CoChat
Sign Up
Cited by 0AI-based Problem Solving and PlanningLogic, Reasoning, and KnowledgeMulti-Agent Systems and Negotiation

Autonomous Task Completion Based on Goal-directed Answer Set Programming

Open access
Tudor, Alexis R.·2025·Electronic Proceedings in Theoretical Computer Science
DOI 10.4204/eptcs.416.39
Open this paper in CoChat — free accountOpen PDF

Abstract

Task planning for autonomous agents has typically been done using deep learning models and simulation-based reinforcement learning. This research proposes combining inductive learning techniques with goal-directed answer set programming to increase the explainability and reliability of systems for task breakdown and completion. Preliminary research has led to the creation of a Python harness that utilizes s(CASP) to solve task problems in a computationally efficient way. Although this research is in the early stages, we are exploring solutions to complex problems in simulated task completion.

AI snapshot

Written from the paper’s full text.

TL;DR

This early-stage paper proposes using goal-directed answer set programming (s(CASP) with a Python harness) for autonomous task planning in the VirtualHome simulator, aiming for more explainable and reliable task completion than deep learning alone. Preliminary results show a dynamic dependency graph can drastically reduce planning computation time on three small tasks.

Key findings

  • A Python harness was created to use s(CASP) with simulators, since s(CASP) lacks a Python API; it converts VirtualHome world states into s(CASP) facts and informs the agent’s next action.
  • A dynamic dependency graph prunes facts and rules irrelevant to the query. On three semi-simple tasks, Table 1 reports: “Grab Remote Control” from 13925.14 s unoptimized to 0.55 s optimized; “Grab Remote Control and Shirt” from 608.28 s to 0.71 s; “Grab Cell Phone and Sit on Couch” from 1771.21 s to 0.64 s.
  • Task completion is treated as a planning problem: each task is represented as a final state, and s(CASP) rules search for actions to reach that state from the initial simulated world state.
  • The paper reports that without optimization, planning can take over twelve hours without concluding in sufficiently large environments, and in a small-scale simulation with only six items, “grab the remote” can produce an inefficient plan that walks to every other object first.

Methods

  • Design: an s(CASP) knowledge base for autonomous task completion in a simulated virtual environment, combined with a Python harness for simulator interaction.
  • Data/sample: the VirtualHome simulator, which provides simulated apartments, multiple agents, mid-level control commands, and a database of high-level task breakdowns; testing also uses a small-scale built-in simulation.
  • Main analysis: represent tasks as final states, use s(CASP) planning rules to generate action sequences, and apply a dynamic dependency graph to remove irrelevant facts/rules; compare optimized and unoptimized completion times on three tasks.

Limitations

  • The paper states the research is in an early stage; computation time remains a limitation, large environments produce intractable answer sets, representing time requires event calculus, and s(CASP) lacks a Python API.
  • The paper states ILP generally needs background knowledge and program templates from a domain expert, and logic programs struggle to represent complex data.
  • Evident from the text: results are preliminary and limited to three semi-simple tasks in a small-scale simulation, with no comparison to deep learning or LLM baselines reported.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.4204/eptcs.416.39

Cites

20 of 20 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

10 of 17 cited sources checked · 4 supported · 0 flagged · accuracy not yet measured

  • Premisecitation
    § 2 Background and Relevant Literature · p. 2

    In autonomous task completion, LLMs struggle with making task breakdowns that are both correct and executable.

    “In the specific arena of autonomous task completion, LLMs struggle with making task breakdowns that are both correct and executable [10].”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [10] [10] Wenlong Huang, Pieter Abbeel, Deepak Pathak & Igor Mordatch (2022): Language Models as ZeroShot Planners: Extracting Actionable Knowledge for Embodied Agents . In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvari, Gang Niu & Sivan Sabato, editors: Proceedings of the 39th International Conference on Machine Learning , Proceedings of Machine Learning Research 162, PMLR, pp. 9118-9147, doi:10.48550/arXiv.2201.07207. Available at https://proceedings.mlr.press/v162/ huang22a.html . - source
  • Premisecitation
    § 2 Background and Relevant Literature · p. 2

    The s(CASP) system solves the grounding problem by performing a top-down goal-oriented search which eliminates the need for grounding.

    “The s(CASP) system [1] solves this problem by performing a top-down goal-oriented search which eliminates the need for grounding.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [1] [1] Joaqin Arias, Manuel Carro, Elmer Salazar, Kyle Marple & Gopal Gupta (2018): Constraint Answer Set Programming without Grounding . Theory and Practice of Logic Programming 18(3-4), p. 337-354, doi:10.1017/S1471068418000285. - source
  • Premisecitation
    § 2 Background and Relevant Literature · p. 2

    LLMs are vulnerable to hallucination, where they provide incorrect responses that are statistically likely.

    “LLMs are vulnerable to “hallucination”, where they provide incorrect responses that are statistically likely [9].”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [9] [9] Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin & Ting Liu (2023): A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions . arXiv:2311.05232. - source

      The Introduction (p. 2) says LLM hallucinations commonly manifest as factual errors. Hallucination Detection (p. 16) notes that a hallucinated token can be assigned high probability, but it does not establish that hallucinated responses generally are statistically likely; that part of the claim is broader than the evidence. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Findingreasoning
    § 3.1 Status · p. 5

    The limitation remains computation time.

    “The limitation remains computation time.”
    Quote in the paperSubstantial claim
  • Findingreasoning
    § 3.2 Preliminary Results · p. 7

    Using the dependency graph, the Python harness can simplify the knowledge base, removing the male/1 and grandparent/2 predicates entirely.

    “Using the dependency graph the Python harness can simplify the above knowledge base, removing the male/1 and grandparent/2 predicates entirely.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 26
  • May combine two claims. 10

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    LLMs can be jailbroken to respond outside of the bounds they were designed for.

    “Additionally, they can be “jailbroken” to respond outside of the bounds they were designed for [2].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [2] [2] Patrick Chao, Alexander Robey, Edgar Dobriban, Hamed Hassani, George J. Pappas & Eric Wong (2023): Jailbreaking Black Box Large Language Models in Twenty Queries . arXiv:2310.08419. - source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    ILP can get results rivaling deep learning models while being inherently interpretable and explainable.

    “ILP can get results rivaling deep learning models while being inherently interpretable and explainable [19].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [19] [19] Zheng Zhang, Liangliang Xu, Levent Yilmaz & Bo Liu (2023): A Critical Review of Inductive Logic Programming Techniques for Explainable AI . arXiv:2112.15319. source

      The Introduction says ILP’s symbolic hypotheses are inspectable by humans and provide transparency, supporting the interpretability and explainability part. The Benchmark ILP problem section reports high accuracy for several ILP systems on selected tasks, but compares ILP systems with one another, not with deep-learning models. The claim that ILP can rival deep learning is therefore not established by these passages. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    Using traditional machine learning algorithms to supplement a logic program increases explainability while utilizing the benefits of deep learning and other machine learning models.

    “This approach increases explainability while utilizing the benefits of deep learning and other machine learning models, such as in the paper by Rajasekharan et al. that uses an s(CASP) knowledge base to constrain an LLM into providing more reliable results [14].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [14] [14] Abhiramon Rajasekharan, Yankai Zeng, Parth Padalkar & Gopal Gupta (2023): Reliable Natural Language Understanding with Large Language Models and Answer Set Programming . Electronic Proceedings in Theoretical Computer Science 385, pp. 274-287, doi:10.4204/EPTCS.385.27. - source

      The Related Work section says the approach combines LLM predicate extraction with s(CASP) reasoning, and that the explicit reasoning is more reliable and explainable. The Introduction also says s(CASP) can provide proof-tree justifications. The paper does not describe the knowledge base as constraining the LLM itself; it assigns reasoning to s(CASP) after the LLM extracts predicates. Page numbers are unavailable in our copy. (LLM, conf=0.84)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation
    § 2 Background and Relevant Literature · p. 1

    Most deep learning systems are black boxes that lack explainability and interpretability.

    “However, most deep learning systems are black boxes that lack explainability and interpretability.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [6] [6] David Gunning, Eric Vorm, Jennifer Yunyan Wang & Matt Turek (2021): DARPA's explainable AI (XAI) program: A retrospective . Applied AI Letters 2(4), p. e61, doi:10.1002/ail2.61. arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1002/ail2.61. - source

  • Classified as probably not a substantial claim.

    Premisecitation
    § 2 Background and Relevant Literature · p. 1

    It is difficult to trust that deep learning answers are correct and unbiased.

    “This makes it difficult to trust that their answers are correct and unbiased, as explored in DARPA's explainable AI retrospective [6].”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [6] [6] David Gunning, Eric Vorm, Jennifer Yunyan Wang & Matt Turek (2021): DARPA's explainable AI (XAI) program: A retrospective . Applied AI Letters 2(4), p. e61, doi:10.1002/ail2.61. arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1002/ail2.61. - source

      The Abstract describes AI systems as seemingly inscrutable and says people wanted to know how to trust them. XAI Program Goals links explainability to appropriate trust and to understanding a system’s weaknesses and mistakes. These passages support difficulty assessing whether answers are correct, but our copy does not address whether the answers are unbiased. (LLM, conf=0.85)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    An s(CASP) knowledge base can constrain an LLM into providing more reliable results.

    “This approach increases explainability while utilizing the benefits of deep learning and other machine learning models, such as in the paper by Rajasekharan et al. that uses an s(CASP) knowledge base to constrain an LLM into providing more reliable results [14].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [14] [14] Abhiramon Rajasekharan, Yankai Zeng, Parth Padalkar & Gopal Gupta (2023): Reliable Natural Language Understanding with Large Language Models and Answer Set Programming . Electronic Proceedings in Theoretical Computer Science 385, pp. 274-287, doi:10.4204/EPTCS.385.27. - source

      The Introduction describes combining LLM-generated predicates with ASP-coded commonsense knowledge to produce reliable responses. But the Related Work section explicitly separates the components: the LLM extracts predicates, while s(CASP) performs the reasoning. The paper therefore supports improved reliability through the combined system, not the claim that the knowledge base constrains the LLM itself. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Premisecitation
    § 3.3 Open Issues and Expected Achievements · p. 7

    Explanation-based learning is a promising paradigm that would allow for generalized knowledge from a small number of examples and works well with answer set programming.

    “To that end, explanation-based learning is a promising paradigm that would allow for generalized knowledge from a small number of examples [16] and works well with answer set programming.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [16] [16] Julia Strout, Ye Zhang & Raymond Mooney (2019): Do Human Rationales Improve Machine Explanations? In: Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP , Association for Computational Linguistics, Florence, Italy, pp. 56-62, doi:10.18653/v1/W19-4807. Available at https://aclanthology.org/W19-4807 . - source

      The Introduction and Conclusion discuss learning from human rationales to improve text-classification accuracy and explanations, not explanation-based learning that generalizes from a small number of examples. Searches for that claim and for answer set programming found no supporting passage in our copy. Page numbers were not provided for these passages. (LLM, conf=0.76)

      Experimental — accuracy not yet measured

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given
    § 3.1 Status · p. 5

    In a simple example, the task of walking towards a remote control can be easily accomplished by the s(CASP) program.

    “In this very simple example, the task is to walk towards a remote control, which can be easily accomplished by the program.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given
    § Autonomous Task Completion Based on Goal-directed Answer Set Programming · p. 1

    Preliminary research has led to the creation of a Python harness that utilizes s(CASP) to solve task problems in a computationally efficient way.

    “Preliminary research has led to the creation of a Python harness that utilizes s(CASP) to solve task problems in a computationally efficient way.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given
    § 3.1 Status · p. 4

    For testing purposes, the Python harness has a small-scale simulation environment built in.

    “For testing purposes, the Python harness has a small-scale simulation environment built in.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingown data
    § 3.1 Status · p. 3

    Using the Python harness, the simulated VirtualHome environment can be instantiated and transformed into an s(CASP) representation of the world state.

    “Using the Python harness mentioned above, the simulated VirtualHome environment can be instantiated and transformed into an s(CASP) representation of the world state:”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingreasoning
    § 3.1 Status · p. 5

    Adding the rule suggest(walk(X),State) :- member(holds(X),State), not member(close(X),State) solves the problem, but the same must then be done for any other state requiring closeness as a prerequisite.

    “This problem can be solved by adding the rule suggest(walk(X),State) :-member(holds(X),State), not member(close(X),State). , however the same then must be done for any other state requiring closeness as a prerequisite.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Findingreasoning
    § 3.3 Open Issues and Expected Achievements · p. 7

    The use of a dependency graph to pare down the knowledge base allows us to perform more complicated tasks, however there can be more optimization.

    “The use of a dependency graph to pare down the knowledge base allows us to perform more complicated tasks, however there can be more optimization.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingreasoning
    § 3.1 Status · p. 4

    The complexity of generating an answer set that accounts for all of these facts and possible worlds quickly becomes a computational obstacle.

    “The complexity of generating an answer set that accounts for all of these facts and possible worlds quickly becomes a computational obstacle.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given
    § 3.1 Status · p. 4

    Even without the use of time, the representation of the world state easily grows to encompass a large amount of facts.

    “Even without the use of time, this representation of the world state easily grows to encompass a large amount of facts.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 3.1 Status · p. 3

    There have already been promising results in producing executable actions for small-scale real-world tasks.

    “Although this research is still in an early stage, there have already been promising results in producing executable actions for small-scale real-world tasks.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 3.2 Preliminary Results · p. 7

    In a program of this size, the computational savings of such optimization is negligible.

    “In a program of this size, the computational savings of such optimization is negligible.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    Recent advances in ILP, such as the FOLD family of algorithms, demonstrate that complex data can be represented in small logic programs using default rules.

    “Recent advances in ILP, such as the FOLD family of algorithms, demonstrate that complex data can be represented in small logic programs using default rules.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] [7] Gopal Gupta, Huaduo Wang, Kinjal Basu, Farhad Shakerin, Elmer Salazar, Sarat Chandra Varanasi, Parth Padalkar & Sopam Dasgupta (2023): Logic-based explainable and incremental machine learning . Prolog: The Next 50 Years , p. 346-358, doi:10.1007/978-3-031-35254-6\_28. - source

  • Classified as probably not a substantial claim. May combine two claims.

    Premisecitation
    § 2 Background and Relevant Literature · p. 2

    s(CASP) is well-suited to the representation of complex world states and provides an advantage over other ASP systems.

    “This advantage makes s(CASP) well-suited to the representation of complex world states and provides an advantage over other ASP systems [5].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] [5] Michael Gelfond & Yulia Kahl (2014): Knowledge representation, reasoning, and the design of Intelligent Agents: The answer-set programming approach . Cambridge University Press, doi:10.1017/CBO9781139342124. - source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2 Background and Relevant Literature · p. 1

    Modern autonomous systems mostly use deep learning models to accomplish tasks.

    “For the most part, modern autonomous systems use deep learning models to accomplish this [11].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [11] [11] Eduardo F Morales, Rafael Murrieta-Cid, Israel Becerra & Marco A Esquivel-Basaldua (2021): A survey on deep learning and deep reinforcement learning in robotics with a tutorial on deep reinforcement learning . Intelligent Service Robotics 14(5), pp. 773-805, doi:10.1007/s11370-021-00398-z. - source

  • May combine two claims.

    Premisecitation
    § 3 Methodology · p. 2

    VirtualHome allows for multiple agents to operate in a variety of simulated apartments and provides a large database of high-level task breakdowns into step-by-step instructions.

    “VirtualHome allows for multiple agents to operate in a variety of simulated apartments, and provides a large database of high-level task breakdowns into step-by-step instructions.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [13] [13] Xavier Puig, Kevin Ra, Marko Boben, Jiaman Li, Tingwu Wang, Sanja Fidler & Antonio Torralba (2018): VirtualHome: Simulating Household Activities via Programs . arXiv:1806.07011. - source

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning
    § 4 Conclusion · p. 8

    Simply creating a Python framework for the use of s(CASP) with simulated environments is an advancement for s(CASP).

    “Simply creating a Python framework for the use of s(CASP) with simulated environments is an advancement for s(CASP), as it is currently lacking a Python API.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: s(CASP) does not have a Python API.

  • Classified as probably not a substantial claim.

    Interpretationreasoning
    § 3.3 Open Issues and Expected Achievements · p. 7

    This produces answer sets that are intractable to generate.

    “This produces answer sets that are intractable to generate.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given
    § 3.3 Open Issues and Expected Achievements · p. 7

    Representing a simulation of any reasonable size leads to an exponential increase in the number of facts available in the world state.

    “Representing a simulation of any reasonable size leads to an exponential increase in the number of facts available in the world state.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning
    § 3.3 Open Issues and Expected Achievements · p. 8

    Once that point is reached, there would still be benefits in combining s(CASP) with more traditional machine learning (and newer deep learning, such as LLMs) to improve performance in the former and explainability of the latter.

    “Once that point is reached, there would still be benefits in combining s(CASP) with more traditional machine learning (and newer deep learning, such as LLMs) to improve performance in the former and explainability of the latter.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given
    § 3.3 Open Issues and Expected Achievements · p. 8

    Likely, s(CASP) by itself cannot encode all of the complexities of a real environment and remain tractable.

    “Likely, s(CASP) by itself cannot encode all of the complexities of a real environment and remain tractable.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 3.3 Open Issues and Expected Achievements · p. 7

    This is a known problem with representing continuous time in logic programming and would require the inclusion of event calculus.

    “This is a known problem with representing continuous time in logic programming and would require the inclusion of event calculus [18].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] [18] Sarat Chandra Varanasi, Joaquín Arias, Elmer Salazar, Fang Li, Kinjal Basu & Gopal Gupta (2022): Modeling and Verification of Real-Time Systems with the Event Calculus and s(CASP) . In: Practical Aspects of Declarative Languages: 24th International Symposium, PADL 2022, Philadelphia, PA, USA, January 17-18, 2022, Proceedings , Springer-Verlag, Berlin, Heidelberg, p. 181-190, doi:10.1007/978-3-030-94479-7\_12. - source

  • May combine two claims.

    Interpretationreasoning
    § 3.1 Status · p. 5

    This decreases the value of logical inference and increases the rules required for simple task planning.

    “This decreases the value of logical inference and increases the rules required for simple task planning.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 3.3 Open Issues and Expected Achievements · p. 7

    The use of time in the knowledge base provides complications related to the ostensibly infinitely divisible nature of time.

    “As mentioned above, the use of time in the knowledge base provides complications related to the ostensibly infinitely divisible nature of time (as posited by the famous Greek philosopher Zeno).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] [18] Sarat Chandra Varanasi, Joaquín Arias, Elmer Salazar, Fang Li, Kinjal Basu & Gopal Gupta (2022): Modeling and Verification of Real-Time Systems with the Event Calculus and s(CASP) . In: Practical Aspects of Declarative Languages: 24th International Symposium, PADL 2022, Philadelphia, PA, USA, January 17-18, 2022, Proceedings , Springer-Verlag, Berlin, Heidelberg, p. 181-190, doi:10.1007/978-3-030-94479-7\_12. - source
29 of 29

  • The paper relies on this without support.

    s(CASP) does not have a Python API.

    “A notable weakness of s(CASP) is that it does not have a Python API, which makes it difficult to run in line with other forms of machine learning.”
    “Simply creating a Python framework for the use of s(CASP) with simulated environments is an advancement for s(CASP), as it is currently lacking a Python API.”

  • The paper relies on this without support.

    Logic programs are inherently interpretable and their errors can be logically understood and solved.

    “One advantage of logic programs is that they are inherently interpretable and their errors can be logically understood and solved.”

  • The paper relies on this without support.

    Traditional ASP grounding is not always guaranteed to be feasible, which can leave some programs with no ASP solution.

    “A disadvantage of this approach is that grounding is not always guaranteed to be feasible, which can leave some programs with no ASP solution.”

  • The paper relies on this without support.

    ILP requires background information and program templates provided by a domain expert.

    “One of the biggest weaknesses of the ILP approach to solving problems is the need for background information and 'program templates'. Program templates are a layout of how the generated information should look in the context of the logic program. A domain expert must provide this program template and explicitly logic program-based background knowledge for most ILP.”

  • The paper relies on this without support.

    Logic programs have a difficult time representing complex data.

    “Additionally, while ILP programs perform very well on data that can be represented in a logic program, logic programs have a difficult time representing complex data.”

  • The paper relies on this without support.

    Deep learning algorithms are dependent on often flawed data they are trained on.

    “This is especially dangerous given how dependent deep learning algorithms are on the (often flawed) data they are trained on.”
6 of 6

Cites

Flagged first, then by citations

Related

Authors
Tudor, A. R.
Affiliations
The University of Texas at Dallas (US)
Corresponding author
Tudor, Alexis R.
ORCID
Tudor, Alexis R.
Venue
Electronic Proceedings in Theoretical Computer Science
Publisher
Open Publishing Association
Published
2025-02-13
Volume
416
Pages
381-389
ISSN
2075-2180 (ISSN-L)
DOI
10.4204/eptcs.416.39
OpenAlex
W4407349683
Semantic Scholar
9fd7dab028e85440fa6064bb170a7dff68a82375
Cited by
0
References
13
FWCI
0.00
Type
journal article
Language
en
Open access
Licence
cc-by-nc-nd
Locations
  • Electronic Proceedings in Theoretical Computer Science (publishedVersion, cc-by-nc-nd)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
  • Physical Sciences › Computer Science › Artificial Intelligence › Logic, Reasoning, and Knowledge
  • Physical Sciences › Computer Science › Artificial Intelligence › Multi-Agent Systems and Negotiation
Keywords
task planning, autonomous agents, task decomposition, explainability, inductive learning, task completion

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-19.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-19.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review