-J., & Huang, G. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i17.29936 · Cited by 1,451
Squire, L. R. (1986) · Science · doi:10.1126/science.3086978 · Cited by 1,278 · cited 1×
“Just as the human brain utilizes prior experiences for adaptive behavior (Squire, 1986; Schwabe et al., 2014), agents require specialized memory mechanisms to effectively manage sequential tasks.”
Emergent Abilities of Large Language Models
Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., et al. (2022) · arXiv · doi:10.48550/arxiv.2206.07682 · Cited by 1,264 · cited 1×
“Numerous studies have validated the exceptional performance of LLMs when appropriately prompted in downstream tasks, showcasing their versatility and intelligence (Bang et al., 2023; Wei et al., 2022).”
The role of emotion in believable agents
Bates, J. (1994) · Communications of the ACM · doi:10.1145/176789.176803 · Cited by 1,177
Reflexion: Language Agents with Verbal Reinforcement Learning
Shinn, N., Cassano, F., Berman, E., Gopinath, A., Narasimhan, K., & Yao, S. (2023) · arXiv · doi:10.48550/arxiv.2303.11366 · Cited by 1,087 · cited 1×
“Therefore, the common practice is to store the agent’s memories externally and provide the necessary memories to the LLM in the form of linguistic feedback during decision-making (Shinn et al., 2023).”
The Rise and Potential of Large Language Model Based Agents: A Survey
Xi, Z., Chen, W., Guo, X., He, W., Ding, Y., Hong, B., Zhang, M., Wang, J., Jin, S., Zhou, E., et al. (2023) · arXiv · doi:10.48550/arxiv.2309.07864 · Cited by 972 · cited 5×
“Among these, personalized AI agents that simulate human behavior have garnered increasing attention and are considered a crucial pathway toward AGI (Xi et al., 2023).”
InfluentialThe rise and potential of large language model based agents: a survey
Xi, Z., Chen, W., Guo, X., He, W., Ding, Y., Hong, B., Zhang, M., Wang, J., Jin, S., Zhou, E., et al. (2025) · Science China Information Sciences · doi:10.1007/s11432-024-4222-0 · Cited by 970
Multi-agent deep reinforcement learning: a survey
Gronauer, S., & Diepold, K. (2022) · Artificial Intelligence Review · doi:10.1007/s10462-021-09996-w · Cited by 957
Serrano, S., & Smith, N. A. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1282 · Cited by 690 · cited 1×
“Furthermore, drawing upon research in explainable deep learning (Serrano and Smith, 2019), we innovatively decided to determine the output of the model, specifically the retrieved memories, not by the weighted memories vector m weighted , but rather through the model’s cross attention mechanism.”
Bang, Y., Cahyawijaya, S., Lee, N., Dai, W., Su, D., Wilie, B., Lovenia, H., Ji, Z., Yu, T., Chung, W., et al. (2023) · Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.ijcnlp-main.45 · Cited by 682
Harnessing the power of llms in practice: a survey on chatGPT and beyond
Yang, J., Jin, H., Tang, R., Han, X., Feng, Q., Jiang, H., Zhong, S., Yin, B., & Hu, X. (2024) · ACM Transactions on Knowledge Discovery from Data · doi:10.1145/3649506 · Cited by 644
Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond
Yang, J., Jin, H., Tang, R., Han, X., Feng, Q., Jiang, H., Yin, B., & Hu, X. (2023) · arXiv · doi:10.48550/arxiv.2304.13712 · Cited by 644
Human-Level AI's Killer Application: Interactive Computer Games
Laird, J., & van Lent, M. (2000) · The AI Magazine · doi:10.1609/aimag.v22i2.1558 · Cited by 551
Cao, Y., Li, S., Liu, Y., Yan, Z., Dai, Y., Yu, P. S., & Sun, L. (2023) · arXiv · doi:10.48550/arxiv.2303.04226 · Cited by 463
Human-Level AI's Killer Application: Interactive Computer Games
Laird, J. E., & van Lent, M. (2000) · National Conference on Artificial Intelligence · Cited by 459
Can Large Language Models Be an Alternative to Human Evaluations?
Chiang, C.-H., & Lee, H.-Y. (2023) · arXiv · doi:10.48550/arxiv.2305.01937 · Cited by 331
Can Large Language Models Be an Alternative to Human Evaluations?
Chiang, C.-H., & Lee, H.-Y. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.870 · Cited by 331
Is ChatGPT A Good Translator? Yes With GPT-4 As The Engine
Jiao, W., Wang, W., Huang, J.-T., Wang, X., Shi, S., & Tu, Z. (2023) · arXiv · doi:10.48550/arxiv.2301.08745 · Cited by 317
MemoryBank: Enhancing Large Language Models with Long-Term Memory
Zhong, W., Guo, L., Gao, Q., Ye, H., & Wang, Y. (2023) · arXiv · doi:10.48550/arxiv.2305.10250 · Cited by 253
Is chatGPT a good nlg evaluator? A preliminary study
Wang, J., Liang, Y., Meng, F., Sun, Z., Shi, H., Li, Z., Xu, J., Qu, J., & Zhou, J. (2023) · Proceedings of the 4th New Frontiers in Summarization Workshop · doi:10.18653/v1/2023.newsum-1.1 · Cited by 235
Reconsolidation of Human Memory: Brain Mechanisms and Clinical Relevance
Schwabe, L., Nader, K., & Pruessner, J. C. (2014) · Biological Psychiatry · doi:10.1016/j.biopsych.2014.03.008 · Cited by 233
(1995) · Lecture Notes in Computer Science · doi:10.1007/3-540-58855-8 · Cited by 232 · cited 1×
“Typically, AI-based agents are designed to perceive their environments through sensors, make decisions, and perform actions using effectors (Wooldridge and Jennings, 1995; Russell and Norvig, 2016).”
LongT5: Efficient Text-To-Text Transformer for Long Sequences
Guo, M., Ainslie, J., Uthus, D., Ontanon, S., Ni, J., Sung, Y.-H., & Yang, Y. (2022) · Findings of the Association for Computational Linguistics: NAACL 2022 · doi:10.18653/v1/2022.findings-naacl.55 · Cited by 230
LongT5: Efficient Text-To-Text Transformer for Long Sequences
Guo, M., Ainslie, J., Uthus, D., Ontanon, S., Ni, J., Sung, Y.-H., & Yang, Y. (2021) · arXiv · doi:10.48550/arxiv.2112.07916 · Cited by 230
FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design
Yu, Y., Li, H., Zhi, C., Jiang, Y., Li, Y., Zhang, D., Liu, R., Suchow, J. W., & Khashanah, K. (2024) · Proceedings of the AAAI Symposium Series · doi:10.1609/aaaiss.v3i1.31290 · Cited by 227
Social Simulacra: Creating Populated Prototypes for Social Computing Systems
Park, J. S., Popowski, L., Cai, C., Morris, M. R., Liang, P., & Bernstein, M. S. (2022) · Proceedings of the 35th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3526113.3545616 · Cited by 212
PromptChainer: Chaining Large Language Model Prompts through Visual Programming
Wu, T., Jiang, E., Donsbach, A., Gray, J., Molina, A., Terry, M., & Cai, C. J. (2022) · CHI Conference on Human Factors in Computing Systems Extended Abstracts · doi:10.1145/3491101.3519729 · Cited by 198 · cited 1×
“LLM-based agents primarily utilize prompt chains (Wu T. et al., 2022) to generate concise natural language descriptions and actions for characters within prototype systems, thereby creating populated prototypes for social computing systems (Park J. S. et al., 2022).”
Yannakakis, G. N. (2012) · Proceedings of the 9th conference on Computing Frontiers · doi:10.1145/2212908.2212954 · Cited by 179
ExpeL: LLM Agents Are Experiential Learners
Zhao, A., Huang, D., Xu, Q., Lin, M., Liu, Y.-J., & Huang, G. (2023) · arXiv · doi:10.48550/arxiv.2308.10144 · Cited by 179
From local to global: a graph rag approach to query-focused summarization
Edge, D., Trinh, H., Cheng, N., Bradley, J., Chao, A., Mody, A., Truitt, S., Metropolitansky, D., Ness, R. O., & Larson, J. (2024) · arXiv · doi:10.48550/arxiv.2404.16130 · Cited by 159
A Survey on the Memory Mechanism of Large Language Model based Agents
Zhang, Z., Bo, X., Ma, C., Li, R., Chen, X., Dai, Q., Zhu, J., Dong, Z., & Wen, J.-R. (2024) · arXiv · doi:10.48550/arxiv.2404.13501 · Cited by 149 · cited 1×
“A recent survey (Zhang et al., 2024) further highlights the significance of memory modules in enabling LLM-based agents to achieve self-evolving capabilities and interact effectively in real-world contexts.”
Extending cognitive architecture with episodic memory
Nuxoll, A. M., & Laird, J. E. (2007) · National Conference on Artificial Intelligence · Cited by 147
Text and Code Embeddings by Contrastive Pre-Training
Neelakantan, A., Xu, T., Puri, R., Radford, A., Han, J. M., Tworek, J., Yuan, Q., Tezak, N., Kim, J. W., Hallacy, C., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2201.10005 · Cited by 145
Efficient Classification of Long Documents Using Transformers
Park, H., Vyas, Y., & Shah, K. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-short.79 · Cited by 98
Pesce, E., & Montana, G. (2020) · Machine Learning · doi:10.1007/s10994-019-05864-5 · Cited by 97
Wu, Y., Rabe, M. N., Hutchins, D., & Szegedy, C. (2022) · arXiv · doi:10.48550/arxiv.2203.08913 · Cited by 90 · cited 1×
“Memorizing Transformers (Wu Y. et al., 2022) leveraged non-differentiable memory lookup systems to retrieve past inputs during inference, enabling real-time memory retrieval.”
ChatDev: Communicative Agents for Software Development
Qian, C., Liu, W., Liu, H., Chen, N., Dang, Y., Li, J., Yang, C., Chen, W., Su, Y., Cong, X., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.07924 · Cited by 81
Memory Augmented Large Language Models are Computationally UniversalPreprint
Schuurmans, D. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2301.04589 · Cited by 76 · cited 1×
“Research by Schuurmans (2023) demonstrated that transformer-based large language models (LLMs), when augmented with external memory, achieve computational universality.”
Dungeons and Dragons as a Dialog Challenge for Artificial Intelligence
Callison-Burch, C., Tomar, G. S., Martin, L. J., Ippolito, D., Bailis, S., & Reitter, D. (2022) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2022.emnlp-main.637 · Cited by 75
X-Former: In-Memory Acceleration of Transformers
Sridharan, S., Stevens, J. R., Roy, K., & Raghunathan, A. (2023) · IEEE Transactions on Very Large Scale Integration (VLSI) Systems · doi:10.1109/tvlsi.2023.3282046 · Cited by 68 · cited 1×
“Additionally, hardware-related research has explored optimizing memory utilization to improve model efficiency (Sridharan et al., 2023).”
Bulatov, A., Kuratov, Y., & Burtsev, M. S. (2022) · arXiv · doi:10.48550/arxiv.2207.06881 · Cited by 55 · cited 1×
“For instance, Memory Transformer and Recurrent Memory Transformer (Burtsev et al., 2020; Bulatov et al., 2022) introduced memory tokens and recurrent mechanisms to improve transformers’ understanding of long-sequence tasks, especially for global context processing.”
Demonstrate-Search-Predict: Composing retrieval and language models for knowledge-intensive NLP
Khattab, O., Santhanam, K., Li, X. L., Hall, D., Liang, P., Potts, C., & Zaharia, M. (2022) · arXiv · doi:10.48550/arxiv.2212.14024 · Cited by 52
LLM As DBAPreprint
Zhou, X., Li, G., & Liu, Z. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2308.05481 · Cited by 51
SCM: Enhancing Large Language Model with Self-Controlled Memory Framework
Wang, B., Liang, X., Yang, J., Huang, H., Wu, Z., Wu, S., Ma, Z., & Li, Z. (2026) · Lecture Notes in Computer Science · doi:10.1007/978-981-95-4158-4_12 · Cited by 36
Memory Sandbox: Transparent and Interactive Memory Management for Conversational Agents
Huang, Z., Gutierrez, S., Kamana, H., & Macneil, S. (2023) · Adjunct Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology · doi:10.1145/3586182.3615796 · Cited by 34
Hou, Y., Tamoto, H., & Miyashita, H. (2024) · Extended Abstracts of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613905.3650839 · Cited by 31
Procedural Generation of Interactive Stories using Language Models
Freiknecht, J., & Effelsberg, W. (2020) · International Conference on the Foundations of Digital Games · doi:10.1145/3402942.3409599 · Cited by 29
Burtsev, M. S., Kuratov, Y., Peganov, A., & Sapunov, G. V. (2020) · arXiv · doi:10.48550/arxiv.2006.11527 · Cited by 19
Training Socially Aligned Language Models on Simulated Social Interactions
Liu, R., Yang, R., Jia, C., Zhang, G., Zhou, D., Dai, A. M., Yang, D., & Vosoughi, S. (2023) · arXiv · doi:10.48550/arxiv.2305.16960 · Cited by 18 · cited 1×
“Consequently, an increasing number of researchers are exploring the use of LLMs to develop Generative Agents with robust content generation capabilities ( Park et al., 2023; Liu et al., 2023; Wang et al., 2024).”
Proceedings of the 9th conference on Computing Frontiers
Feo, J., Faraboschi, P., & Villa, O. (2012) · doi:10.1145/2212908 · Cited by 16
AgentSims: An Open-Source Sandbox for Large Language Model Evaluation
Lin, J., Zhao, H., Zhang, A., Wu, Y., Ping, H., & Chen, Q. (2023) · arXiv · doi:10.48550/arxiv.2308.04026 · Cited by 14
Agent‐Oriented Software Engineering Methodologies: Analysis and Future Directions
Abdalla, R., & Mishra, A. (2021) · Complexity · doi:10.1155/2021/1629419 · Cited by 13
ChatDB: Augmenting LLMs with Databases as Their Symbolic Memory
Hu, C., Fu, J., Du, C., Luo, S., Zhao, J., & Zhao, H. (2023) · arXiv · doi:10.48550/arxiv.2306.03901 · Cited by 13
RET-LLM: Towards a General Read-Write Memory for Large Language Models
Modarressi, A., Imani, A., Fayyaz, M., & Schütze, H. (2023) · arXiv · doi:10.48550/arxiv.2305.14322 · Cited by 8
RAP: Retrieval-Augmented Planning with Contextual Memory for Multimodal LLM Agents
Kagaya, T., Yuan, T. J., Lou, Y., Karlekar, J., Pranata, S., Kinose, A., Oguri, K., Wick, F., & You, Y. (2024) · arXiv · doi:10.48550/arxiv.2402.03610 · Cited by 8
An agent-based model of opinion polarization driven by emotionsPreprint
Schweitzer, F., Kriváchy, T., & García, D. (2020) · SocArXiv (OSF Preprints) · doi:10.31235/osf.io/8m2wq · Cited by 6 · cited 1×
“Despite this, the development of AI agents that can accurately and credibly simulate complex human behaviors has proven to be challenging (Schweitzer et al., 2020; Abdalla and Mishra, 2021).”
FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design
Yu, Y., Li, H., Chen, Z., Jiang, Y., Li, Y., Zhang, D., Liu, R., Suchow, J. W., & Khashanah, K. (2023) · arXiv · doi:10.48550/arxiv.2311.13743 · Cited by 6
SCM: Enhancing Large Language Model with Self-Controlled Memory Framework
Wang, B., Liang, X., Yang, J., Huang, H., Wu, S., Wu, P., Lu, L., Ma, Z., & Li, Z. (2023) · arXiv · doi:10.48550/arxiv.2304.13343 · Cited by 5
Landmark Attention: Random-Access Infinite Context Length for Transformers
Mohtashami, A., & Jaggi, M. (2023) · arXiv · doi:10.48550/arxiv.2305.16300 · Cited by 4
Toward conversational agents with context and time sensitive long-term memory
Alonso, N., Figliolia, T., Ndirango, A., & Millidge, B. (2024) · arXiv · doi:10.48550/arxiv.2406.00057 · Cited by 2
OpenAI, Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F. L., Almeida, D., Altenschmidt, J., Altman, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.08774
Serrano, S., & Smith, N. A. (2019) · arXiv · doi:10.48550/arxiv.1906.03731
Research works printed but not linked: 2
- Musacchio, 2024, Adapting large language models to narrative content, Complexity, 6, 8
- Zalta, 1995, Stanford Encyclopedia of Philosophy
Cited by
From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanisms
Luo, J., Tian, Y., Cao, C., Luo, Z., Lin, H., Li, K., Kong, C., Yang, R., & Ma, J. (2026) · Findings of the Association for Computational Linguistics: ACL 2026 · doi:10.18653/v1/2026.findings-acl.2069 · Cited by 3
Zhang, M., Yang, Y., Yu, C., & Diao, Y. (2025) · Frontiers in Psychology · doi:10.3389/fpsyg.2025.1615342 · Cited by 4
Lightweight LLM Agent Memory with Small Language Models
Zhang, J., Zhang, C., Chen, S., Huang, Z., Zheng, P., Wang, Z., Guo, P., Mo, F., Bae, S.-H., Zou, J., et al. (2026) · Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2026.acl-long.588 · Cited by 35
Memory in LLM-Based Multi-agent Systems: Mechanisms, Challenges, and Collective Intelligence
Wu, S., & Shu, K. (2026) · Lecture Notes in Computer Science · doi:10.1007/978-981-92-1468-6_10 · Cited by 2
Dishop, C. R. (2026) · Computers in Human Behavior · doi:10.1016/j.chb.2026.108994 · Cited by 2
Metadata
- Authors
- Hong, C., & He, Q.
- Affiliations
- Southwestern University of Finance and Economics (CN)
- Venue
- Frontiers in Psychology
- Publisher
- Frontiers Media SA
- Published
- 2025-05-07
- Volume
- 16
- Pages
- 1591618-1591618
- ISSN
- 1664-1078 (ISSN-L)
- PMC
- PMC12092450
- OpenAlex
- W4410182127
- Cited by
- 7
- References
- 61
- Citations by year
- 2026: 4 · 2025: 3
- FWCI
- 2.80
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Psychology (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Psychiatry, Mental Health, Neuroscience
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Keywords
- generative AI agents, memory retrieval, cross-attention network, large language models, artificial general intelligence, loss function design, agent-based modeling, behavioral consistency
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.