pith. machine review for the scientific record. sign in

arxiv: 2604.16263 · v1 · submitted 2026-04-17 · 💻 cs.RO

Recognition: unknown

Semantic Area Graph Reasoning for Multi-Robot Language-Guided Search

Ruiyang Wang , Hao-Lun Hsu , Jiwoo Kim , Miroslav Pajic
Authors on Pith no claims yet

Pith reviewed 2026-05-10 08:13 UTC · model grok-4.3

classification 💻 cs.RO
keywords multi-robot coordinationsemantic searchlanguage-guided explorationlarge language modelssemantic area graphindoor environmentsfrontier planning
0
0 comments X

The pith

A hierarchical framework lets large language models coordinate multi-robot semantic search by reasoning over an incrementally built graph of room instances and frontiers.

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

The paper introduces Semantic Area Graph Reasoning (SAGR), a method that converts a semantic occupancy map into a compact graph encoding rooms, their connections, available frontiers, and robot locations. This graph serves as the input for an LLM to make high-level decisions about which robots should search which room types based on task context, while deterministic planners handle movement and coverage inside the assigned rooms. Across 100 test scenarios drawn from the Habitat-Matterport3D dataset, the approach matches the coverage speed of leading geometric exploration methods yet reduces the time needed to locate semantically specified targets, with gains reaching 18.8 percent in larger environments.

Core claim

SAGR incrementally constructs a semantic area graph from a semantic occupancy map, encoding room instances, connectivity, frontier availability, and robot states into a compact task-relevant representation for LLM reasoning. The LLM performs high-level semantic room assignment based on spatial structure and task context, while deterministic frontier planning and local navigation handle geometric execution within assigned rooms.

What carries the argument

The semantic area graph, a structured abstraction that encodes room instances, their connectivity, frontier availability, and robot states to enable LLM-based high-level assignment.

Load-bearing premise

The approach assumes a semantic occupancy map can be built accurately enough to identify distinct room instances and types, and that the LLM will assign rooms correctly from spatial structure and task context without errors or hallucinations.

What would settle it

An experiment in which the semantic occupancy map misclassifies room types or the LLM assigns robots to semantically incorrect rooms, resulting in search efficiency no better than or worse than pure frontier-based baselines.

Figures

Figures reproduced from arXiv: 2604.16263 by Hao-Lun Hsu, Jiwoo Kim, Miroslav Pajic, Ruiyang Wang.

Figure 1
Figure 1. Figure 1: Overview of the SAGR framework. The semantic reasoning layer [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example semantic area graph constructed from the observed semantic [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example environment from the HM3D dataset, showing an indoor [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
read the original abstract

Coordinating multi-robot systems (MRS) to search in unknown environments is particularly challenging for tasks that require semantic reasoning beyond geometric exploration. Classical coordination strategies rely on frontier coverage or information gain and cannot incorporate high-level task intent, such as searching for objects associated with specific room types. We propose \textit{Semantic Area Graph Reasoning} (SAGR), a hierarchical framework that enables Large Language Models (LLMs) to coordinate multi-robot exploration and semantic search through a structured semantic-topological abstraction of the environment. SAGR incrementally constructs a semantic area graph from a semantic occupancy map, encoding room instances, connectivity, frontier availability, and robot states into a compact task-relevant representation for LLM reasoning. The LLM performs high-level semantic room assignment based on spatial structure and task context, while deterministic frontier planning and local navigation handle geometric execution within assigned rooms. Experiments on the Habitat-Matterport3D dataset across 100 scenarios show that SAGR remains competitive with state-of-the-art exploration methods while consistently improving semantic target search efficiency, with up to 18.8\% in large environments. These results highlight the value of structured semantic abstractions as an effective interface between LLM-based reasoning and multi-robot coordination in complex indoor environments.

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 paper proposes Semantic Area Graph Reasoning (SAGR), a hierarchical framework for multi-robot language-guided semantic search in unknown environments. It incrementally builds a semantic area graph from a semantic occupancy map that encodes room instances, connectivity, frontiers, and robot states; an LLM performs high-level room assignment based on spatial structure and task context, while deterministic frontier planning and local navigation handle geometric execution. Experiments on the Habitat-Matterport3D dataset across 100 scenarios show SAGR remains competitive with state-of-the-art exploration methods while improving semantic target search efficiency by up to 18.8% in large environments.

Significance. If the semantic-map and LLM components prove robust, the work provides a practical interface between high-level semantic reasoning and multi-robot geometric planning, addressing a gap in language-guided MRS tasks. The compact graph abstraction and separation of LLM reasoning from deterministic execution are clear strengths that could generalize beyond the tested indoor scenarios.

major comments (2)
  1. [§4] §4 (Experiments): The headline claim of up to 18.8% search-efficiency improvement on 100 Habitat-Matterport3D scenarios is load-bearing, yet the manuscript provides no quantitative evaluation of semantic occupancy map fidelity (room instance/type segmentation accuracy) or LLM room-assignment error rates under incremental multi-robot partial observability. Without these, it is impossible to confirm that the reported gains arise from the proposed semantic reasoning rather than from the underlying geometric planners.
  2. [§3.2] §3.2 (Semantic Area Graph Construction): The framework delegates geometric execution to deterministic planners once rooms are assigned, so any systematic error in map construction or LLM output directly removes the claimed advantage; however, no ablation or sensitivity analysis on mapping noise, frontier detection errors, or LLM prompt variations is reported to bound the robustness of the central efficiency result.
minor comments (2)
  1. [Abstract] Abstract: the efficiency claim ends with 'with up to 18.8% in large environments'; the missing noun ('improvement') reduces readability.
  2. [§3] Notation: the distinction between 'semantic area graph' and 'semantic occupancy map' is introduced without an explicit diagram or equation linking the two representations, making the incremental construction harder to follow.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which highlight important aspects of experimental validation. We address each major comment below and will revise the manuscript to incorporate additional quantitative analyses of the semantic components.

read point-by-point responses

  1. Referee: The headline claim of up to 18.8% search-efficiency improvement on 100 Habitat-Matterport3D scenarios is load-bearing, yet the manuscript provides no quantitative evaluation of semantic occupancy map fidelity (room instance/type segmentation accuracy) or LLM room-assignment error rates under incremental multi-robot partial observability. Without these, it is impossible to confirm that the reported gains arise from the proposed semantic reasoning rather than from the underlying geometric planners.

    Authors: We agree that direct quantitative metrics on semantic occupancy map fidelity and LLM assignment accuracy under partial observability would strengthen the attribution of gains to the semantic reasoning layer. The current evaluation focuses on end-to-end semantic search efficiency across 100 scenarios while remaining competitive with SOTA geometric exploration baselines, which indirectly supports the contribution of the LLM-based room assignment. To address the concern explicitly, the revised manuscript will add: (i) room instance and type segmentation accuracy metrics for the semantic occupancy map, and (ii) LLM room-assignment accuracy and error rates measured under incremental multi-robot partial observability. These additions will help isolate the semantic component's role. revision: yes

  2. Referee: The framework delegates geometric execution to deterministic planners once rooms are assigned, so any systematic error in map construction or LLM output directly removes the claimed advantage; however, no ablation or sensitivity analysis on mapping noise, frontier detection errors, or LLM prompt variations is reported to bound the robustness of the central efficiency result.

    Authors: We acknowledge that explicit sensitivity and ablation studies on mapping noise, frontier detection errors, and LLM prompt variations are valuable for bounding robustness, especially given the delegation to deterministic planners. The reported results use realistic incremental mapping on Habitat-Matterport3D data, but the manuscript does not include dedicated ablations. In the revision we will add sensitivity analyses, including controlled injection of mapping noise, variations in frontier detection, and alternative LLM prompts, to quantify their impact on the observed efficiency gains. revision: yes

Circularity Check

0 steps flagged

No circularity: new framework validated on external dataset benchmarks

full rationale

The paper introduces SAGR as a hierarchical construction that builds a semantic area graph incrementally from a semantic occupancy map, delegates high-level room assignment to an LLM, and uses deterministic planners for geometric execution. The efficiency claims (up to 18.8% improvement) are supported by direct comparison against state-of-the-art exploration baselines on the external Habitat-Matterport3D dataset across 100 scenarios. No equations, fitted parameters renamed as predictions, or self-citation chains appear in the provided text that would reduce the reported gains to internal definitions or prior author results by construction. The derivation is therefore self-contained against independent benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on two domain assumptions about map quality and LLM reliability plus one new abstraction; no free parameters or invented physical entities are described in the abstract.

axioms (2)
  • domain assumption Semantic occupancy maps can be constructed to reliably identify room instances, types, and connectivity.
    The pipeline begins with a semantic occupancy map as input to graph construction.
  • domain assumption Large language models can perform accurate high-level semantic room assignment from the graph structure and task context.
    This is the core reasoning step that separates SAGR from purely geometric methods.
invented entities (1)
  • Semantic Area Graph no independent evidence
    purpose: Compact representation that encodes room instances, connectivity, frontier availability, and robot states for LLM reasoning.
    New structured abstraction introduced to bridge semantic reasoning and geometric execution.

pith-pipeline@v0.9.0 · 5519 in / 1377 out tokens · 34453 ms · 2026-05-10T08:13:02.417887+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

42 extracted references · 4 canonical work pages · 2 internal anchors

  1. [1]

    A formal analysis and taxonomy of task allocation in multi-robot systems,

    B. P. Gerkey and M. J. Matari ´c, “A formal analysis and taxonomy of task allocation in multi-robot systems,”The International journal of robotics research, vol. 23, no. 9, pp. 939–954, 2004

    2004

  2. [2]

    A comprehensive taxonomy for multi-robot task allocation,

    G. A. Korsah, A. Stentz, and M. B. Dias, “A comprehensive taxonomy for multi-robot task allocation,”The International Journal of Robotics Research, vol. 32, pp. 1495 – 1512, 2013. [Online]. Available: https://api.semanticscholar.org/CorpusID:12515065

    2013

  3. [3]

    Coordinated multi-robot exploration,

    W. Burgard, M. Moors, C. Stachniss, and F. E. Schneider, “Coordinated multi-robot exploration,”IEEE Transactions on robotics, vol. 21, no. 3, pp. 376–386, 2005

    2005

  4. [4]

    Optimization techniques for multi-robot task allocation problems: Review on the state- of-the-art,

    H. Chakraa, F. Gu ´erin, E. Leclercq, and D. Lefebvre, “Optimization techniques for multi-robot task allocation problems: Review on the state- of-the-art,”Robotics Auton. Syst., vol. 168, p. 104492, 2023. [Online]. Available: https://api.semanticscholar.org/CorpusID:260122837

    2023

  5. [5]

    Market-based multirobot coordination: A survey and analysis,

    M. B. Dias, R. Zlot, N. Kalra, and A. Stentz, “Market-based multirobot coordination: A survey and analysis,”Proceedings of the IEEE, vol. 94, no. 7, pp. 1257–1270, 2006

    2006

  6. [6]

    Market-based multirobot coordination for complex tasks,

    R. Zlot and A. Stentz, “Market-based multirobot coordination for complex tasks,”The International Journal of Robotics Research, vol. 25, no. 1, pp. 73–101, 2006

    2006

  7. [7]

    A frontier-based approach for autonomous exploration,

    B. Yamauchi, “A frontier-based approach for autonomous exploration,” inProceedings 1997 IEEE Int. Symposium on Computational Intelli- gence in Robotics and Automation CIRA’97. ’Towards New Computa- tional Principles for Robotics and Automation’, 1997, pp. 146–151

    1997

  8. [8]

    Efficient autonomous exploration planning of large-scale 3-d environments,

    M. Selin, M. Tiger, D. Duberg, F. Heintz, and P. Jensfelt, “Efficient autonomous exploration planning of large-scale 3-d environments,” IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 1699–1706, 2019

    2019

  9. [9]

    Sensor-based multi-robot coverage control with spatial separation in unstructured environments,

    X. Wang, J. Xu, C. Gao, Y . Chen, J. Zhang, C. Wang, Y . Ding, and B. M. Chen, “Sensor-based multi-robot coverage control with spatial separation in unstructured environments,” in2024 IEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 10 623–10 629

    2024

  10. [10]

    Racer: Rapid collaborative exploration with a decentralized multi-uav system,

    B. Zhou, H. Xu, and S. Shen, “Racer: Rapid collaborative exploration with a decentralized multi-uav system,”IEEE Transactions on Robotics, vol. 39, no. 3, pp. 1816–1835, 2023

    2023

  11. [11]

    Semantics for robotic mapping, perception and interaction: A survey,

    S. Garg, N. Sunderhauf, F. Dayoub, D. Morrison, A. Cosgun, G. Carneiro, Q. Wu, T.-J. Chin, I. D. Reid, S. Gould, P. Corke, and M. Milford, “Semantics for robotic mapping, perception and interaction: A survey,”Foundations and Trends® in Robotics 8, vol. 1-2, 2020. 8 JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021

    2020

  12. [12]

    Object goal navigation using goal-oriented semantic exploration,

    D. S. Chaplot, D. P. Gandhi, A. Gupta, and R. R. Salakhutdinov, “Object goal navigation using goal-oriented semantic exploration,”Advances in Neural Information Processing Systems, vol. 33, pp. 4247–4258, 2020

    2020

  13. [13]

    Navigating to objects in the real world,

    T. Gervet, S. Chintala, D. Batra, J. Malik, and D. S. Chaplot, “Navigating to objects in the real world,”Science Robotics, vol. 8, 2022

    2022

  14. [14]

    Do As I Can, Not As I Say: Grounding Language in Robotic Affordances

    M. Ahn, A. Brohan, N. Brown, Y . Chebotar, O. Cortes, B. David, C. Finn, C. Fu, K. Gopalakrishnan, K. Hausmanet al., “Do as i can, not as i say: Grounding language in robotic affordances,”arXiv preprint arXiv:2204.01691, 2022

    work page internal anchor Pith review arXiv 2022

  15. [15]

    Vima: Robot manipulation with multimodal prompts,

    Y . Jiang, A. Gupta, Z. Zhang, G. Wang, Y . Dou, Y . Chen, L. Fei-Fei, A. Anandkumar, Y . Zhu, and L. Fan, “Vima: Robot manipulation with multimodal prompts,” 2023

    2023

  16. [16]

    Rt-2: Vision-language-action models transfer web knowledge to robotic control,

    B. Zitkovich, T. Yu, S. Xu, P. Xu, T. Xiao, F. Xia, J. Wu, P. Wohlhart, S. Welker, A. Wahidet al., “Rt-2: Vision-language-action models transfer web knowledge to robotic control,” inConference on Robot Learning. PMLR, 2023, pp. 2165–2183

    2023

  17. [17]

    Co-navgpt: Multi-robot co- operative visual semantic navigation using large language models.arXiv preprint arXiv:2310.07937, 2023a

    B. Yu, H. Kasaei, and M. Cao, “Co-navgpt: Multi-robot cooperative visual semantic navigation using large language models,”arXiv preprint arXiv:2310.07937, 2023

    work page arXiv 2023

  18. [18]

    Enhancing multi- robot semantic navigation through multimodal chain-of-thought score collaboration,

    Z. Shen, H. Luo, K. Chen, F. Lv, and T. Li, “Enhancing multi- robot semantic navigation through multimodal chain-of-thought score collaboration,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 14, 2025, pp. 14 664–14 672

    2025

  19. [19]

    Comres- vlm: Coordinated multi-robot exploration and search using vision lan- guage models,

    R. Wang, H.-L. Hsu, D. Hunt, J. Kim, S. Luo, and M. Pajic, “Comres- vlm: Coordinated multi-robot exploration and search using vision lan- guage models,” 2025

    2025

  20. [20]

    Habitat-Matterport 3D Dataset (HM3D): 1000 Large-scale 3D Environments for Embodied AI

    S. K. Ramakrishnan, A. Gokaslan, E. Wijmans, O. Maksymets, A. Clegg, J. M. Turner, E. Undersander, W. Galuba, A. Westbury, A. X. Chang, M. Savva, Y . Zhao, and D. Batra, “Habitat-matterport 3d dataset (HM3d): 1000 large-scale 3d environments for embodied AI,” inThirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track,...

    work page internal anchor Pith review arXiv 2021

  21. [21]

    Information based adaptive robotic exploration,

    F. Bourgault, A. A. Makarenko, S. B. Williams, B. Grocholsky, and H. F. Durrant-Whyte, “Information based adaptive robotic exploration,” inIEEE/RSJ international conference on intelligent robots and systems, vol. 1. IEEE, 2002, pp. 540–545

    2002

  22. [22]

    Exploration with active loop-closing for fastslam,

    C. Stachniss, D. Hahnel, and W. Burgard, “Exploration with active loop-closing for fastslam,” in2004 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)(IEEE Cat. No. 04CH37566), vol. 2. IEEE, 2004, pp. 1505–1510

    2004

  23. [23]

    Information-theoretic mapping using cauchy-schwarz quadratic mutual information,

    B. Charrow, S. Liu, V . Kumar, and N. Michael, “Information-theoretic mapping using cauchy-schwarz quadratic mutual information,” in2015 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2015, pp. 4791–4798

    2015

  24. [24]

    Receding horizon

    A. Bircher, M. Kamel, K. Alexis, H. Oleynikova, and R. Siegwart, “Receding horizon” next-best-view” planner for 3d exploration,” in 2016 IEEE international conference on robotics and automation (ICRA). IEEE, 2016, pp. 1462–1468

    2016

  25. [25]

    The hungarian method for the assignment problem,

    H. W. Kuhn, “The hungarian method for the assignment problem,”Naval research logistics quarterly, vol. 2, no. 1-2, pp. 83–97, 1955

    1955

  26. [26]

    Multi-hierarchical semantic maps for mobile robotics,

    C. Galindo, A. Saffiotti, S. Coradeschi, P. Buschka, J.-A. Fernandez- Madrigal, and J. Gonz´alez, “Multi-hierarchical semantic maps for mobile robotics,” in2005 IEEE/RSJ international conference on intelligent robots and systems. IEEE, 2005, pp. 2278–2283

    2005

  27. [27]

    Multi- modal semantic place classification,

    A. Pronobis, O. Martinez Mozos, B. Caputo, and P. Jensfelt, “Multi- modal semantic place classification,”The International Journal of Robotics Research, vol. 29, no. 2-3, pp. 298–320, 2010

    2010

  28. [28]

    Semantic mapping for mobile robotics tasks: A survey,

    I. Kostavelis and A. Gasteratos, “Semantic mapping for mobile robotics tasks: A survey,”Robotics and Autonomous Systems, vol. 66, pp. 86– 103, 2015

    2015

  29. [29]

    Semantic information for robot navigation: A survey,

    J. Crespo, J. C. Castillo, O. M. Mozos, and R. Barber, “Semantic information for robot navigation: A survey,”Applied Sciences, vol. 10, no. 2, p. 497, 2020

    2020

  30. [30]

    Proba- bilistic data association for semantic slam,

    S. L. Bowman, N. Atanasov, K. Daniilidis, and G. J. Pappas, “Proba- bilistic data association for semantic slam,” in2017 IEEE international conference on robotics and automation (ICRA). IEEE, 2017, pp. 1722– 1729

    2017

  31. [31]

    Semanticfu- sion: Dense 3d semantic mapping with convolutional neural networks,

    J. McCormac, A. Handa, A. Davison, and S. Leutenegger, “Semanticfu- sion: Dense 3d semantic mapping with convolutional neural networks,” in2017 IEEE International Conference on Robotics and automation (ICRA). IEEE, 2017, pp. 4628–4635

    2017

  32. [32]

    Kimera: an open- source library for real-time metric-semantic localization and mapping,

    A. Rosinol, M. Abate, Y . Chang, and L. Carlone, “Kimera: an open- source library for real-time metric-semantic localization and mapping,” in2020 IEEE international conference on robotics and automation (ICRA). IEEE, 2020, pp. 1689–1696

    2020

  33. [33]

    3d dynamic scene graphs: Actionable spatial perception with places, objects, and humans,

    A. Rosinol, A. Gupta, M. Abate, J. Shi, and L. Carlone, “3d dynamic scene graphs: Actionable spatial perception with places, objects, and humans,”arXiv preprint arXiv:2002.06289, 2020

    work page arXiv 2002

  34. [34]

    Robohop: Segment-based topological map represen- tation for open-world visual navigation,

    S. Garg, K. Rana, M. Hosseinzadeh, L. Mares, N. S ¨underhauf, F. Day- oub, and I. Reid, “Robohop: Segment-based topological map represen- tation for open-world visual navigation,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 4090–4097

    2024

  35. [35]

    Palm-e: An embodied multimodal language model,

    D. Driess, F. Xia, M. S. M. Sajjadi, C. Lynch, A. Chowdhery, B. Ichter, A. Wahid, J. Tompson, Q. H. Vuong, T. Yu, W. Huang, Y . Chebotar, P. Sermanet, D. Duckworth, S. Levine, V . Vanhoucke, K. Hausman, M. Toussaint, K. Greff, A. Zeng, I. Mordatch, and P. R. Florence, “Palm-e: An embodied multimodal language model,” inInternational Conference on Machine L...

    2023

  36. [36]

    Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments,

    P. Anderson, Q. Wu, D. Teney, J. Bruce, M. Johnson, N. S ¨underhauf, I. D. Reid, S. Gould, and A. van den Hengel, “Vision-and-language navigation: Interpreting visually-grounded navigation instructions in real environments,”2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3674–3683, 2018

    2018

  37. [37]

    Vision-and- language navigation: A survey of tasks, methods, and future directions,

    J. Gu, E. Stefani, Q. Wu, J. Thomason, and X. E. Wang, “Vision-and- language navigation: A survey of tasks, methods, and future directions,” inAnnual Meeting of the Association for Computational Linguistics, 2022

    2022

  38. [38]

    Language models as zero-shot planners: Extracting actionable knowledge for embodied agents,

    W. Huang, P. Abbeel, D. Pathak, and I. Mordatch, “Language models as zero-shot planners: Extracting actionable knowledge for embodied agents,”International conference on machine learning (ICML), 2022

    2022

  39. [39]

    Code as policies: Language model programs for embodied control,

    J. Liang, W. Huang, F. Xia, P. Xu, K. Hausman, B. Ichter, P. Florence, and A. Zeng, “Code as policies: Language model programs for embodied control,” inIEEE International Conference on Robotics and Automation (ICRA), 2023

    2023

  40. [40]

    Places: A 10 million image database for scene recognition,

    B. Zhou, A. Lapedriza, A. Khosla, A. Oliva, and A. Torralba, “Places: A 10 million image database for scene recognition,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017

    2017

  41. [41]

    Coverage control for mobile sensing networks,

    J. Cortes, S. Martinez, T. Karatas, and F. Bullo, “Coverage control for mobile sensing networks,”IEEE Transactions on robotics and Automation, vol. 20, no. 2, pp. 243–255, 2004

    2004

  42. [42]

    Graph of thoughts: Solving elaborate problems with large language models,

    M. Besta, N. Blach, A. Kubicek, R. Gerstenberger, M. Podstawski, L. Gianinazzi, J. Gajda, T. Lehmann, H. Niewiadomski, P. Nyczyket al., “Graph of thoughts: Solving elaborate problems with large language models,” inProceedings of the AAAI conference on artificial intelligence, vol. 38, no. 16, 2024, pp. 17 682–17 690

    2024