pith. sign in

arxiv: 2605.25680 · v1 · pith:TBGTC66Knew · submitted 2026-05-25 · 💻 cs.CL · cs.AI

Simulating Human Memory with Language Models

Qihan Wang , Nicholas Tomlin , Michael Hu , Brian Dillon , Tal Linzen This is my paper

Pith reviewed 2026-06-29 21:58 UTC · model grok-4.3

classification 💻 cs.CL cs.AI
keywords language modelshuman memoryuser simulatorsmemory experimentsforgettingeducation taskpsychology benchmarks
0
0 comments X

The pith

Language models remember information more reliably than humans across classic psychology tasks, even when prompted to imitate human behavior.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper compares language models and human participants on standard memory experiments from psychology. It finds that unmodified models retain more details than people do. Introducing specific prompting methods and a compactor allows the models to forget in patterns closer to human performance. Models adjusted this way then act as more effective simulators of real users during an education interaction task. The authors also release human data and benchmarks to enable further work on memory simulation.

Core claim

Out-of-the-box language models exhibit better memory than humans across a series of classic psychology experiments, even when prompted to imitate human behavior. Better prompting strategies and the use of a compactor can cause language models to forget content in a more human-like way. Language models with these human-like memory constraints function as more effective user simulators in a downstream education task.

What carries the argument

A compactor together with prompting strategies that induce human-like forgetting in language models.

If this is right

  • Out-of-the-box language models retain more information than humans in memory tasks.
  • Prompting strategies plus a compactor produce forgetting patterns that match human data more closely.
  • Language models with human-like memory constraints improve performance as user simulators in education tasks.
  • New human reference data and benchmarks are now available for testing memory simulation methods.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same adjustment methods could be tested in other interactive domains such as customer support or collaborative problem solving.
  • If the memory constraints scale, they might allow simulators to replace some human-subject studies in interface design.
  • The gap in raw memory reliability may explain why current simulators sometimes produce unrealistically consistent user responses.
  • Future work could measure whether the compactor affects other cognitive aspects beyond memory alone.

Load-bearing premise

The classic memory experiments from psychology serve as valid proxies for the memory demands that arise when language models act as user simulators in interactive tasks such as education.

What would settle it

A direct comparison in which language models with induced forgetting perform no better than unmodified models when simulating users in actual educational dialogues would falsify the claim that human-like memory constraints improve simulator effectiveness.

Figures

Figures reproduced from arXiv: 2605.25680 by Brian Dillon, Michael Hu, Nicholas Tomlin, Qihan Wang, Tal Linzen.

Figure 1
Figure 1. Figure 1: Three sample tasks from our benchmark. In reverse digit span, participants must read and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Human–model performance comparisons across tasks and for two representative models, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Human–model similarity across models and prompt conditions. Error bars indicate [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Few-shot prompting experiments. Human–model similarity across prompting conditions [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Proof-of-concept application: rereanking educational documents. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The main instruction and example user interfaces from our experimental platform. Top: [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Ablation results for the COMPACTOR. We find that COMPACTOR, which uses a key-value memory store, achieves higher humanlikeness than a simpler method which prompts a model to produce an abstractive summary of its context before completing the task. • Human prompt: The human will have three minutes to read a passage, after which the text will disappear. The human will then be asked to answer ten questions ab… view at source ↗
Figure 8
Figure 8. Figure 8: Although Llama 3 8B with COMPACTOR achieves higher pairwise accuracy across condi￾tions, the relative ordering of document difficulty still differs from human behavior, suggesting there remains room for improvement. 25 [PITH_FULL_IMAGE:figures/full_fig_p025_8.png] view at source ↗
read the original abstract

Language models are increasingly being deployed as user simulators, but their memory is far more reliable than that of real users. To measure this gap, we run a series of classic memory experiments from psychology on both humans and language models. Across tasks, we find that out-of-the-box language models exhibit better memory than humans, even when prompted to imitate human behavior. We then show that better prompting strategies and the use of a compactor can cause language models to forget content in a more human-like way. Using these methods, we show preliminary evidence that language models with human-like memory constraints can function as more effective user simulators in a downstream education task. Finally, we release human reference data and benchmarks to support future work on simulating human memory with language models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper claims that out-of-the-box language models exhibit superior memory performance compared to humans across classic psychology memory experiments, even when prompted to imitate human behavior. It further claims that improved prompting and a 'compactor' mechanism can induce more human-like forgetting patterns in LMs, and provides preliminary evidence that LMs constrained in this way serve as more effective user simulators in a downstream education task. Human reference data and benchmarks are released to support future work.

Significance. If the empirical comparisons and downstream results hold after proper statistical validation and proxy testing, the work could support development of more realistic LM-based user simulators for interactive applications such as education by addressing the reliability gap in memory. The release of human data and benchmarks is a constructive contribution for reproducibility in this area.

major comments (3)
  1. [Abstract] Abstract: the abstract states clear empirical findings but provides no sample sizes, statistical tests, error bars, or details on how the compactor is implemented or evaluated; the central claim cannot be assessed from the given text alone.
  2. [Downstream education task] Downstream education task: the claim that language models with human-like memory constraints function as more effective user simulators rests on the untested assumption that forgetting patterns from classic one-shot recall or list-learning experiments transfer to memory failures in multi-turn interactive tutoring (context tracking, personalization, error recovery); no evidence is provided that the specific human-like forgetting (vs. any capacity reduction) drives the reported gain.
  3. [Methods] Methods/implementation: details on the compactor's architecture, training or prompting procedure, and quantitative evaluation against human forgetting curves are absent, which are load-bearing for reproducing the human-like forgetting result and the education-task improvement.
minor comments (1)
  1. [Abstract] Abstract: the term 'compactor' is introduced without a brief definition or citation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback. We address each major comment below and indicate the planned revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the abstract states clear empirical findings but provides no sample sizes, statistical tests, error bars, or details on how the compactor is implemented or evaluated; the central claim cannot be assessed from the given text alone.

    Authors: We agree that the abstract would benefit from these specifics. In revision we will expand the abstract to report sample sizes for the human and LM experiments, reference the statistical tests performed, note error bars in the relevant figures, and include a concise description of the compactor and its evaluation. revision: yes

  2. Referee: [Downstream education task] Downstream education task: the claim that language models with human-like memory constraints function as more effective user simulators rests on the untested assumption that forgetting patterns from classic one-shot recall or list-learning experiments transfer to memory failures in multi-turn interactive tutoring (context tracking, personalization, error recovery); no evidence is provided that the specific human-like forgetting (vs. any capacity reduction) drives the reported gain.

    Authors: We acknowledge the point. The manuscript presents preliminary evidence that the constrained models improve simulator performance on the education task, yet it does not contain controls isolating the contribution of the specific human-like forgetting curve versus general capacity reduction. We will revise the discussion to explicitly state this limitation and its implications for causal interpretation. revision: partial

  3. Referee: [Methods] Methods/implementation: details on the compactor's architecture, training or prompting procedure, and quantitative evaluation against human forgetting curves are absent, which are load-bearing for reproducing the human-like forgetting result and the education-task improvement.

    Authors: We agree these details are necessary. The revised manuscript will expand the Methods section to describe the compactor architecture, the prompting and compaction procedures, and the quantitative metrics used to compare induced forgetting curves against the released human benchmarks. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical comparison with no derivations or self-referential fits

full rationale

The paper reports direct experimental results from running classic psychology memory tasks on humans and LMs, followed by empirical tests of prompting/compactor methods and a downstream education task. No equations, parameter fits, uniqueness theorems, or self-citations are invoked as load-bearing steps in any derivation chain. All reported outcomes (better LM memory, human-like forgetting via interventions, improved simulator performance) are measured quantities, not quantities defined by the authors' own modeling choices. This is the standard case of a self-contained empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The work is empirical and relies on the validity of transferring psychology memory tasks to the simulator setting; no free parameters or invented physical entities are described in the abstract.

axioms (1)
  • domain assumption Classic psychology memory experiments measure the memory properties relevant to language-model user simulation in downstream tasks.
    This premise is required for the claim that human-like memory improves simulator effectiveness; it is invoked when the authors move from the memory experiments to the education-task result.
invented entities (1)
  • compactor no independent evidence
    purpose: Component that forces language models to forget content in a more human-like manner.
    Introduced as a new method to adjust model memory; no independent evidence outside the paper is described in the abstract.

pith-pipeline@v0.9.1-grok · 5652 in / 1337 out tokens · 22311 ms · 2026-06-29T21:58:10.843396+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

87 extracted references · 35 canonical work pages · 18 internal anchors

  1. [1]

    A systematic comparison of syllogistic reasoning in humans and language models

    Tiwalayo Eisape, Michael Tessler, Ishita Dasgupta, Fei Sha, Sjoerd Steenkiste, and Tal Linzen. A systematic comparison of syllogistic reasoning in humans and language models. In Kevin Duh, Helena Gomez, and Steven Bethard, editors,Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langu...

    2024

  2. [2]

    User simulators bridge RL with real-world interaction

    Jessy Lin and Nicholas Tomlin. User simulators bridge RL with real-world interaction. https: //jessylin.com/2025/07/10/user-simulators-1/

    2025

  3. [3]

    TutorUp: What if your students were simulated? Training tutors to address engagement challenges in online learning

    Sitong Pan, Robin Schmucker, Bernardo Garcia Bulle Bueno, Salome Aguilar Llanes, Fernanda Albo Alarcón, Hangxiao Zhu, Adam Teo, and Meng Xia. TutorUp: What if your students were simulated? Training tutors to address engagement challenges in online learning. InProceedings of the 2025 CHI conference on human factors in computing systems, pages 1–18, 2025

    2025

  4. [4]

    Can LLM-simulated practice and feedback upskill human counselors? A randomized study with 90+ novice counselors

    Ryan Louie, Raj Sanjay Shah, Ifdita Hasan Orney, Juan Pablo Pacheco, Emma Brunskill, and Diyi Yang. Can LLM-simulated practice and feedback upskill human counselors? A randomized study with 90+ novice counselors. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems, pages 1–31, 2026

    2026

  5. [5]

    LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals

    Joon Sung Park, Carolyn Q Zou, Aaron Shaw, Benjamin Mako Hill, Carrie Cai, Meredith Ringel Morris, Robb Willer, Percy Liang, and Michael S Bernstein. Generative agent simulations of 1,000 people.arXiv preprint arXiv:2411.10109, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  6. [6]

    Building machines that learn and think like people.Behavioral and brain sciences, 40:e253, 2017

    Brenden M Lake, Tomer D Ullman, Joshua B Tenenbaum, and Samuel J Gershman. Building machines that learn and think like people.Behavioral and brain sciences, 40:e253, 2017

    2017

  7. [7]

    To model human linguistic prediction, make LLMs less superhuman

    Byung-Doh Oh and Tal Linzen. To model human linguistic prediction, make LLMs less superhuman.arXiv preprint arXiv:2510.05141, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  8. [8]

    Characterizing verbatim short-term memory in neural language models

    Kristijan Armeni, Christopher Honey, and Tal Linzen. Characterizing verbatim short-term memory in neural language models. InProceedings of the 26th Conference on Computational Natural Language Learning (CoNLL), pages 405–424, 2022

    2022

  9. [9]

    Benchmarks for models of short-term and working memory.Psychological Bulletin, 144(9):885, 2018

    Klaus Oberauer, Stephan Lewandowsky, Edward Awh, Gordon DA Brown, Andrew Conway, Nelson Cowan, Christopher Donkin, Simon Farrell, Graham J Hitch, Mark J Hurlstone, et al. Benchmarks for models of short-term and working memory.Psychological Bulletin, 144(9):885, 2018

    2018

  10. [10]

    The magical number 4 in short-term memory: A reconsideration of mental storage capacity.Behavioral and Brain Sciences, 24(1):87–114, 2001

    Nelson Cowan. The magical number 4 in short-term memory: A reconsideration of mental storage capacity.Behavioral and Brain Sciences, 24(1):87–114, 2001. 10

    2001

  11. [11]

    Analyzing memory effects in large language models through the lens of cognitive psychology.arXiv preprint arXiv:2509.17138, 2025

    Zhaoyang Cao, Lael Schooler, and Reza Zafarani. Analyzing memory effects in large language models through the lens of cognitive psychology.arXiv preprint arXiv:2509.17138, 2025

    work page arXiv 2025

  12. [12]

    Are frequent phrases directly retrieved like idioms? An investigation with self-paced reading and language models

    Giulia Rambelli, Emmanuele Chersoni, Marco SG Senaldi, Philippe Blache, and Alessandro Lenci. Are frequent phrases directly retrieved like idioms? An investigation with self-paced reading and language models. InProceedings of the 19th Workshop on Multiword Expressions (MWE 2023), pages 87–98, 2023

    2023

  13. [13]

    R Thomas McCoy, Paul Smolensky, Tal Linzen, Jianfeng Gao, and Asli Celikyilmaz. How much do language models copy from their training data? evaluating linguistic novelty in text generation using raven.Transactions of the Association for Computational Linguistics, 11:652–670, 2023

    2023

  14. [14]

    Humans and language models diverge when predicting repeating text

    Aditya Vaidya, Javier Turek, and Alexander Huth. Humans and language models diverge when predicting repeating text. InProceedings of the 27th Conference on Computational Natural Language Learning (CoNLL), pages 58–69, 2023

    2023

  15. [15]

    Bigger is not always better: The importance of human-scale language modeling for psycholinguistics.Journal of Memory and Language, 144:104650, 2025

    Ethan Gotlieb Wilcox, Michael Y Hu, Aaron Mueller, Alex Warstadt, Leshem Choshen, Chengxu Zhuang, Adina Williams, Ryan Cotterell, and Tal Linzen. Bigger is not always better: The importance of human-scale language modeling for psycholinguistics.Journal of Memory and Language, 144:104650, 2025

    2025

  16. [16]

    Human-like fleeting memory improves language learning but impairs reading time prediction in transformer language models

    Abishek Thamma and Micha Heilbron. Human-like fleeting memory improves language learning but impairs reading time prediction in transformer language models.arXiv preprint arXiv:2508.05803, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  17. [17]

    Locally biased transformers better align with human reading times

    Andrea De Varda and Marco Marelli. Locally biased transformers better align with human reading times. InProceedings of the workshop on cognitive modeling and computational linguistics, pages 30–36, 2024

    2024

  18. [18]

    Linear recency bias during training improves transformers’ fit to reading times

    Christian Clark, Byung-Doh Oh, and William Schuler. Linear recency bias during training improves transformers’ fit to reading times. InProceedings of the 31st International Conference on Computational Linguistics, pages 7735–7747, 2025

    2025

  19. [19]

    A foundation model to predict and capture human cognition.Nature, 644(8078):1002–1009, 2025

    Marcel Binz, Elif Akata, Matthias Bethge, Franziska Brändle, Fred Callaway, Julian Coda- Forno, Peter Dayan, Can Demircan, Maria K Eckstein, Noémi Éltet˝o, et al. A foundation model to predict and capture human cognition.Nature, 644(8078):1002–1009, 2025

    2025

  20. [20]

    HippoRAG: Neurobio- logically inspired long-term memory for large language models.Advances in neural information processing systems, 37:59532–59569, 2024

    Bernal J Gutiérrez, Yiheng Shu, Yu Gu, Michihiro Yasunaga, and Yu Su. HippoRAG: Neurobio- logically inspired long-term memory for large language models.Advances in neural information processing systems, 37:59532–59569, 2024

    2024

  21. [21]

    arXiv preprint arXiv:2407.09450 (2024)

    Zafeirios Fountas, Martin A Benfeghoul, Adnan Oomerjee, Fenia Christopoulou, Gerasimos Lampouras, Haitham Bou-Ammar, and Jun Wang. Human-inspired episodic memory for infinite context LLMs.arXiv preprint arXiv:2407.09450, 2024

    work page arXiv 2024

  22. [22]

    Towards large language models with human-like episodic memory.Trends in Cognitive Sciences, 2025

    Cody V Dong, Qihong Lu, Kenneth A Norman, and Sebastian Michelmann. Towards large language models with human-like episodic memory.Trends in Cognitive Sciences, 2025

    2025

  23. [23]

    Sangyeop Kim, Yohan Lee, Sanghwa Kim, Hyunjong Kim, and Sungzoon Cho

    Kuang-Huei Lee, Xinyun Chen, Hiroki Furuta, John Canny, and Ian Fischer. A human-inspired reading agent with gist memory of very long contexts.arXiv preprint arXiv:2402.09727, 2024

    work page arXiv 2024

  24. [24]

    Himes: Hippocampus-inspired memory system for personalized AI assistants.arXiv preprint arXiv:2601.06152, 2026

    Hailong Li, Feifei Li, Wenhui Que, and Xingyu Fan. Himes: Hippocampus-inspired memory system for personalized AI assistants.arXiv preprint arXiv:2601.06152, 2026

    work page arXiv 2026

  25. [25]

    Transformer-XL: Attentive language models beyond a fixed-length context

    Zihang Dai, Zhilin Yang, Yiming Yang, Jaime G Carbonell, Quoc Le, and Ruslan Salakhutdinov. Transformer-XL: Attentive language models beyond a fixed-length context. InProceedings of the 57th annual meeting of the association for computational linguistics, pages 2978–2988, 2019

    2019

  26. [26]

    Recurrent memory transformer.Advances in Neural Information Processing Systems, 35:11079–11091, 2022

    Aydar Bulatov, Yury Kuratov, and Mikhail Burtsev. Recurrent memory transformer.Advances in Neural Information Processing Systems, 35:11079–11091, 2022. 11

    2022

  27. [27]

    MemoryLLM: Towards self-updatable large language models.arXiv preprint arXiv:2402.04624, 2024

    Yu Wang, Yifan Gao, Xiusi Chen, Haoming Jiang, Shiyang Li, Jingfeng Yang, Qingyu Yin, Zheng Li, Xian Li, Bing Yin, et al. MemoryLLM: Towards self-updatable large language models.arXiv preprint arXiv:2402.04624, 2024

    work page arXiv 2024

  28. [28]

    M+: Extending MemoryLLM with scalable long-term memory.arXiv preprint arXiv:2502.00592, 2025

    Yu Wang, Dmitry Krotov, Yuanzhe Hu, Yifan Gao, Wangchunshu Zhou, Julian McAuley, Dan Gutfreund, Rogerio Feris, and Zexue He. M+: Extending MemoryLLM with scalable long-term memory.arXiv preprint arXiv:2502.00592, 2025

    work page arXiv 2025

  29. [29]

    Lost in the middle: How language models use long contexts.Transactions of the association for computational linguistics, 12:157–173, 2024

    Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. Lost in the middle: How language models use long contexts.Transactions of the association for computational linguistics, 12:157–173, 2024

    2024

  30. [30]

    YaRN: Efficient Context Window Extension of Large Language Models

    Bowen Peng, Jeffrey Quesnelle, Honglu Fan, and Enrico Shippole. YaRN: Efficient context window extension of large language models.arXiv preprint arXiv:2309.00071, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  31. [31]

    A controlled study on long context extension and generalization in LLMs, 2025

    Yi Lu, Jing Nathan Yan, Songlin Yang, Justin T Chiu, Siyu Ren, Fei Yuan, Wenting Zhao, Zhiyong Wu, and Alexander M Rush. A controlled study on long context extension and generalization in LLMs, 2025

    2025

  32. [32]

    Extending Context Window of Large Language Models via Positional Interpolation

    Shouyuan Chen, Sherman Wong, Liangjian Chen, and Yuandong Tian. Extending context window of large language models via positional interpolation.arXiv preprint arXiv:2306.15595, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  33. [33]

    Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    2024

  34. [34]

    Self- updatable large language models by integrating context into model parameters.arXiv preprint arXiv:2410.00487, 2024

    Yu Wang, Xinshuang Liu, Xiusi Chen, Sean O’Brien, Junda Wu, and Julian McAuley. Self- updatable large language models by integrating context into model parameters.arXiv preprint arXiv:2410.00487, 2024

    work page arXiv 2024

  35. [35]

    Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

    Xin Cheng, Wangding Zeng, Damai Dai, Qinyu Chen, Bingxuan Wang, Zhenda Xie, Kezhao Huang, Xingkai Yu, Zhewen Hao, Yukun Li, et al. Conditional memory via scalable lookup: A new axis of sparsity for large language models.arXiv preprint arXiv:2601.07372, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  36. [36]

    Continual learning via sparse memory finetuning, 2025

    Jessy Lin, Luke Zettlemoyer, Gargi Ghosh, Wen-Tau Yih, Aram Markosyan, Vincent-Pierre Berges, and Barlas O˘guz. Continual learning via sparse memory finetuning, 2025

    2025

  37. [37]

    Hierarchical memory for high-efficiency long-term reasoning in LLM agents.arXiv preprint arXiv:2507.22925, 2025

    Haoran Sun and Shaoning Zeng. Hierarchical memory for high-efficiency long-term reasoning in LLM agents.arXiv preprint arXiv:2507.22925, 2025

    work page arXiv 2025

  38. [38]

    A-MEM: Agentic Memory for LLM Agents

    Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. A-MEM: Agentic memory for LLM agents.arXiv preprint arXiv:2502.12110, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  39. [39]

    Zep: A Temporal Knowledge Graph Architecture for Agent Memory

    Preston Rasmussen, Pavlo Paliychuk, Travis Beauvais, Jack Ryan, and Daniel Chalef. Zep: A temporal knowledge graph architecture for agent memory.arXiv preprint arXiv:2501.13956, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  40. [40]

    Mensink, Andrew Elfenbein, and Dongyeop Kang

    Karin de Langis, Jong Inn Park, Bin Hu, Khanh Chi Le, Andreas Schramm, Michael C. Mensink, Andrew Elfenbein, and Dongyeop Kang. Strong memory, weak control: An empirical study of executive functioning in LLMs. In Vera Demberg, Kentaro Inui, and Lluís Marquez, editors,Proceedings of the 19th Conference of the European Chapter of the Association for Computa...

  41. [41]

    Association for Computational Linguistics

  42. [42]

    Evaluating very long-term conversational memory of LLM agents

    Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. Evaluating very long-term conversational memory of LLM agents. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13851–13870, 2024

    2024

  43. [43]

    LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory

    Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, Kai-Wei Chang, and Dong Yu. Long- MemEval: Benchmarking chat assistants on long-term interactive memory.arXiv preprint arXiv:2410.10813, 2024. 12

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  44. [44]

    Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions

    Yuanzhe Hu, Yu Wang, and Julian McAuley. Evaluating memory in LLM agents via incremental multi-turn interactions.arXiv preprint arXiv:2507.05257, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  45. [45]

    MEMTRACK: Evaluating long-term memory and state tracking in multi-platform dynamic agent environments.arXiv preprint arXiv:2510.01353, 2025

    Darshan Deshpande, Varun Gangal, Hersh Mehta, Anand Kannappan, Rebecca Qian, and Peng Wang. MEMTRACK: Evaluating long-term memory and state tracking in multi-platform dynamic agent environments.arXiv preprint arXiv:2510.01353, 2025

    work page arXiv 2025

  46. [46]

    Generative agents: Interactive simulacra of human behavior

    Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. Generative agents: Interactive simulacra of human behavior. InProceed- ings of the 36th annual acm symposium on user interface software and technology, pages 1–22, 2023

    2023

  47. [47]

    Scaling Synthetic Data Creation with 1,000,000,000 Personas

    Tao Ge, Xin Chan, Xiaoyang Wang, Dian Yu, Haitao Mi, and Dong Yu. Scaling synthetic data creation with 1,000,000,000 personas.arXiv preprint arXiv:2406.20094, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  48. [48]

    The need for a socially-grounded persona framework for user simulation.arXiv preprint arXiv:2601.07110, 2026

    Pranav Narayanan Venkit, Yu Li, Yada Pruksachatkun, and Chien-Sheng Wu. The need for a socially-grounded persona framework for user simulation.arXiv preprint arXiv:2601.07110, 2026

    work page arXiv 2026

  49. [49]

    LLM personas as a substitute for field experiments in method benchmarking.arXiv preprint arXiv:2512.21080, 2025

    Enoch Hyunwook Kang. LLM personas as a substitute for field experiments in method benchmarking.arXiv preprint arXiv:2512.21080, 2025

    work page arXiv 2025

  50. [50]

    User behavior simulation with large language model-based agents.ACM Transactions on Information Systems, 43(2):1–37, 2025

    Lei Wang, Jingsen Zhang, Hao Yang, Zhi-Yuan Chen, Jiakai Tang, Zeyu Zhang, Xu Chen, Yankai Lin, Hao Sun, Ruihua Song, et al. User behavior simulation with large language model-based agents.ACM Transactions on Information Systems, 43(2):1–37, 2025

    2025

  51. [51]

    S$^3$: Social-network Simulation System with Large Language Model-Empowered Agents

    Chen Gao, Xiaochong Lan, Zhihong Lu, Jinzhu Mao, Jinghua Piao, Huandong Wang, Depeng Jin, and Yong Li. S3: Social-network simulation system with large language model-empowered agents.arXiv preprint arXiv:2307.14984, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  52. [52]

    AgentSociety: Large-scale simulation of LLM-driven generative agents advances understanding of human behaviors and society

    Jinghua Piao, Yuwei Yan, Jun Zhang, Nian Li, Junbo Yan, Xiaochong Lan, Zhihong Lu, Zhiheng Zheng, Jing Yi Wang, Di Zhou, et al. AgentSociety: Large-scale simulation of LLM-driven generative agents advances understanding of human behaviors and society. 2025

    2025

  53. [53]

    OASIS: Open agent social interaction simulations with one million agents.arXiv preprint arXiv:2411.11581, 2024

    Ziyi Yang, Zaibin Zhang, Zirui Zheng, Yuxian Jiang, Ziyue Gan, Zhiyu Wang, Zijian Ling, Jinsong Chen, Martz Ma, Bowen Dong, et al. OASIS: Open agent social interaction simulations with one million agents.arXiv preprint arXiv:2411.11581, 2024

    work page arXiv 2024

  54. [54]

    Mosaic: Modeling social AI for content dissemination and regulation in multi-agent simulations

    Genglin Liu, Vivian T Le, Salman Rahman, Elisa Kreiss, Marzyeh Ghassemi, and Saadia Gabriel. Mosaic: Modeling social AI for content dissemination and regulation in multi-agent simulations. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 6401–6428, 2025

    2025

  55. [55]

    The automated but risky game: Modeling and benchmarking agent-to-agent negotiations and transactions in consumer markets.arXiv preprint arXiv:2506.00073, 2025

    Shenzhe Zhu, Jiao Sun, Yi Nian, Tobin South, Alex Pentland, and Jiaxin Pei. The automated but risky game: Modeling and benchmarking agent-to-agent negotiations and transactions in consumer markets.arXiv preprint arXiv:2506.00073, 2025

    work page arXiv 2025

  56. [56]

    Decision-oriented dialogue for human-AI collaboration.Transactions of the Association for Computational Linguistics, 12:892–911, 08 2024

    Jessy Lin, Nicholas Tomlin, Jacob Andreas, and Jason Eisner. Decision-oriented dialogue for human-AI collaboration.Transactions of the Association for Computational Linguistics, 12:892–911, 08 2024

    2024

  57. [57]

    $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment

    Victor Barres, Honghua Dong, Soham Ray, Xujie Si, and Karthik Narasimhan. τ 2-Bench: Eval- uating conversational agents in a dual-control environment.arXiv preprint arXiv:2506.07982, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  58. [58]

    Training proactive and personalized LLM agents.arXiv preprint arXiv:2511.02208, 2025

    Weiwei Sun, Xuhui Zhou, Weihua Du, Xingyao Wang, Sean Welleck, Graham Neubig, Maarten Sap, and Yiming Yang. Training proactive and personalized LLM agents.arXiv preprint arXiv:2511.02208, 2025

    work page arXiv 2025

  59. [59]

    Mind the sim2real gap in user simulation for agentic tasks.arXiv preprint arXiv:2603.11245, 2026

    Xuhui Zhou, Weiwei Sun, Qianou Ma, Yiqing Xie, Jiarui Liu, Weihua Du, Sean Welleck, Yiming Yang, Graham Neubig, Sherry Tongshuang Wu, et al. Mind the sim2real gap in user simulation for agentic tasks.arXiv preprint arXiv:2603.11245, 2026. 13

    work page arXiv 2026

  60. [60]

    SimBench: Benchmarking the Ability of Large Language Models to Simulate Human Behaviors

    Tiancheng Hu, Joachim Baumann, Lorenzo Lupo, Nigel Collier, Dirk Hovy, and Paul Röttger. SimBench: Benchmarking the ability of large language models to simulate human behaviors. arXiv preprint arXiv:2510.17516, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  61. [61]

    Do LLMs exhibit human-like response biases? a case study in survey design.Transactions of the Association for Computational Linguistics, 12:1011–1026, 2024

    Lindia Tjuatja, Valerie Chen, Tongshuang Wu, Ameet Talwalkwar, and Graham Neubig. Do LLMs exhibit human-like response biases? a case study in survey design.Transactions of the Association for Computational Linguistics, 12:1011–1026, 2024

    2024

  62. [62]

    Can LLM Agents Simulate Multi-Turn Human Behavior? Evidence from Real Online Customer Behavior Data

    Yuxuan Lu, Jing Huang, Yan Han, Bingsheng Yao, Sisong Bei, Jiri Gesi, Yaochen Xie, Qi He, Dakuo Wang, et al. Can LLM agents simulate multi-turn human behavior? Evidence from real online customer behavior data.arXiv preprint arXiv:2503.20749, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  63. [63]

    LLM-based human simulations have not yet been reliable.arXiv preprint arXiv:2501.08579, 2025

    Qian Wang, Jiaying Wu, Zichen Jiang, Zhenheng Tang, Bingqiao Luo, Nuo Chen, Wei Chen, and Bingsheng He. LLM-based human simulations have not yet been reliable.arXiv preprint arXiv:2501.08579, 2025

    work page arXiv 2025

  64. [64]

    The magical number seven, plus or minus two: Some limits on our capacity for processing information.Psychological review, 63(2):81, 1956

    George A Miller. The magical number seven, plus or minus two: Some limits on our capacity for processing information.Psychological review, 63(2):81, 1956

    1956

  65. [65]

    Effect of age on forward and backward digit spans.Aging, neuropsychology, and cognition, 4(2):140–149, 1997

    Jacques Grégoire and Martial Van Der Linden. Effect of age on forward and backward digit spans.Aging, neuropsychology, and cognition, 4(2):140–149, 1997

    1997

  66. [66]

    Digit span is (mostly) related linearly to general intelligence: Every extra bit of span counts.Psychological Assessment, 27(4):1312, 2015

    Gilles E Gignac and Lawrence G Weiss. Digit span is (mostly) related linearly to general intelligence: Every extra bit of span counts.Psychological Assessment, 27(4):1312, 2015

    2015

  67. [67]

    Development of the WAIS-III: A brief overview, history, and description

    Marc A Silva. Development of the WAIS-III: A brief overview, history, and description. Graduate Journal of Counseling Psychology, 1(1):11, 2008

    2008

  68. [68]

    The digit span backwards task.European Journal of Psychological Assessment, 2014

    Sven Hilbert, Tristan T Nakagawa, Patricia Puci, Alexandra Zech, and Markus Bühner. The digit span backwards task.European Journal of Psychological Assessment, 2014

    2014

  69. [69]

    Reporting and interpreting working memory performance in n-back tasks

    Adrian Meule. Reporting and interpreting working memory performance in n-back tasks. Frontiers in psychology, 8:352, 2017

    2017

  70. [70]

    N-back working memory paradigm: A meta-analysis of normative functional neuroimaging studies.Human brain mapping, 25(1):46–59, 2005

    Adrian M Owen, Kathryn M McMillan, Angela R Laird, and Ed Bullmore. N-back working memory paradigm: A meta-analysis of normative functional neuroimaging studies.Human brain mapping, 25(1):46–59, 2005

    2005

  71. [71]

    Complex span versus updating tasks of working memory: The gap is not that deep

    Florian Schmiedek, Andrea Hildebrandt, Martin Lövdén, Oliver Wilhelm, and Ulman Linden- berger. Complex span versus updating tasks of working memory: The gap is not that deep. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4):1089, 2009

    2009

  72. [72]

    The components of working memory updating: An experimental decomposition and individual differences

    Ullrich KH Ecker, Stephan Lewandowsky, Klaus Oberauer, and Abby EH Chee. The components of working memory updating: An experimental decomposition and individual differences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(1):170, 2010

    2010

  73. [73]

    Retrieval processes in continuous recognition.Journal of Experimental Psychology: Learning, Memory, and Cognition, 8(6):497, 1982

    William E Hockley. Retrieval processes in continuous recognition.Journal of Experimental Psychology: Learning, Memory, and Cognition, 8(6):497, 1982

    1982

  74. [74]

    Naturalistic Free Recall

    Omri Raccah, Phoebe Chen, Todd M Gureckis, David Poeppel, and Vy A V o. The “Naturalistic Free Recall” dataset: four stories, hundreds of participants, and high-fidelity transcriptions. Scientific Data, 11(1):1317, 2024

    2024

  75. [75]

    Cognitive maps: What are they and why study them?Journal of environ- mental psychology, 14(1):1–19, 1994

    Robert M Kitchin. Cognitive maps: What are they and why study them?Journal of environ- mental psychology, 14(1):1–19, 1994

    1994

  76. [76]

    Human relational memory requires time and sleep.Proceedings of the National Academy of Sciences, 104(18):7723–7728, 2007

    Jeffrey M Ellenbogen, Peter T Hu, Jessica D Payne, Debra Titone, and Matthew P Walker. Human relational memory requires time and sleep.Proceedings of the National Academy of Sciences, 104(18):7723–7728, 2007

    2007

  77. [77]

    OpenAI GPT-5 System Card

    OpenAI. GPT-5 system card.arXiv preprint arXiv:2601.03267, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  78. [78]

    System card: Claude Opus 4.6

    Anthropic. System card: Claude Opus 4.6. 2026

    2026

  79. [79]

    The Llama 3 Herd of Models

    Llama Team. The Llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024. 14

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  80. [80]

    Qwen3 Technical Report

    Qwen Team. Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

Showing first 80 references.