Pith. sign in

REVIEW 4 major objections 7 minor 40 references

AdaptBot: Combining LLM with Knowledge Graphs and Human Input for Generic-to-Specific Task Decomposition and Knowledge Refinement

T0 review · 4 major / 7 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read An embodied agent can adapt to new tasks by combining a large language model's generic plans with knowledge-graph repair and occasional human corrections, and this loop substantially outperforms the language model alone in simulation.

desk verdict A sensible LLM+KG+human-in-the-loop integration whose reported gains are inflated by an evaluation that lets the human oracle both fix the plan and score the result. read the letter →

arxiv 2502.02067 v2 pith:7BNIRZE2 submitted 2025-02-04 cs.RO cs.AIcs.CLcs.LG

Shivam Singh , Karthik Swaminathan , Nabanita Dash , Ramandeep Singh , Snehasis Banerjee , Mohan Sridharan , Madhava Krishna This is my paper · ORCID
classification cs.ROcs.AIcs.CLcs.LG
keywords LargeLanguageModelsKnowledgeGraphsHuman-in-the-LoopLearningTaskDecompositionEmbodiedAgentsCookingTasksCleaningRefinement
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to show that a robot can handle a new task without retraining by letting three components compensate for each other: a large language model produces a rough step-by-step plan, a knowledge graph of the specific environment filters and repairs that plan, and a human supplies the few missing details the graph does not know. The central claim is that the loop formed by these three pieces outperforms using the language model alone, and also outperforms the language model plus knowledge graph without human help. The authors support this with simulated cooking and cleaning experiments in which the full loop raises success rates from roughly 45% to 91% with one language model and from 25% to 92% with another, while reducing token usage. The payoff, if true, is a practical route to adaptable assistive robots that accumulate knowledge incrementally and transparently rather than needing comprehensive pretraining.

What carries the argument

The load-bearing mechanism is a two-layer knowledge graph used as a filter-and-repair layer between the LLM and the physical actions: a state graph stores instances with their current attributes and locations, and an attribute graph stores class-level properties such as 'sliceable' or 'fryable', both encoded in the RDF triple format. The paper checks every LLM-proposed action and object against these graphs with SPARQL queries; mismatches are repaired by direct word matching over hypernyms and hyponyms, or flagged as unknown. The second half of the machinery is a feedback controller: unresolved unknowns or execution errors send a feedback prompt back to the LLM up to a fixed number of times, and only then draw on a human, whose corrections update the graphs. This explicit triple-based update is what makes the knowledge growth transparent and reusable across subsequent tasks.

What would settle it

Run the full framework on the same 30 cooking tasks but replace the human oracle with a random-answer script and score the executions with evaluators who do not know which configuration produced each trace; if success rates stay near 91–92%, human accuracy is not load-bearing, whereas a collapse toward the LLM-only baseline would confirm that the reported gains rely on correct human input.

Watch

Extended reading notes

Core claim

The paper's discovery is that the bottleneck in LLM-based robot task execution is not the plan but the fit between the plan and the robot's actual world, and that this fit can be achieved by explicit knowledge repair. An LLM is prompted with an in-context example to produce a step-by-step action sequence for a new task. That sequence is checked against a knowledge graph stored as RDF (a standard triple-based graph format) representing states and attributes; known objects and actions pass, unknown ones are replaced with hypernym or hyponym matches from the graph, and anything that cannot be resolved triggers re-prompting of the LLM with feedback. If mismatches or execution errors persist beyond a threshold, the robot asks a human, uses the answer to correct the action, and writes the new object, attribute, or state into the knowledge graph as explicit triples. The reported result is that this combined loop succeeds far more often than the LLM alone, and that the improvement is tied to the knowledge expansion: for example, adding a single missing object such as a mopping cloth to the graph produced a 31–56% rise in success on cleaning tasks.

Load-bearing premise

The system's reported success rates depend on humans giving correct answers when asked and on the evaluators of the executions being impartial, so the framework alone cannot reproduce those numbers if the human oracle is wrong or the scoring is biased.

Editorial extensions

If this is right

  • If the reported results hold, an assistive robot can take on a new task class after only a handful of human-supplied knowledge insertions, with no retraining or large labeled datasets.
  • Because the knowledge graph catches and repairs errors before re-prompting, the framework uses fewer tokens than the LLM-only baseline while succeeding more often.
  • The framework transfers from cooking to cleaning and clearing tasks, so the same architecture supports qualitatively different task families.
  • As the graph expands, the need for human intervention decreases, so the system becomes progressively more autonomous on the tasks it has seen.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper assumes human answers are accurate; a natural stress test would be to inject occasional incorrect answers and measure how quickly errors propagate through the knowledge graph.
  • Because all knowledge updates are explicit RDF triples, the learned knowledge from one task class could in principle be audited or transferred to another domain, a possibility the paper leaves open.
  • The current repair step uses direct word matching, which suggests a probable failure mode when the LLM and the graph use different phrasing; an embedding-based matcher might extend the approach to paraphrase-heavy inputs.
  • The success metric comes from human evaluators who were not necessarily blinded to which configuration produced each execution; a follow-up with blinded scoring would clarify how much of the reported gain is due to the loop itself.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes AdaptBot, a framework that combines LLM-generated generic action sequences with a Knowledge Graph encoding domain-specific prior knowledge and with human-in-the-loop feedback. The LLM first produces a high-level action sequence for a task; the KG is used to refine the sequence by replacing unknown or inapplicable actions, objects, and attributes; unresolved mismatches or execution errors trigger re-prompting of the LLM and, beyond a threshold, solicitation of human input that expands or corrects the KG. The system is evaluated on 30 cooking tasks sampled from Recipe1M+ and 12 cleaning/clearing tasks in simulation, comparing three configurations: LLM only, LLM+KG, and LLM+KG+Human. Tables I and II report higher success rates, lower token usage, and higher ingredient overlap for the full configuration, and the paper claims that the interplay between LLM, KG, and human input leads to substantial performance gains and enables adaptation to new task classes through incremental knowledge refinement.

Significance. If the reported results were properly supported, the paper would make a useful contribution to robot task planning: the idea of using an LLM for generic decomposition and a KG for transparent, incremental refinement is attractive, and the authors include an explicit algorithm (Algorithm 1), transparent knowledge-update equations, linked trials across baselines, and two task domains with two LLMs. The main limitation is that the experimental protocol as described cannot separate the contribution of the framework from the contribution of the human oracle: the human can supply missing actions and object attributes during execution, and human evaluators then score the executed outputs. The observed gains in Tables I and II may therefore be due to the oracle rather than to the LLM+KG mechanism. The core architectural idea is defensible, but the evidence presented in support of the central claim needs re-scoping or a substantially stronger evaluation protocol.

major comments (4)
  1. [Section 4.1.2 and Algorithm 1, lines 14-18] The evaluation conflates the framework's adaptation with the human oracle. In the LLM+KG+Human arm, unresolved mismatches are resolved by ask_human, which can supply missing action names, object attributes, and existence judgments, and Section 4.1.2 states that this human input is 'assumed to be accurate.' The executed plan therefore contains human-authored steps, but Tables I and II report only final aggregated success rates; no quantity is reported for how many final action steps were human-supplied, nor for which tasks human intervention occurred. Under this protocol, a system that simply asked the human to complete the plan whenever it was stuck would receive high scores, so the reported jumps (e.g., GPT-4o cooking from 45.2% to 91.14%) cannot be read as evidence that the LLM+KG component is doing the adaptation. Please report autonomous-only performance (e.g., performance before any human correction, or performance on later trials after KG expansion), per-task human-intervention counts, and the fraction of final action steps authored by the human.
  2. [Section 4.1.4] The evaluation protocol does not establish that the scoring is fair across conditions. The manuscript says that 18 evaluators scored the execution outputs, that at least three raters scored each task, and that evaluators were 'not involved in the design,' but it does not state whether the evaluators were blinded to which condition produced each output, whether progress lines were shown for all conditions, or whether the people who supplied corrections during execution were the same population as the evaluators. Without blinding and separation, the success-rate scores may measure how closely the final output matches the evaluators' own expectations, which is especially problematic when the final output contains corrections supplied by a human. Please describe the evaluation interface, the blinding procedure, inter-rater agreement, and the relationship between the correcting participants and the scoring participants.
  3. [Section 4.1.4 and Tables I and II] No measures of uncertainty are reported for any of the headline numbers. Each table cell is a single average over 30 or 12 tasks with at least three ratings per task, and the claims of 'substantial improvement' in Section 4.2 are based on these single numbers without confidence intervals, standard deviations, or significance tests. Given the small number of tasks and the large reported differences, per-task score distributions, bootstrap confidence intervals, or pairwise significance tests are needed to determine whether the differences are robust; the paper should also report the value of Fmax used and any sensitivity analysis for this threshold, since it controls how often the human is queried.
  4. [Section 4.2, H2 and Table I] The token-use reduction is not clear evidence of efficiency gains from the KG. Because unresolved mismatches in the human arm are resolved by asking the human rather than by additional LLM feedback prompts, the lower average token count is a direct consequence of the protocol: the human's answer replaces the LLM calls that the baselines continue making. The paper should report token counts separately for the LLM-generated portion and for the human-query overhead (including the interactive attribute acquisition described in Section 3.4), and should avoid presenting the reduction as evidence of the KG's contribution without an ablation that holds the number of feedback rounds fixed across conditions.
minor comments (7)
  1. [Section 2] The sentence 'KGs have have been used to model prior knowledge' contains a duplicated word ('have have').
  2. [Section 3.1] The sentence 'We describe out framework's components below' should read 'our framework's components'.
  3. [Text following Algorithm 1] The phrase 'until the tasks is completed' should be corrected to 'until the task is completed.'
  4. [Section 3.4] In the definition of Scurrent, the second tuple is written as (sn, sn); it should presumably be (sn, vn) to match the first tuple and the surrounding notation.
  5. [Section 4.2, H2] The reported differences of 45.94 and 34.19 percentage points are labeled 'increase in success rate'; these are percentage-point differences relative to the baselines, not percentage increases, and should be stated as percentage-point gains to avoid ambiguity.
  6. [Section 4.1] The experimental section does not name the simulator or environment used for the cooking and cleaning tasks, although the introduction cites three candidate simulators; please specify the simulator, version, and task definitions so that the experiments are reproducible.
  7. [Section 4.1.5] The success-rate measure is described as 'computed based on the scores assigned by the human participants,' but the exact evaluation question, the rubric for the 0-20 scale, and the mapping from scores to binary success are not provided; please include the precise instructions given to evaluators.

Circularity Check

1 steps flagged · score 3.0 of 10

Reported HITL gains largely measure the human oracle: human-supplied corrections are assumed accurate, executed, and then scored by human evaluators.

  1. fitted input called prediction [Section 4.1.2 Baselines; Algorithm 1 lines 14–18; Section 4.1.4 Evaluation Strategy]
    "The LLM+ KG + human configuration represents our framework, in which unresolved mismatches are addressed using human input, which is assumed to be accurate; the other two configurations serve as baselines."

    Algorithm 1 shows that when unresolved mismatches persist, ask_human returns a refined action sequence (T′refined) that is then executed directly. Because that human input is explicitly assumed accurate, any error the LLM+KG components could not fix is replaced by a correct human answer. The success rate in Tables I–II is then obtained by human evaluators scoring the executed output of this human-corrected sequence. The claimed improvement of LLM+KG+Human over LLM/LLM+KG is thus partly forced by construction: the framework receives the very information needed to complete the task, and the evaluators score the result.

full rationale

This is an empirical systems paper with no mathematical derivation chain, so there is no proof-theoretic circularity. The only substantial circularity is in the evaluation of H2/H3: the human-in-the-loop configuration is allowed to query a human for the missing action/object/attribute, the human response is assumed accurate, and the resulting action sequence is executed and scored by human evaluators. Under this protocol a system that asked the human for every unresolved step would receive the human oracle's score, so the observed jumps (e.g., GPT-4o cooking success from 45.2% to 91.14% when human input is added) cannot be attributed to the LLM+KG adaptation mechanism. This is a real limitation but it is openly disclosed in the assumption of accurate human input, and the framework's token-reduction and KG-expansion results retain independent content. There is no load-bearing self-citation: the prior-work citations (FOON, Recipe1M+, ADaPT, etc.) are external and not used to force the central claim. The paper is best read as 'performance with an accurate human oracle in the loop' rather than as evidence that the LLM+KG components alone produce the reported gains. Overall circularity is moderate: the central HITL comparison is partially by construction, but the framework's autonomous components are still evaluated in the LLM and LLM+KG baselines, so the work is not wholly circular. Score 3.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The framework's central claims rest on the accuracy of human corrections, the executability of actions, the availability of lexical relationships, and the usefulness of LLM decomposition; none of these are verified by the experiments, and the human-oracle setup inflates the success-rate measure.

free parameters (1)
  • Fmax (feedback threshold)
    The algorithm's maximum number of re-prompt attempts before human consultation is a hyperparameter set by the designers; its value is not reported, yet it controls the division of labor between LLM re-prompting and human input.
assumptions (4)
  • domain assumption The agent can execute every action in the refined action sequence
    Stated in Section 3: 'The agent is assumed to be able to execute these actions.' Invalid if robot capabilities do not match the action vocabulary.
  • domain assumption Human input is accurate
    Section 4.1.2 states 'human input, which is assumed to be accurate.' The entire HITL improvement rests on this; inaccurate humans would reduce success rates.
  • domain assumption Hypernym/hyponym relationships are available for all relevant object and action terms
    Section 3.3 uses direct word matching with hypernyms/hyponyms for refinement, but the source of the lexical hierarchy is not specified.
  • domain assumption LLM provides a generic action sequence that is a useful starting point for the target task
    The framework relies on the LLM's decomposition being approximately correct; if the LLM output is entirely wrong, KG/human refinement may not recover. This assumption is central to H1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdaptBot: Combining LLM with Knowledge Graphs and Human Input for Generic-to-Specific Task Decomposition and Knowledge Refinement." pith.science (2026). https://pith.science/paper/7BNIRZE2

@misc{pith2026250202067,
  author       = {Pith},
  title        = {Pith review of: AdaptBot: Combining LLM with Knowledge Graphs and Human Input for Generic-to-Specific Task Decomposition and Knowledge Refinement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7BNIRZE2}},
  note         = {Machine review of arXiv:2502.02067}
}
read the original abstract

An embodied agent assisting humans is often asked to complete new tasks, and there may not be sufficient time or labeled examples to train the agent to perform these new tasks. Large Language Models (LLMs) trained on considerable knowledge across many domains can be used to predict a sequence of abstract actions for completing such tasks, although the agent may not be able to execute this sequence due to task-, agent-, or domain-specific constraints. Our framework addresses these challenges by leveraging the generic predictions provided by LLM and the prior domain knowledge encoded in a Knowledge Graph (KG), enabling an agent to quickly adapt to new tasks. The robot also solicits and uses human input as needed to refine its existing knowledge. Based on experimental evaluation in the context of cooking and cleaning tasks in simulation domains, we demonstrate that the interplay between LLM, KG, and human input leads to substantial performance gains compared with just using the LLM. Project website{\S}: https://sssshivvvv.github.io/adaptbot/

Figures

Figures reproduced from arXiv: 2502.02067 by the authors.

Figure 1
Figure 1. For any given task, an LLM provides a generic sequence of abstract actions [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Framework overview for cooking tasks: (a) Input Chain-of-Thought (COT) prompt contains target dish, available ingredients, and an example of input and output action [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Example of a node onion in Gk (top) and Gs (bottom). edges Ek ⊆ Ik × Pk × Vk represent class properties, e.g., (apple, sliceable, true) implies apples can be sliced. The available actions include moving, picking up, and putting down objects; using tools; cleaning, toggling, slic￾ing, stirring, and mopping*. Such a KG can be learned automatically based on information extracted from datasets or sensor streams. The fea… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Progress line [8] showing use of each ingredient when preparing an omelette. H1: Combining generic prediction of action sequences from LLMs with KG-based specific prior knowledge im￾proves performance compared with just LLMs. H2: Soliciting and using human feedback as …
Figure 5
Figure 5. Figure 5: 12 variants of tasks that involve the agent assisting with cleaning different [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 23 canonical work pages

  1. [1]

    Coppeliasim (formerly v-rep): a versatile and scalable robot simulation framework,

    E. Rohmer et al. , “Coppeliasim (formerly v-rep): a versatile and scalable robot simulation framework,” in Proc. of The Interna- tional Conference on Intelligent Robots and Systems (IROS) , 2013, www.coppeliarobotics.com

    work page 2013

  2. [2]

    Virtualhome: Simulating household activities via programs,

    X. Puig et al. , “Virtualhome: Simulating household activities via programs,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , June 2018

    work page 2018

  3. [3]

    Ai2-thor: An interactive 3d environment for visual ai,

    E. Kolve et al. , “Ai2-thor: An interactive 3d environment for visual ai,” 2022. [Online]. Available: https://arxiv.org/abs/1712.05474

    arXiv 2022

  4. [4]

    Decomposed prompting: A modular approach for solving complex tasks,

    T. Khot et al. , “Decomposed prompting: A modular approach for solving complex tasks,” 2023. [Online]. Available: https: //arxiv.org/abs/2210.02406

    arXiv 2023

  5. [5]

    Iterated decomposition: Improving science q&a by supervising reasoning processes,

    J. Reppert et al. , “Iterated decomposition: Improving science q&a by supervising reasoning processes,” 2023. [Online]. Available: https://arxiv.org/abs/2301.01751

    arXiv 2023

  6. [6]

    Delta: Decomposed efficient long-term robot task planning using large language models,

    Y . Liu et al. , “Delta: Decomposed efficient long-term robot task planning using large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2404.03275

    arXiv 2024

  7. [7]

    Approximate task tree retrieval in a knowledge network for robotic cooking,

    M. S. Sakib et al. , “Approximate task tree retrieval in a knowledge network for robotic cooking,” IEEE Robotics and Automation Letters , vol. 7, no. 4, pp. 11 492–11 499, 2022

    work page 2022

  8. [8]

    From cooking recipes to robot task trees– improving planning correctness and task efficiency by leveraging llms with a knowledge network,

    M. S. Sakib and Y . Sun, “From cooking recipes to robot task trees– improving planning correctness and task efficiency by leveraging llms with a knowledge network,” in 2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 12 704–12 711

    work page 2024

Show all 40 references
  1. [9]

    Gpt-4 technical report,

    OpenAI et al. , “Gpt-4 technical report,” 2024. [Online]. Available: https://arxiv.org/abs/2303.08774

    2024 arXiv

  2. [10]

    Gemma 2: Improving open language models at a practical size,

    G. Team et al. , “Gemma 2: Improving open language models at a practical size,” 2024. [Online]. Available: https://arxiv.org/abs/2408. 00118

    2024

  3. [11]

    The llama 3 herd of models,

    Dubey et al. , “The llama 3 herd of models,” 2024. [Online]. Available: https://arxiv.org/abs/2407.21783

    2024 arXiv

  4. [12]

    Learning task decomposition to assist humans in competitive programming,

    J. Wen et al. , “Learning task decomposition to assist humans in competitive programming,” arXiv preprint arXiv:2406.04604 , 2024

    2024 arXiv

  5. [13]

    Semantically aligned task decomposition in multi-agent reinforcement learning,

    W. Li et al., “Semantically aligned task decomposition in multi-agent reinforcement learning,” arXiv preprint arXiv:2305.10865 , 2023

    2023 arXiv

  6. [14]

    Auxiliary task update decomposition: The good, the bad and the neutral,

    L. M. Dery et al. , “Auxiliary task update decomposition: The good, the bad and the neutral,” 2021. [Online]. Available: https://arxiv.org/abs/2108.11346

    2021 arXiv

  7. [15]

    Taskbench: Benchmarking large language models for task automation,

    Y . Shen et al., “Taskbench: Benchmarking large language models for task automation,” arXiv preprint arXiv:2311.18760 , 2023

    2023 arXiv

  8. [16]

    Tdag: A multi-agent framework based on dy- namic task decomposition and agent generation,

    Y . Wang et al. , “Tdag: A multi-agent framework based on dy- namic task decomposition and agent generation,” arXiv preprint arXiv:2402.10178, 2024

    2024 arXiv

  9. [17]

    Semantic task planning for service robots in open worlds,

    G. Cui, W. Shuai, and X. Chen, “Semantic task planning for service robots in open worlds,” Future Internet, vol. 13, no. 2, p. 49, 2021

    2021

  10. [18]

    Adapt: As-needed decomposition and planning with language models,

    A. Prasad et al., “Adapt: As-needed decomposition and planning with language models,” arXiv preprint arXiv:2311.05772 , 2023

    2023 arXiv

  11. [19]

    A knowledge-based task planning approach for robot multi-task manipulation,

    D. Zheng et al., “A knowledge-based task planning approach for robot multi-task manipulation,” Complex & Intelligent Systems , vol. 10, 07 2023

    2023

  12. [20]

    Roboplanner: autonomous robotic action planning via knowledge graph queries,

    A. Kattepur and B. P, “Roboplanner: autonomous robotic action planning via knowledge graph queries,” in Proceedings of the 34th ACM/SIGAPP Symposium on Applied Computing , ser. SAC ’19. New York, NY , USA: Association for Computing Machinery, 2019, p. 953â ˘A¸ S956. [Online]. ...

    2019 doi

  13. [21]

    Towards robust one-shot task execution using knowl- edge graph embeddings,

    A. Daruna et al., “Towards robust one-shot task execution using knowl- edge graph embeddings,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) , 2021, pp. 11 118–11 124

    2021

  14. [22]

    Knowledge graphs as context sources for llm-based explanations of learning recommendations,

    H. Abu-Rasheed et al. , “Knowledge graphs as context sources for llm-based explanations of learning recommendations,” 2024. [Online]. Available: https://arxiv.org/abs/2403.03008

    2024 arXiv

  15. [23]

    Unifying large language models and knowledge graphs: A roadmap,

    S. Pan et al., “Unifying large language models and knowledge graphs: A roadmap,” IEEE Transactions on Knowledge and Data Engineering, 2024

    2024

  16. [24]

    Openfmnav: Towards open-set zero-shot object navigation via vision-language foundation models,

    Y . Kuang et al. , “Openfmnav: Towards open-set zero-shot object navigation via vision-language foundation models,” 2024. [Online]. Available: https://arxiv.org/abs/2402.10670

    2024 arXiv

  17. [25]

    Functional object-oriented network for manipulation learning,

    D. Paulius et al., “Functional object-oriented network for manipulation learning,” in 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2016, pp. 2655–2662

    2016

  18. [26]

    Robot task planning and situation handling in open worlds,

    Y . Ding et al. , “Robot task planning and situation handling in open worlds,” 2022. [Online]. Available: https://arxiv.org/abs/2210.01287

    2022 arXiv

  19. [27]

    Grounding llms for robot task planning using closed-loop state feedback,

    V . Bhat et al. , “Grounding llms for robot task planning using closed-loop state feedback,” 2024. [Online]. Available: https: //arxiv.org/abs/2402.08546

    2024

  20. [28]

    Task planning in robotics: an empirical comparison of pddl-and asp-based systems,

    Y .-q. Jiang et al., “Task planning in robotics: an empirical comparison of pddl-and asp-based systems,” Frontiers of Information Technology & Electronic Engineering , vol. 20, pp. 363–373, 2019

    2019

  21. [29]

    Recipe1m+: A dataset for learning cross-modal embeddings for cooking recipes and food images,

    J. Marın et al. , “Recipe1m+: A dataset for learning cross-modal embeddings for cooking recipes and food images,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 1, pp. 187– 203, 2021

    2021

  22. [30]

    Towards hierarchical task decomposition using deep reinforcement learning for pick and place subtasks,

    L. Marzari et al., “Towards hierarchical task decomposition using deep reinforcement learning for pick and place subtasks,” in 2021 20th International Conference on Advanced Robotics (ICAR). IEEE, 2021, pp. 640–645

    2021

  23. [31]

    Robot task planning based on large language model representing knowledge with directed graph structures,

    Y . Zhen et al. , “Robot task planning based on large language model representing knowledge with directed graph structures,” 2023. [Online]. Available: https://arxiv.org/abs/2306.05171

    2023 arXiv

  24. [32]

    Reflexion: Language agents with verbal reinforce- ment learning,

    N. Shinn et al. , “Reflexion: Language agents with verbal reinforce- ment learning,” Advances in Neural Information Processing Systems , vol. 36, 2024

    2024

  25. [33]

    Hddl: An extension to pddl for expressing hierar- chical planning problems,

    D. Höller et al. , “Hddl: An extension to pddl for expressing hierar- chical planning problems,” in Proceedings of the AAAI conference on artificial intelligence, vol. 34, no. 06, 2020, pp. 9883–9891

    2020

  26. [34]

    Robotic task oriented knowledge graph for human- robot collaboration in disassembly,

    Y . Ding et al. , “Robotic task oriented knowledge graph for human- robot collaboration in disassembly,” Procedia CIRP , vol. 83, pp. 105–110, 2019, 11th CIRP Conference on Industrial Product-Service Systems. [Online]. Available: https://www.sciencedirect.com/science/ article/...

    2019

  27. [35]

    A dynamic knowledge graph approach to distributed self-driving laboratories,

    J. Bai et al. , “A dynamic knowledge graph approach to distributed self-driving laboratories,” Nature Communications, vol. 15, 01 2024

    2024

  28. [36]

    Vital: Visual teleoperation to enhance robot learning through human-in-the-loop corrections,

    H. Kasaei and M. Kasaei, “Vital: Visual teleoperation to enhance robot learning through human-in-the-loop corrections,” 2024. [Online]. Available: https://arxiv.org/abs/2407.21244

    2024 arXiv

  29. [37]

    Robot learning on the job: Human-in-the-loop autonomy and learning during deployment,

    H. Liu, S. Nasiriany, L. Zhang, Z. Bao, and Y . Zhu, “Robot learning on the job: Human-in-the-loop autonomy and learning during deployment,” 2023. [Online]. Available: https://arxiv.org/abs/ 2211.08416

    2023 arXiv

  30. [38]

    Human-in-the-loop robotic manipulation planning for collaborative assembly,

    M. Raessa et al. , “Human-in-the-loop robotic manipulation planning for collaborative assembly,”IEEE Transactions on Automation Science and Engineering, vol. 17, no. 4, pp. 1800–1813, 2020

    2020

  31. [39]

    Human-in-the-loop machine learning for safe and ethical autonomous vehicles: Princi- ples, challenges, and opportunities,

    Y . Emami, K. Li, L. Almeida, W. Ni, and Z. Han, “Human-in-the-loop machine learning for safe and ethical autonomous vehicles: Princi- ples, challenges, and opportunities,” arXiv preprint arXiv:2408.12548, 2024

    2024 arXiv

  32. [40]

    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 Com- puter Systems, vol. 135, pp. 364–381, 2022

    2022

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.