pith. sign in

arxiv: 2505.00472 · v2 · submitted 2025-05-01 · 💻 cs.AI · cs.DC· cs.MA· cs.NI

UserCentrix: An Agentic Memory-augmented AI Framework for Smart Spaces

Alaa Saleh , Sasu Tarkoma , Praveen Kumar Donta , Anders Lindgren , Naser Hossein Motlagh , Schahram Dustdar , Susanna Pirttikangas , Lauri Lov\'en This is my paper

Pith reviewed 2026-05-22 17:11 UTC · model grok-4.3

classification 💻 cs.AI cs.DCcs.MAcs.NI
keywords agentic AIsmart spacesintent-driven systemsresource managementedge computingautonomous decision making
0
0 comments X

The pith

UserCentrix uses detected user intent as a control signal to prioritize decisions and balance speed, accuracy, and cost in smart spaces.

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

The paper introduces UserCentrix as a hybrid agentic orchestration framework for smart spaces that adds autonomous modules capable of intent-driven decisions. It treats user intent as the key signal that sets priorities, adjusts resource use, and adapts how the system trades off latency against quality. This matters in settings with limited computing power because it promises to keep operations responsive without constant high-resource demands. Experiments show the approach supports real-time monitoring and efficient processing especially when running on edge devices.

Core claim

UserCentrix is a hybrid agentic orchestration framework for smart spaces that optimizes resource management and enhances user experience through urgency-aware and intent-driven decision-making mechanisms. The framework integrates interactive modules equipped with agentic behavior and autonomous decision-making capabilities to dynamically balance latency, accuracy, and computational cost. User intent functions as a governing control signal that prioritizes decisions, regulates task execution and resource allocation, and guides the adaptation of decision-making strategies to balance trade-offs between speed and accuracy. Experimental results demonstrate that the framework autonomously enables

What carries the argument

Intent-driven decision-making where user intent acts as the governing control signal to prioritize tasks, regulate execution, and adapt strategies for balancing speed and accuracy.

Load-bearing premise

User intent can be reliably detected from interactions and used as an accurate governing control signal that correctly prioritizes decisions and regulates task execution without introducing significant errors or overhead.

What would settle it

A test showing that intent detection often misidentifies goals leading to wrong task priorities, or that the framework's overhead exceeds that of simpler methods in the same edge conditions.

Figures

Figures reproduced from arXiv: 2505.00472 by Alaa Saleh, Anders Lindgren, Lauri Lov\'en, Naser Hossein Motlagh, Praveen Kumar Donta, Sasu Tarkoma, Schahram Dustdar, Susanna Pirttikangas.

Figure 1
Figure 1. Figure 1: UserCentrix Framework. resource allocation through three core modules equipped with adaptive agentic behavior and autonomous decision￾making capabilities. Subsequent sections detail the workflow, module functionalities, and agent roles. Algorithms and use cases, along with illustrative figures presenting inputs, outputs, and intermediate processes, are included to clarify the framework’s operation. 3.1 Use… view at source ↗
Figure 2
Figure 2. Figure 2: User Intent Processing Use Case within UserCentrix Framework. Algorithm 2 presents the Sub-tasks Execution Module, which is designed to execute a set of sub-task solutions T = {1 : k} using a corresponding dataset D. The execution process begins with the creation of k low-level agents, where each agent is assigned a specific sub-task Ti (Line 1). For each sub-task Ti , the respective agent retrieves a comp… view at source ↗
Figure 3
Figure 3. Figure 3: High-urgency Workflow within UserCentrix Framework. execution and to the environment agent for monitoring. The environment agents continuously monitors for any changes between user preferences and the Smart Campus dataset during the booking period. If changes are detected, it generates new commands to adjust the environment accordingly, ensuring real-time adaptation to user needs. 3.3.3 Low-urgency Scenari… view at source ↗
Figure 4
Figure 4. Figure 4: Low-urgency Workflow within UserCentrix Framework. • Mistral (7.25B)10, an open-source model developed by Mistral AI with advanced reasoning capabilities and rapid inference speed. 10https://mistral.ai/ 12 [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Personal Agent Performance Evaluation [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Classifier Agent Performance Evaluation. [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation of Low-urgency and High-urgency Agent Performance at the Edge. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Evaluation of Low-urgency High-urgency Agent Performance at the Cloud. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Evaluation of Low-Urgency Agent Performance Metrics and Pareto Analysis. (a) Gemini-1.5 flash (b) [PITH_FULL_IMAGE:figures/full_fig_p018_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Evaluation of Low-Level Agents Performance at Cloud: (a) Gemini-1.5 flash (b) Command-r7b (c) Claude [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Low-urgency Learning Performance Evaluation before learning (a) Response of Metrics (b) Elapsed time [PITH_FULL_IMAGE:figures/full_fig_p020_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Low-urgency Learning Performance Evaluation after learning (a) Response of Metrics (b) Elapsed time (c) [PITH_FULL_IMAGE:figures/full_fig_p021_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Environment Agent Performance Evaluation. [PITH_FULL_IMAGE:figures/full_fig_p021_13.png] view at source ↗
read the original abstract

Agentic Artificial Intelligence (AI) constitutes a transformative paradigm in the evolution of intelligent agents and decision-support systems, redefining smart environments by enhancing operational efficiency, optimizing resource allocation, and strengthening systemic resilience. This paper presents UserCentrix, a hybrid agentic orchestration framework for smart spaces that optimizes resource management and enhances user experience through urgency-aware and intent-driven decision-making mechanisms. The framework integrates interactive modules equipped with agentic behavior and autonomous decision-making capabilities to dynamically balance latency, accuracy, and computational cost. User intent functions as a governing control signal that prioritizes decisions, regulates task execution and resource allocation, and guides the adaptation of decision-making strategies to balance trade-offs between speed and accuracy. Experimental results demonstrate that the framework autonomously enables efficient intent processing and real-time monitoring, while balancing reasoning quality and computational efficiency, particularly under resource-constrained edge conditions.

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

2 major / 0 minor

Summary. The paper presents UserCentrix, a hybrid agentic orchestration framework for smart spaces that uses interactive modules with autonomous decision-making to optimize resource management and user experience. User intent acts as a governing control signal to prioritize decisions, regulate task execution and resource allocation, and adapt strategies for balancing latency, accuracy, and computational cost. The abstract claims that experiments demonstrate efficient intent processing, real-time monitoring, and balanced reasoning quality under resource-constrained edge conditions.

Significance. If the performance claims hold with quantitative support, the work could contribute to agentic AI systems by demonstrating an intent-driven approach to dynamic trade-off management in edge-based smart environments.

major comments (2)
  1. Abstract: the claim that 'Experimental results demonstrate that the framework autonomously enables efficient intent processing and real-time monitoring, while balancing reasoning quality and computational efficiency' supplies no metrics (e.g., precision/recall for intent detection, latency/accuracy trade-offs, or baselines), datasets, or implementation details. This is load-bearing for the central claim because the framework treats intent inference as a reliable control signal without evidence that detection errors do not propagate into scheduling or allocation failures.
  2. Framework description (throughout): the assumption that user intent can be reliably detected from interactions and used to correctly prioritize decisions without significant overhead or errors is presupposed rather than measured. No quantification of intent-module accuracy or downstream impact appears, leaving the autonomous balancing mechanism unverified.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive review. We agree that the abstract and framework sections would be improved by including explicit quantitative metrics and analysis for intent detection accuracy and its downstream effects. We will revise the manuscript to address these concerns directly.

read point-by-point responses

  1. Referee: Abstract: the claim that 'Experimental results demonstrate that the framework autonomously enables efficient intent processing and real-time monitoring, while balancing reasoning quality and computational efficiency' supplies no metrics (e.g., precision/recall for intent detection, latency/accuracy trade-offs, or baselines), datasets, or implementation details. This is load-bearing for the central claim because the framework treats intent inference as a reliable control signal without evidence that detection errors do not propagate into scheduling or allocation failures.

    Authors: We accept this criticism. The abstract is currently high-level and does not report concrete numbers. In the revised version we will replace the generic claim with specific results: intent detection precision/recall, measured latency versus accuracy trade-offs under varying edge constraints, baseline comparisons, and the datasets and implementation details used. We will also add a short robustness analysis showing that the scheduling and allocation layers remain stable under realistic intent detection error rates. revision: yes

  2. Referee: Framework description (throughout): the assumption that user intent can be reliably detected from interactions and used to correctly prioritize decisions without significant overhead or errors is presupposed rather than measured. No quantification of intent-module accuracy or downstream impact appears, leaving the autonomous balancing mechanism unverified.

    Authors: We agree that explicit quantification is needed. We will expand the framework and evaluation sections to report intent-module accuracy (precision, recall, F1), computational overhead of the intent inference step, and an ablation or sensitivity study that measures how detection errors propagate into task prioritization, resource allocation, and the final latency-accuracy balance. These additions will directly verify the autonomous balancing claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper presents UserCentrix as a hybrid agentic orchestration framework for smart spaces, describing its architecture, modules with agentic behavior, and the role of user intent as a governing control signal for prioritization and resource allocation. It reports experimental outcomes on efficient intent processing and balancing of quality/efficiency under edge constraints. No equations, mathematical derivations, predictions, or fitted parameters are present in the provided text. There are no self-definitional reductions, fitted inputs renamed as predictions, load-bearing self-citations, uniqueness theorems imported from prior author work, or ansatzes smuggled via citation. The central claims rest on system design choices and direct experimental reporting rather than any chain that reduces to its own inputs by construction. This is a standard self-contained architecture paper with independent content.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that intent can be extracted and used as a reliable control signal; the framework itself is the primary invented entity. No free parameters or mathematical axioms are stated.

axioms (1)
  • domain assumption User intent can be reliably detected from interactions and used as an accurate governing control signal that correctly prioritizes decisions and regulates task execution without introducing significant errors or overhead.
    Invoked throughout the abstract as the mechanism that 'prioritizes decisions, regulates task execution and resource allocation'.
invented entities (1)
  • UserCentrix hybrid agentic orchestration framework no independent evidence
    purpose: To integrate interactive modules with autonomous decision-making for intent-driven resource management in smart spaces
    Introduced as the main contribution of the paper.

pith-pipeline@v0.9.0 · 5718 in / 1321 out tokens · 57432 ms · 2026-05-22T17:11:36.839617+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages

  1. [1]

    A survey on evaluation of large language models,

    Y . Chang, X. Wang, J. Wang, Y . Wu, L. Yang, K. Zhu, H. Chen, X. Yi, C. Wang, Y . Wang, W. Ye, Y . Zhang, Y . Chang, P. S. Yu, Q. Yang, and X. Xie, “A survey on evaluation of large language models,”ACM Trans. Intell. Syst. Technol., vol. 15, Mar. 2024

    work page 2024

  2. [2]

    Digital twins for smart spaces—beyond iot analytics,

    N. H. Motlagh, M. A. Zaidan, L. Lovén, P. L. Fung, T. Hänninen, R. Morabito, P. Nurmi, and S. Tarkoma, “Digital twins for smart spaces—beyond iot analytics,”IEEE internet of things journal, vol. 11, no. 1, pp. 573–583, 2023

    work page 2023

  3. [3]

    Chatfive: Enhancing user experience in likert scale personality test through interactive conversation with llm agents,

    J. Lee, Y . Choi, M. Song, and S. Park, “Chatfive: Enhancing user experience in likert scale personality test through interactive conversation with llm agents,” inProceedings of the 6th ACM Conference on Conversational User Interfaces, CUI ’24, (New York, NY , USA), Association for Computing Machinery, 2024

    work page 2024

  4. [4]

    Sasha: Creative goal-oriented reasoning in smart homes with large language models,

    E. King, H. Yu, S. Lee, and C. Julien, “Sasha: Creative goal-oriented reasoning in smart homes with large language models,”Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 8, Mar. 2024

    work page 2024

  5. [5]

    Reasoning abilities of large language models: In-depth analysis on the abstraction and reasoning corpus,

    S. Lee, W. Sim, D. Shin, W. Seo, J. Park, S. Lee, S. Hwang, S. Kim, and S. Kim, “Reasoning abilities of large language models: In-depth analysis on the abstraction and reasoning corpus,”ACM Trans. Intell. Syst. Technol., Jan. 2025. Just Accepted

    work page 2025

  6. [6]

    Generative ai for self-adaptive systems: State of the art and research roadmap,

    J. Li, M. Zhang, N. Li, D. Weyns, Z. Jin, and K. Tei, “Generative ai for self-adaptive systems: State of the art and research roadmap,”ACM Trans. Auton. Adapt. Syst., vol. 19, Sept. 2024

    work page 2024

  7. [7]

    Bespoke: Using llm agents to generate just-in-time interfaces by reasoning about user intent,

    P. Nandy, S. O. Adalgeirsson, A. K. Sinha, T. Kraljic, M. Cleron, L. Shi, A. Singh, A. Chaudhary, A. Ganti, C. A. Melancon, S. Zhang, D. Robishaw, H. Ciurdar, J. Secor, K. A. Robertsen, K. Climer, M. Le, M. Venkatesan, P. Chi, P. Li, P. F. McDermott, R. Shim, S. Onsan, S. Vaishnav, and S. Guamán, “Bespoke: Using llm agents to generate just-in-time interfa...

    work page 2024

  8. [8]

    Gesturegpt: Toward zero-shot free-form hand gesture understanding with large language model agents,

    X. Zeng, X. Wang, T. Zhang, C. Yu, S. Zhao, and Y . Chen, “Gesturegpt: Toward zero-shot free-form hand gesture understanding with large language model agents,”Proc. ACM Hum.-Comput. Interact., vol. 8, Oct. 2024

    work page 2024

  9. [9]

    Improving llm-based recommender systems with user- controllable profiles,

    S. Wo´ zniak, J. Duszenko, J. Koco´n, and P. Kazienko, “Improving llm-based recommender systems with user- controllable profiles,” inCompanion Proceedings of the ACM on Web Conference 2025, WWW ’25, (New York, NY , USA), p. 2102–2111, Association for Computing Machinery, 2025

    work page 2025

  10. [10]

    Recommender ai agent: Integrating large language models for interactive recommendations,

    X. Huang, J. Lian, Y . Lei, J. Yao, D. Lian, and X. Xie, “Recommender ai agent: Integrating large language models for interactive recommendations,”ACM Trans. Inf. Syst., vol. 43, June 2025

    work page 2025

  11. [11]

    Expressing the needs in smart home: What is the end users’ favorite way,

    X. Chen, S. Chen, Z. Jin, H. Bian, Z. Chen, and H. Li, “Expressing the needs in smart home: What is the end users’ favorite way,”ACM Trans. Comput.-Hum. Interact., vol. 32, Apr. 2025

    work page 2025

  12. [12]

    Prompt2task: Automating ui tasks on smartphones from textual prompts,

    T. Huang, C. Yu, W. Shi, Z. Peng, D. Yang, W. Sun, and Y . Shi, “Prompt2task: Automating ui tasks on smartphones from textual prompts,”ACM Trans. Comput.-Hum. Interact., vol. 32, June 2025

    work page 2025

  13. [13]

    A survey on the memory mechanism of large language model based agents,

    Z. Zhang, Q. Dai, X. Bo, C. Ma, R. Li, X. Chen, J. Zhu, Z. Dong, and J.-R. Wen, “A survey on the memory mechanism of large language model based agents,”ACM Trans. Inf. Syst., July 2025. Just Accepted

    work page 2025

  14. [14]

    Memindex: Agentic event-based distributed memory management for multi-agent systems,

    A. Saleh, S. Tarkoma, A. Lindgren, P. K. Donta, S. Dustdar, S. Pirttikangas, and L. Lovén, “Memindex: Agentic event-based distributed memory management for multi-agent systems,”ACM Transactions on Autonomous and Adaptive Systems, 2025

    work page 2025

  15. [15]

    Llm-powered smart spaces with multi-agent user-centric adaptation,

    A. Saleh, M. Hassaan, Q. Zhang, S. Tarkoma, S. Pirttikangas, and L. Lovén, “Llm-powered smart spaces with multi-agent user-centric adaptation,” in2025 IEEE 45th International Conference on Distributed Computing Systems Workshops (ICDCSW), pp. 576–581, IEEE, 2025

    work page 2025

  16. [16]

    Towards message brokers for generative ai: Survey, challenges, and opportunities,

    A. Saleh, R. Morabito, S. Dustdar, S. Tarkoma, S. Pirttikangas, and L. Lovén, “Towards message brokers for generative ai: Survey, challenges, and opportunities,”ACM Comput. Surv., June 2025. Just Accepted

    work page 2025

  17. [17]

    Follow-Me AI: Energy- Efficient User Interaction with Smart Environments,

    A. Saleh, P. K. Donta, R. Morabito, N. Hossein Motlagh, S. Tarkoma, and L. Lovén, “Follow-Me AI: Energy- Efficient User Interaction with Smart Environments,”IEEE Pervasive Computing, pp. 1–10, 02 2025

    work page 2025

  18. [18]

    Towards lifelong learning of large language models: A survey,

    J. Zheng, S. Qiu, C. Shi, and Q. Ma, “Towards lifelong learning of large language models: A survey,”ACM Comput. Surv., vol. 57, Mar. 2025

    work page 2025

  19. [19]

    Resource-efficient algorithms and systems of foundation models: A survey,

    M. Xu, D. Cai, W. Yin, S. Wang, X. Jin, and X. Liu, “Resource-efficient algorithms and systems of foundation models: A survey,”ACM Comput. Surv., vol. 57, Jan. 2025

    work page 2025

  20. [20]

    Biswas and W

    A. Biswas and W. Talukdar,Building Agentic AI Systems: Create intelligent, autonomous AI agents that can reason, plan, and adapt. Online: Packt Publishing Ltd, 2025

    work page 2025

  21. [21]

    A survey of human-in-the-loop for machine learning,

    X. Wu, L. Xiao, Y . Sun, J. Zhang, T. Ma, and L. He, “A survey of human-in-the-loop for machine learning,”Future Generation Computer Systems, vol. 135, pp. 364–381, 2022

    work page 2022

  22. [22]

    A survey on llm-as-a-judge,

    J. Gu, X. Jiang, Z. Shi, H. Tan, X. Zhai, C. Xu, W. Li, Y . Shen, S. Ma, H. Liu, S. Wang, K. Zhang, Y . Wang, W. Gao, L. Ni, and J. Guo, “A survey on llm-as-a-judge,” 2025

    work page 2025

  23. [23]

    Revisiting Edge AI: Opportunities and Challenges,

    T. Meuser, L. Lovén, M. Bhuyan, S. G. Patil, S. Dustdar, A. Aral, S. Bayhan, C. Becker, E. de Lara, A. Y . Ding, et al., “Revisiting Edge AI: Opportunities and Challenges,”IEEE Internet Computing, vol. 28, no. 4, pp. 49–59, 2024

    work page 2024

  24. [24]

    Large language models empowered autonomous edge ai for connected intelligence,

    Y . Shen, J. Shao, X. Zhang, Z. Lin, H. Pan, D. Li, J. Zhang, and K. B. Letaief, “Large language models empowered autonomous edge ai for connected intelligence,”IEEE Communications Magazine, 2024

    work page 2024

  25. [25]

    Edgeshard: Efficient llm inference via collaborative edge computing,

    M. Zhang, X. Shen, J. Cao, Z. Cui, and S. Jiang, “Edgeshard: Efficient llm inference via collaborative edge computing,”IEEE Internet of Things Journal, 2024

    work page 2024

  26. [26]

    Hybrid slm and llm for edge-cloud collaborative inference,

    Z. Hao, H. Jiang, S. Jiang, J. Ren, and T. Cao, “Hybrid slm and llm for edge-cloud collaborative inference,” in Proceedings of the Workshop on Edge and Mobile Foundation Models, pp. 36–41, 2024

    work page 2024

  27. [27]

    Edge-llm: Enabling efficient large language model adaptation on edge devices via unified compression and adaptive layer voting,

    Z. Yu, Z. Wang, Y . Li, R. Gao, X. Zhou, S. R. Bommu, Y . Zhao, and Y . Lin, “Edge-llm: Enabling efficient large language model adaptation on edge devices via unified compression and adaptive layer voting,” inProceedings of the 61st ACM/IEEE Design Automation Conference, pp. 1–6, 2024

    work page 2024

  28. [28]

    Hybrid LLM: Cost-efficient and quality-aware query routing,

    D. Ding, A. Mallick, C. Wang, R. Sim, S. Mukherjee, V . Rühle, L. V . S. Lakshmanan, and A. H. Awadallah, “Hybrid LLM: Cost-efficient and quality-aware query routing,” inThe Twelfth International Conference on Learning Representations, 2024. 24 USERCENTRIX: ANAGENTICMEMORY-AUGMENTEDAI FRAMEWORK FORSMARTSPACES

    work page 2024

  29. [29]

    Dlora: Distributed parameter-efficient fine-tuning solution for large language model,

    C. Gao and S. Zhang, “Dlora: Distributed parameter-efficient fine-tuning solution for large language model,” in EMNLP 2024 - 2024 Conference on Empirical Methods in Natural Language Processing, Findings of EMNLP 2024(Y . Al-Onaizan, M. Bansal, and Y .-N. Chen, eds.), EMNLP 2024 - 2024 Conference on Empirical Methods in Natural Language Processing, Finding...

    work page 2024

  30. [30]

    Cached model-as-a-resource: Provisioning large language model agents for edge intelligence in space–air–ground integrated networks,

    M. Xu, D. Niyato, H. Zhang, J. Kang, Z. Xiong, S. Mao, and Z. Han, “Cached model-as-a-resource: Provisioning large language model agents for edge intelligence in space–air–ground integrated networks,”IEEE Transactions on Networking, vol. 34, pp. 2850–2864, 2026

    work page 2026

  31. [31]

    Embodied LLM agents learn to cooperate in organized teams,

    X. Guo, K. Huang, J. Liu, W. Fan, N. Vélez, Q. Wu, H. Wang, T. L. Griffiths, and M. Wang, “Embodied LLM agents learn to cooperate in organized teams,” inLanguage Gamification - NeurIPS 2024 Workshop, 2024

    work page 2024

  32. [32]

    Map: Multi-user personalization with collaborative llm-powered agents,

    C. P. Lee, J. Choi, and B. Mutlu, “Map: Multi-user personalization with collaborative llm-powered agents,” in Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, CHI EA ’25, (New York, NY , USA), Association for Computing Machinery, 2025

    work page 2025

  33. [33]

    Mixture-of-agents enhances large language model capabilities,

    J. Wang, J. W ANG, B. Athiwaratkun, C. Zhang, and J. Zou, “Mixture-of-agents enhances large language model capabilities,” inThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  34. [34]

    Scaling LLM test-time compute optimally can be more effective than scaling parameters for reasoning,

    C. V . Snell, J. Lee, K. Xu, and A. Kumar, “Scaling LLM test-time compute optimally can be more effective than scaling parameters for reasoning,” inThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  35. [35]

    Agents thinking fast and slow: A talker-reasoner architecture,

    K. Christakopoulou, S. Mourad, and M. Mataric, “Agents thinking fast and slow: A talker-reasoner architecture,” inNeurIPS 2024 Workshop on Open-World Agents, 2024

    work page 2024

  36. [36]

    Mutual reasoning makes smaller LLMs stronger problem-solver,

    Z. Qi, M. MA, J. Xu, L. L. Zhang, F. Yang, and M. Yang, “Mutual reasoning makes smaller LLMs stronger problem-solver,” inThe Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  37. [37]

    Thinking LLMs: General instruction following with thought generation,

    T. Wu, J. Lan, W. Yuan, J. Jiao, J. E. Weston, and S. Sukhbaatar, “Thinking LLMs: General instruction following with thought generation,” inForty-second International Conference on Machine Learning, 2025

    work page 2025

  38. [38]

    Quiet-STar: Language models can teach themselves to think before speaking,

    E. Zelikman, G. R. Harik, Y . Shao, V . Jayasiri, N. Haber, and N. Goodman, “Quiet-STar: Language models can teach themselves to think before speaking,” inFirst Conference on Language Modeling, 2024

    work page 2024

  39. [39]

    Efficient management of llm-based coaching agents’ reasoning while maintaining interaction quality and speed,

    A. Göldi, R. Rietsche, and L. Ungar, “Efficient management of llm-based coaching agents’ reasoning while maintaining interaction quality and speed,” inProceedings of the 2025 CHI Conference on Human Factors in Computing Systems, CHI ’25, (New York, NY , USA), Association for Computing Machinery, 2025

    work page 2025

  40. [40]

    " my agent understands me better

    Y . Hou, H. Tamoto, and H. Miyashita, “" my agent understands me better": Integrating dynamic human-like memory recall and consolidation in llm-based agents,” inExtended Abstracts of the CHI Conference on Human Factors in Computing Systems, pp. 1–7, 2024

    work page 2024

  41. [41]

    S. J. Russell and P. Norvig,Artificial intelligence: a Modern Approach, Fourth Edition. Pearson, 2020

    work page 2020

  42. [42]

    Case-based reasoning: A review,

    I. Watson and F. Marir, “Case-based reasoning: A review,”The knowledge engineering review, vol. 9, no. 4, pp. 327–354, 1994

    work page 1994

  43. [43]

    Smart campus oulu indoor climate, air-quality and motion

    U. of Oulu, “Smart campus oulu indoor climate, air-quality and motion.” https://doi.org/10.23729/ b9adb0a2-7381-45db-b32f-7e78ae1bc9e3, 6 2021. University of Oulu, CWC - Verkot ja järjestelmät

    work page 2021

  44. [44]

    Monitoring social distancing in smart spaces using infrastructure-based sensors,

    N. H. Motlagh, P. Toivonen, M. A. Zaidan, E. Lagerspetz, E. Peltonen, E. Gilman, P. Nurmi, and S. Tarkoma, “Monitoring social distancing in smart spaces using infrastructure-based sensors,” in2021 IEEE 7th World Forum on Internet of Things (WF-IoT), pp. 124–129, IEEE, 2021

    work page 2021

  45. [45]

    Langchain

    LangChain, “Langchain.”https://www.langchain.com/. Last accessed: April 8, 2026. 25

    work page 2026