pith. sign in

arxiv: 2603.24858 · v2 · pith:6OXYP6AGnew · submitted 2026-03-25 · 💻 cs.HC · cs.MA

Context-Mediated Domain Adaptation in Multi-Agent Sensemaking Systems

Anton Wolter , Leon Haag , Vaishali Dhanoa , Niklas Elmqvist This is my paper

Pith reviewed 2026-05-22 10:05 UTC · model grok-4.3

classification 💻 cs.HC cs.MA
keywords domain adaptationmulti-agent systemssensemakingtacit knowledgehuman-AI collaborationimplicit specificationsLLM reasoningedit patterns
0
0 comments X

The pith

User modifications to AI-generated artifacts serve as implicit domain specifications that reshape multi-agent LLM reasoning.

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

The paper proposes context-mediated domain adaptation for multi-agent sensemaking systems. Domain experts reveal tacit knowledge when they edit AI outputs by correcting terms, restructuring arguments, or shifting emphasis. Rather than treating these changes as one-off fixes, the approach converts them into specifications that update how the agents reason going forward. This creates a loop where vague starting prompts grow into precise domain rules through repeated human-AI interaction. A small evaluation with experts produced 46 extracted knowledge entries from edit patterns.

Core claim

Context-mediated domain adaptation treats user modifications to system-generated artifacts as implicit domain specifications that reshape LLM-powered multi-agent reasoning behavior, creating bidirectional semantic links between artifacts and reasoning, specification bootstrapping from vague prompts, implicit knowledge transfer via reverse-engineered edits, and in-context adaptation based on correction patterns, as shown in the Seedentia system where 46 domain knowledge entries were extracted from expert edits to research questions.

What carries the argument

context-mediated domain adaptation, the mechanism that converts observed user edits into domain specifications which then guide subsequent multi-agent LLM reasoning through reverse engineering and in-context learning

Load-bearing premise

User modifications to system-generated artifacts reliably encode tacit domain knowledge that can be reverse-engineered into precise specifications capable of reshaping subsequent multi-agent LLM reasoning behavior.

What would settle it

A larger controlled study in which the same initial prompts are run once with the extracted knowledge entries incorporated and once without, then measuring whether independent raters find statistically significant differences in domain accuracy or coherence of the final outputs.

Figures

Figures reproduced from arXiv: 2603.24858 by Anton Wolter, Leon Haag, Niklas Elmqvist, Vaishali Dhanoa.

Figure 1
Figure 1. Figure 1: Context-Mediated Domain Adaptation transforms ephemeral user interactions into persistent domain knowledge. As user knowledge and LLM model knowledge deviate we analyze user interaction and edits in order to extract implicit domain knowledge. Through iterative refinement our approach expands the shared context substantially, capturing domain-specific terminol￾ogy, conventions, and patterns. This accumulate… view at source ↗
Figure 2
Figure 2. Figure 2: shows the first two interaction modes, as context-based generation runs initially and ideally requires no user interaction. Based on hovering certain elements triggering the said interactions, the affected elements of the interactions are highlighted. (a) Direct manipulation mode. Interface showing direct content editing capabilities where users can modify research questions inline. The border is highlight… view at source ↗
Figure 3
Figure 3. Figure 3: Prompt-based generation. Complete workflow for prompt-based artifact regeneration showing the input dialog for natural language instructions and the asynchronous generation process. The interface maintains application responsiveness during AI processing, demonstrating the fire-and-forget architecture that decouples user interactions from computational workloads. Manuscript submitted to ACM [PITH_FULL_IMAG… view at source ↗
Figure 4
Figure 4. Figure 4: Edit history visualization. The AIContentWrapper component provides integrated edit history functionality that powers context-mediated domain adaptation. 4.2.3 Real-time State Management and Persistence. The system implements optimistic UI updates through React state management, providing immediate feedback while asynchronous save operations complete in the background. The Next.js application functions as … view at source ↗
Figure 5
Figure 5. Figure 5: Agentic task processing graph. The backend workflow graph is centered on the planner router node, which conditionally dispatches tasks to specialized nodes for paper retrieval, context-based research question generation, and edit-driven knowledge extraction. Node outputs are merged back into a unified state and persisted via the agent tasks infrastructure, enabling asynchronous execution while maintaining … view at source ↗
Figure 6
Figure 6. Figure 6: Context-mediated domain adaptation workflow tracing. Langfuse tracing demonstrates how the bidirectional learning cycle operates: user modifications flow through the extract_implicit_knowledge node (shown processing three user interactions), with extracted knowledge subsequently injected into the generate_evaluation_questions node’s system prompt. Notice how the interface makes the complete knowledge trans… view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation protocol interface. Key components of the controlled study we conducted for assessing context-mediated domain adaptation effectiveness. The system captures baseline quality assessments and provides standardized interaction protocols to ensure consistent evaluation of bidirectional learning mechanisms across participants and sessions. 5.1.2 Materials and Study Design. Our evaluation employed a se… view at source ↗
read the original abstract

Domain experts possess tacit knowledge that they cannot easily articulate through explicit specifications. When experts modify AI-generated artifacts by correcting terminology, restructuring arguments, and adjusting emphasis, these edits reveal domain understanding that remains latent in traditional prompt-based interactions. Current systems treat such modifications as endpoint corrections rather than as implicit specifications that could reshape subsequent reasoning. We propose context-mediated domain adaptation, a paradigm where user modifications to system-generated artifacts serve as implicit domain specification that reshapes LLM-powered multi-agent reasoning behavior. Through our system Seedentia, a web-based multi-agent framework for sense-making, we demonstrate bidirectional semantic links between generated artifacts and system reasoning. Our approach enables specification bootstrapping where vague initial prompts evolve into precise domain specifications through iterative human-AI collaboration, implicit knowledge transfer through reverse-engineered user edits, and in-context learning where agent behavior adapts based on observed correction patterns. We present results from an evaluation with domain experts who generated and modified research questions from academic papers. Our system extracted 46 domain knowledge entries from user modifications, demonstrating the feasibility of capturing implicit expertise through edit patterns, though the limited sample size constrains conclusions about systematic quality improvements.

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 / 2 minor

Summary. The manuscript proposes context-mediated domain adaptation, a paradigm in which user modifications to AI-generated artifacts in the Seedentia multi-agent sensemaking system are treated as implicit domain specifications that can reshape subsequent LLM-powered multi-agent reasoning. Through a user study with domain experts modifying research questions derived from academic papers, the system extracted 46 domain knowledge entries from edit patterns, illustrating feasibility of capturing tacit expertise, specification bootstrapping, and in-context adaptation, albeit with constraints due to limited sample size.

Significance. If the bidirectional adaptation mechanisms are shown to function as described, the work could advance human-AI collaboration in sensemaking by converting latent edits into reusable specifications that improve multi-agent outputs without explicit re-prompting. The approach of reverse-engineering user corrections into precise domain knowledge offers a promising direction for implicit knowledge transfer, though the present evaluation focuses on extraction counts rather than demonstrated behavioral change.

major comments (2)
  1. [Evaluation] Evaluation section: The reported results consist only of the extraction of 46 domain knowledge entries from user modifications to research-question artifacts. No controlled before/after comparisons of agent reasoning traces, output distributions, expert-rated quality, or behavioral metrics are provided to demonstrate that the extracted entries actually reshape multi-agent LLM reasoning as required by the central claim in the abstract and introduction.
  2. [Abstract and Evaluation] Abstract and Evaluation: The extraction process for the 46 entries is described without details on method, quality metrics, inter-rater reliability, baselines, or error analysis, leaving the reliability of the implicit-knowledge-capture step unassessed and weakening support for the specification-bootstrapping mechanism.
minor comments (2)
  1. [Abstract] Abstract: The distinction between the general paradigm of context-mediated domain adaptation and its concrete realization in the Seedentia implementation could be stated more explicitly to avoid conflating the two.
  2. [Throughout] Throughout manuscript: Terminology for core concepts such as 'context-mediated domain adaptation,' 'specification bootstrapping,' and 'implicit knowledge transfer' should be used consistently to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address the major concerns point by point below, agreeing where the evaluation can be strengthened and outlining specific revisions to improve clarity and support for our claims without overstating the current results.

read point-by-point responses

  1. Referee: [Evaluation] Evaluation section: The reported results consist only of the extraction of 46 domain knowledge entries from user modifications to research-question artifacts. No controlled before/after comparisons of agent reasoning traces, output distributions, expert-rated quality, or behavioral metrics are provided to demonstrate that the extracted entries actually reshape multi-agent LLM reasoning as required by the central claim in the abstract and introduction.

    Authors: We agree that direct evidence of behavioral change in multi-agent reasoning would provide stronger support for the full adaptation loop. The present evaluation was scoped to demonstrate the feasibility of extracting implicit domain knowledge from edits as the foundational step in context-mediated adaptation, consistent with the limited sample size noted in the abstract. In the revision we will add a new subsection with qualitative before/after examples drawn from the study data, showing how extracted entries are injected into subsequent agent prompts and the resulting shifts in reasoning traces and output structure. We will also explicitly discuss the absence of quantitative behavioral metrics as a limitation and describe planned follow-up experiments. revision: yes

  2. Referee: [Abstract and Evaluation] Abstract and Evaluation: The extraction process for the 46 entries is described without details on method, quality metrics, inter-rater reliability, baselines, or error analysis, leaving the reliability of the implicit-knowledge-capture step unassessed and weakening support for the specification-bootstrapping mechanism.

    Authors: We accept that the extraction methodology requires more transparent documentation. The revised manuscript will expand the Evaluation section to describe the step-by-step process by which the 46 entries were derived from edit patterns, including the coding scheme, any quantitative quality metrics used, inter-rater reliability statistics (or note if single-coder), comparison against simple keyword baselines, and an error analysis of cases where extraction was ambiguous. These additions will directly address the reliability of the implicit-knowledge-capture step and better ground the specification-bootstrapping claim. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical extraction count stands independent of inputs

full rationale

The paper advances a system implementation (Seedentia) and reports an empirical count of 46 extracted domain knowledge entries from user modifications to research-question artifacts. No equations, first-principles derivations, or predictive models are presented whose outputs reduce to fitted parameters or self-referential definitions by construction. The central feasibility claim rests on the observed extraction tally and system description rather than any load-bearing self-citation chain, ansatz smuggling, or renaming of known results. The evaluation measures extraction volume directly from the study data; it does not rename a fitted quantity as a prediction or invoke prior author work to close a logical loop. This is a standard self-contained empirical demonstration.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the domain assumption that edits encode extractable tacit knowledge and on the invented paradigm itself; no numeric free parameters are mentioned.

axioms (1)
  • domain assumption User modifications to AI-generated artifacts encode implicit domain knowledge that can be systematically extracted and applied to adapt agent reasoning.
    Invoked throughout the proposal as the basis for specification bootstrapping and implicit knowledge transfer.
invented entities (1)
  • Context-mediated domain adaptation no independent evidence
    purpose: Paradigm treating user edits as implicit domain specifications that reshape multi-agent reasoning.
    New framing introduced to organize the bidirectional links and bootstrapping process.

pith-pipeline@v0.9.0 · 5734 in / 1419 out tokens · 43869 ms · 2026-05-22T10:05:10.626663+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

55 extracted references · 55 canonical work pages · 1 internal anchor

  1. [1]

    Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz

    Saleema Amershi, Dan Weld, Mihaela Vorvoreanu, Adam Fourney, Besmira Nushi, Penny Collisson, Jina Suh, Shamsi Iqbal, Paul N. Bennett, Kori Inkpen, Jaime Teevan, Ruth Kikin-Gil, and Eric Horvitz. 2019. Guidelines for Human-AI Interaction. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 3:1–3:13. doi:10.114...

    work page doi:10.1145/3290605.3300233 2019

  2. [2]

    Glassman

    Ian Arawjo, Chelse Swoopes, Priyan Vaithilingam, Martin Wattenberg, and Elena L. Glassman. 2024. ChainForge: A Visual Toolkit for Prompt Engineering and LLM Hypothesis Testing. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 304:1–304:18. doi:10.1145/3613904.3642016

    work page doi:10.1145/3613904.3642016 2024

  3. [3]

    Sriram Karthik Badam, Niklas Elmqvist, and Jean-Daniel Fekete. 2017. Steering the craft: UI elements and visualizations for supporting progressive visual analytics.Computer Graphics Forum36, 3 (2017), 491–502. doi:10.1111/cgf.13205

    work page doi:10.1111/cgf.13205 2017

  4. [4]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Christopher Hesse, Mark Chen, Eric Sigler, Mateusz Litwin...

    work page 2020

  5. [5]

    Minsuk Chang, Yao Wang, Huichen Will Wang, Yuanhong Zhou, Andreas Bulling, and Cindy Xiong Bearfield. 2025. Tell Me Without Telling Me: Two-Way Prediction of Visualization Literacy and Visual Attention. arXiv:2508.03713 [cs.HC] https://arxiv.org/abs/2508.03713

    work page arXiv 2025

  6. [6]

    Xinyun Chen, Maxwell Lin, Nathanael Schärli, and Denny Zhou. 2024. Teaching Large Language Models to Self-Debug. InInternational Conference on Learning Representations (ICLR). https://openreview.net/forum?id=KuPixIqPiq

    work page 2024

  7. [7]

    Yongchao Chen, Jacob Arkin, Yilun Hao, Yang Zhang, Nicholas Roy, and Chuchu Fan. 2024. PRompt Optimization in Multi-Step Tasks (PROMST): Integrating Human Feedback and Heuristic-based Sampling. InProceedings of the Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, Stroudsburg, PA, USA, 3859–3920. do...

    work page doi:10.18653/v1/2024.emnlp-main.226 2024

  8. [8]

    Zhe Cui, Sriram Karthik Badam, M Adil Yalçin, and Niklas Elmqvist. 2019. DataSite: Proactive visual data exploration with computation of insight-based recommendations.Information Visualization18, 2 (2019), 251–267. doi:10.1177/1473871618806555

    work page doi:10.1177/1473871618806555 2019

  9. [9]

    Amit Kumar Das, Mohammad Tarun, and Klaus Mueller. 2025. Charts-of-Thought: Enhancing LLM Visualization Literacy Through Structured Data Extraction. arXiv:2508.04842 [cs.HC] https://arxiv.org/abs/2508.04842

    work page arXiv 2025

  10. [10]

    Michael Desmond and Michelle Brachman. 2024. Exploring Prompt Engineering Practices in the Enterprise.CoRRabs/2403.08950 (2024), 9 pages. arXiv:2403.08950 doi:10.48550/ARXIV.2403.08950

    work page doi:10.48550/arxiv.2403.08950 2024

  11. [11]

    Vaishali Dhanoa, Anton Wolter, Gabriela Molina León, Hans-Jörg Schulz, and Niklas Elmqvist. 2025. Agentic Visualization: Extracting Agent-based Design Patterns from Visualization Systems.IEEE Computer Graphics and Applications45, 6 (2025), 89–100. doi:10.1109/MCG.2025.3607741

    work page doi:10.1109/mcg.2025.3607741 2025

  12. [12]

    Marian Dörk, Boris Müller, Jan-Erik Stange, Johanna Herseni, and Katrin Dittrich. 2020. Co-Designing Visualizations for Information Seeking and Knowledge Management.Open Information Science4, 1 (2020), 217–235. doi:10.1515/opis-2020-0102

    work page doi:10.1515/opis-2020-0102 2020

  13. [13]

    Hung Du, Srikanth Thudumu, Rajesh Vasa, and Kon Mouzakis. 2024. A Survey on Context-Aware Multi-Agent Systems: Techniques, Challenges and Future Directions.CoRRabs/2402.01968 (2024), 28 pages. arXiv:2402.01968 doi:10.48550/ARXIV.2402.01968

    work page doi:10.48550/arxiv.2402.01968 2024

  14. [14]

    Niklas Elmqvist, Pierre Dragicevic, and Jean-Daniel Fekete. 2008. Rolling the Dice: Multidimensional Visual Exploration Using Scatterplot Matrix Navigation.IEEE transactions on visualization and computer graphics14 (11 2008), 1141–8. doi:10.1109/TVCG.2008.153

    work page doi:10.1109/tvcg.2008.153 2008

  15. [15]

    Bertelsen, Akhil Arora, Kaj Grønbæk, Susanne Bødker, Clemens Nylandsted Klokmose, Rachel Charlotte Smith, Sebastian Hubenschmid, Christoph A

    Niklas Elmqvist, Eve Hoggan, Hans-Jörg Schulz, Marianne Graves Petersen, Peter Dalsgaard, Ira Assent, Olav W. Bertelsen, Akhil Arora, Kaj Grønbæk, Susanne Bødker, Clemens Nylandsted Klokmose, Rachel Charlotte Smith, Sebastian Hubenschmid, Christoph A. Johns, Gabriela Molina León, Anton Wolter, Johannes Ellemose, Vaishali Dhanoa, Simon Aagaard Enni, Mille ...

    work page arXiv 2025

  16. [16]

    Alex Endert, Patrick Fiaux, and Chris North. 2012. Semantic interaction for visual text analytics. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 473–482. doi:10.1145/2207676.2207741

    work page doi:10.1145/2207676.2207741 2012

  17. [17]

    Will Epperson, Gagan Bansal, Victor C Dibia, Adam Fourney, Jack Gerrits, Erkang (Eric) Zhu, and Saleema Amershi. 2025. Interactive Debugging and Steering of Multi-Agent AI Systems. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 156:1–156:15. doi:10.1145/3706598.3713581 Manuscript submitted to ACM Context...

    work page doi:10.1145/3706598.3713581 2025

  18. [18]

    Mohamed Amine Ferrag, Norbert Tihanyi, and Mérouane Debbah. 2025. From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review. CoRRabs/2504.19678 (2025), 44 pages. arXiv:2504.19678 doi:10.48550/ARXIV.2504.19678

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.19678 2025

  19. [19]

    Skov, and Jesper Kjeldskov

    Anders Gammelgård-Larsen, Niels van Berkel, Mikael B. Skov, and Jesper Kjeldskov. 2024. Designing for Human-AI Interaction: Comparing Intermittent, Continuous, and Proactive Interactions for a Music Application. InExtended Abstracts of the CHI Conference on Human Factors in Computing Systems(Honolulu, HI, USA)(CHI EA ’24). Association for Computing Machin...

    work page 2024

  20. [20]

    Ge Gao, Alexey Taymanov, Eduardo Salinas, Paul Mineiro, and Dipendra Misra. 2024. Aligning LLM Agents by Learning Latent Preference from User Edits. InAdvances in Neural Information Processing Systems. Curran Associates Inc., Red Hook, NY, USA, 24 pages. https://openreview.net/ forum?id=DlYNGpCuwa

    work page 2024

  21. [21]

    Péter Ferenc Gyarmati, Manfred Klaffenböck, Laura Koesten, and Torsten Möller. 2025. Do Vision-Language Models See Visualizations Like Humans? Alignment in Chart Categorization. arXiv:2509.05718 [cs.HC]

    work page arXiv 2025

  22. [22]

    Péter Ferenc Gyarmati, Dominik Moritz, Torsten Möller, and Laura Koesten. 2025. A Composable Agentic System for Automated Visual Data Reporting. arXiv:2509.05721 [cs.HC]

    work page arXiv 2025

  23. [23]

    Hearst, James F

    Marti A. Hearst, James F. Allen, Eric Horvitz, and Curry I. Guinn. 1999. Trends & Controversies: Mixed-initiative interaction.IEEE Intelligent Systems14, 5 (1999), 14–24. doi:10.1109/5254.796083

    work page doi:10.1109/5254.796083 1999

  24. [24]

    Eric Horvitz. 1999. Principles of Mixed-Initiative User Interfaces. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 159–166. doi:10.1145/302979.303030

    work page doi:10.1145/302979.303030 1999

  25. [25]

    Naveen Krishnan. 2025. Advancing Multi-Agent Systems Through Model Context Protocol: Architecture, Implementation, and Applications.CoRR abs/2504.21030 (2025), 118 pages. arXiv:2504.21030 doi:10.48550/ARXIV.2504.21030

    work page doi:10.48550/arxiv.2504.21030 2025

  26. [26]

    H. Li, G. Appleby, C. Brumar, R. Chang, and A. Suh. 2023. Knowledge Graphs in Practice: Characterizing their Users, Challenges, and Visualization Opportunities.IEEE Transactions on Visualization and Computer Graphics30, 1 (2023), 584–594. doi:10.1109/tvcg.2023.3326904

    work page doi:10.1109/tvcg.2023.3326904 2023

  27. [27]

    H. Li, Y. Wang, S. Zhang, Y. Song, and H. Qu. 2021. KG4Vis: A Knowledge Graph-Based Approach for Visualization Recommendation.IEEE Transactions on Visualization and Computer Graphics28, 1 (2021), 195–205. doi:10.1109/tvcg.2021.3114863

    work page doi:10.1109/tvcg.2021.3114863 2021

  28. [28]

    S. Liu, H. Miao, Z. Li, M. Olson, V. Pascucci, and P-T. Bremer. 2024. AVA: Towards Autonomous Visualization Agents through Visual Perception-Driven Decision-Making.Computer Graphics Forum43, 3 (2024), e15093. doi:10.1111/cgf.15093

    work page doi:10.1111/cgf.15093 2024

  29. [29]

    Angela Locoro, Silvia Golia, and Davide Falessi. 2025. DRIVE-T: A Methodology for Discriminative and Representative Data Viz Item Selection for Literacy Construct and Assessment. arXiv:2508.04160 [cs.HC] https://arxiv.org/abs/2508.04160

    work page arXiv 2025

  30. [30]

    Tymoteusz Miller, Irmina Durlik, Adrianna Łobodzińska, Lech Dorobczyński, and Robert Jasionowski. 2024. AI in Context: Harnessing Domain Knowledge for Smarter Machine Learning.Applied Sciences14, 24 (2024), 25 pages. doi:10.3390/app142411612

    work page doi:10.3390/app142411612 2024

  31. [31]

    Aditi Mishra, Bretho Danzy, Utkarsh Soni, Anjana Arunkumar, Jinbin Huang, Bum Chul Kwon, and Chris Bryan. 2025. PromptAid: Visual Prompt Exploration, Perturbation, Testing and Iteration for Large Language Models.IEEE Transactions on Visualization and Computer Graphics31 (2025), 6946–6962

    work page 2025

  32. [32]

    Munguia-Galeano, A

    F. Munguia-Galeano, A. Tan, and Z. Ji. 2023. Deep Reinforcement Learning With Explicit Context Representation.IEEE Transactions on Neural Networks and Learning Systems36 (2023), 419–432. doi:10.1109/tnnls.2023.3325633

    work page doi:10.1109/tnnls.2023.3325633 2023

  33. [33]

    Rupayan Neogy, Jonathan Zong, and Arvind Satyanarayan. 2020. Representing Real-Time Multi-User Collaboration in Visualizations. InProceedings of the IEEE Visualization Conference. IEEE Computer Society, Los Alamitos, CA, USA, 146–150. doi:10.1109/VIS47514.2020.00036

    work page doi:10.1109/vis47514.2020.00036 2020

  34. [34]

    Andrea Passerini, Aryo Gema, Pasquale Minervini, Burcu Sayin, and Katya Tentori. 2024. Fostering effective hybrid human-LLM reasoning and decision making.Frontiers Artificial Intelligence7 (2024), 10 pages. doi:10.3389/FRAI.2024.1464690

    work page doi:10.3389/frai.2024.1464690 2024

  35. [35]

    Pierce, Herbert H

    Robert Earl Patterson, Byron J. Pierce, Herbert H. Bell, and Gary Klein. 2010. Implicit Learning, Tacit Knowledge, Expertise Development, and Naturalistic Decision Making.Journal of Cognitive Engineering and Decision Making4, 4 (2010), 289–303. doi:10.1177/155534341000400403

    work page doi:10.1177/155534341000400403 2010

  36. [36]

    G. Peng, H. Wang, H. Zhang, Y. Zhao, and A. Johnson. 2017. A collaborative system for capturing and reusing in-context design knowledge with an integrated representation model.Advanced Engineering Informatics33 (2017), 314–329. doi:10.1016/j.aei.2016.12.007

    work page doi:10.1016/j.aei.2016.12.007 2017

  37. [37]

    2022.Human-Centered AI

    Ben Shneiderman. 2022.Human-Centered AI. Oxford University Press, Oxford, UK

    work page 2022

  38. [38]

    Kumar Shridhar, Koustuv Sinha, Andrew Cohen, Tianlu Wang, Ping Yu, Ram Pasunuru, Mrinmaya Sachan, Jason Weston, and Asli Celikyilmaz

    work page

  39. [39]

    arXiv:2311.07961 [cs.CL]

    The ART of LLM Refinement: Ask, Refine, and Trust. arXiv:2311.07961 [cs.CL]

    work page arXiv

  40. [40]

    Arjun Srinivasan and Vidya Setlur. 2021. Snowy: Recommending Utterances for Conversational Visual Analysis. InProceedings of the ACM Symposium on User Interface Software and Technology. ACM, New York, NY, USA, 864–880. doi:10.1145/3472749.3474792

    work page doi:10.1145/3472749.3474792 2021

  41. [41]

    Roy Turner. 1998. Context-mediated behavior for intelligent agents.International Journal of Human-Computer Studies48, 3 (1998), 307–330. doi:10.1006/ijhc.1997.0173

    work page doi:10.1006/ijhc.1997.0173 1998

  42. [42]

    Cem Celal Tutum, Suhaib Abdulquddos, and Risto Miikkulainen. 2021. Generalization of Agent Behavior through Explicit Representation of Context. InProceedings of the IEEE Conference on Games. IEEE Computer Society, Los Alamitos, CA, USA, 1–7. doi:10.1109/COG52621.2021.9619141

    work page doi:10.1109/cog52621.2021.9619141 2021

  43. [43]

    Skov, and Jesper Kjeldskov

    Niels van Berkel, Mikael B. Skov, and Jesper Kjeldskov. 2021. Human-AI interaction: intermittent, continuous, and proactive.Interactions28, 6 (2021), 67–71. doi:10.1145/3486941

    work page doi:10.1145/3486941 2021

  44. [44]

    Chenglong Wang, John Thompson, and Bongshin Lee. 2024. Data Formulator: AI-Powered Concept-Driven Visualization Authoring.IEEE Transactions on Visualization and Computer Graphics30, 1 (2024), 1128–1138. doi:10.1109/TVCG.2023.3326585

    work page doi:10.1109/tvcg.2023.3326585 2024

  45. [45]

    M. Weck, I. Humala, P. Tamminen, and F. Ferreira. 2021. Knowledge management visualisation in regional innovation system collaborative decision-making.Management Decision60, 4 (2021), 1017–1038. doi:10.1108/md-01-2021-0064 Manuscript submitted to ACM 34 Wolter et al

    work page doi:10.1108/md-01-2021-0064 2021

  46. [46]

    Luoxuan Weng, Xingbo Wang, Junyu Lu, Yingchaojie Feng, Yihan Liu, Haozhe Feng, Danqing Huang, and Wei Chen. 2025. InsightLens: Augmenting LLM-Powered Data Analysis With Interactive Insight Management and Navigation.IEEE Transactions on Visualization and Computer Graphics31, 6 (2025), 3719–3732. doi:10.1109/TVCG.2025.3567131

    work page doi:10.1109/tvcg.2025.3567131 2025

  47. [47]

    Anton Wolter, Georgios Vidalakis, Michael Yu, Ankit Grover, and Vaishali Dhanoa. 2025. Multi-Agent Data Visualization and Narrative Generation. arXiv:2509.00481 [cs.AI]

    work page arXiv 2025

  48. [48]

    Junde Wu, Jiayuan Zhu, Yuyuan Liu, Min Xu, and Yueming Jin. 2025. Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools. InProceedings of the Annual Meeting of the Association for Computational Linguistics. ACL, Stroudsburg, PA, USA, 28489–28503

    work page 2025

  49. [49]

    Fumeng Yang. 2021. How Do Visual Explanations Foster End Users’ Appropriate Trust in Machine Learning? doi:10.1145/3377325.3377480

    work page doi:10.1145/3377325.3377480 2021

  50. [50]

    Raquib Bin Yousuf, Nicholas Defelice, Mandar Sharma, Shengzhe Xu, and Naren Ramakrishnan. 2024. LLM Augmentations to support Analytical Reasoning over Multiple Documents. InProceedings of the IEEE Conference on Big Data. IEEE Computer Society, Los Alamitos, CA, USA, 1892–1901. doi:10.1109/BIGDATA62323.2024.10826114

    work page doi:10.1109/bigdata62323.2024.10826114 2024

  51. [51]

    J. D. Zamfirescu-Pereira, Richmond Y. Wong, Bjoern Hartmann, and Qian Yang. 2023. Why Johnny Can’t Prompt: How Non-AI Experts Try (and Fail) to Design LLM Prompts. InProceedings of the ACM Conference on Human Factors in Computing Systems. ACM, New York, NY, USA, 437:1–437:21. doi:10.1145/3544548.3581388

    work page doi:10.1145/3544548.3581388 2023

  52. [52]

    Yusen Zhang, Ruoxi Sun, Yanfei Chen, Tomas Pfister, Rui Zhang, and Sercan Ö. Arik. 2024. Chain of Agents: Large Language Models Collaborating on Long-Context Tasks.CoRRabs/2406.02818 (2024), 19 pages. arXiv:2406.02818 doi:10.48550/ARXIV.2406.02818

    work page doi:10.48550/arxiv.2406.02818 2024

  53. [53]

    Haiyan Zhao, Hanjie Chen, Fan Yang, Ninghao Liu, Huiqi Deng, Hengyi Cai, Shuaiqiang Wang, Dawei Yin, and Mengnan Du. 2024. Explainability for Large Language Models: A Survey.ACM Transactions on Intelligent Systems and Technology15, 2 (2024), 1–38

    work page 2024

  54. [54]

    Yongchao Zhou, Andrei Ioan Muresanu, Ziwen Han, Keiran Paster, Silviu Pitis, Harris Chan, and Jimmy Ba. 2023. Large Language Models Are Human-Level Prompt Engineers. InInternational Conference on Learning Representations (ICLR). https://openreview.net/forum?id=92gvk82DE-

    work page 2023

  55. [55]

    Yufan Zhuang, Xiaodong Yu, Jialian Wu, Ximeng Sun, Ze Wang, Jiang Liu, Yusheng Su, Jingbo Shang, Zicheng Liu, and Emad Barsoum. 2025. Self-Taught Agentic Long Context Understanding. InProceedings of the Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, Stroudsburg, PA, USA, 5525–5537. https://aclan...

    work page 2025