GenPlanX. Generation of Plans and Execution

Alberto Pozanco; Alfredo Garrach\'on; Charese Smiley; Daniel Borrajo; Giuseppe Canonaco; Keshav Ramani; Manuela Veloso; Marianela Morales; Pietro Totis; Simerjot Kaur

REVIEW 3 major objections 7 minor 45 references

GenPlanX. Generation of Plans and Execution

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read GenPlanX is an architecture that turns natural-language office requests into cost-optimal, executable plans by giving the LLM the job of writing the problem and a classical planner the job of solving it.

desk verdict A clean integration of known LLM-plus-planner components whose load-bearing translation step is unmeasured and, in one example, demonstrably wrong—worth reviewing, not ready to be believed. read the letter →

arxiv 2506.10897 v1 pith:TACTGM2Q submitted 2025-06-12 cs.AI

Daniel Borrajo , Giuseppe Canonaco , Tomás de la Rosa , Alfredo Garrachón , Sriram Gopalakrishnan , Simerjot Kaur , Marianela Morales , Sunandita Patra

show 5 more authors

Alberto Pozanco Keshav Ramani Charese Smiley Pietro Totis Manuela Veloso

This is my paper · ORCID

classification cs.AI

keywords LLMplanningclassicalAIPDDLexecutionmonitoringreplanningofficeautomationnaturallanguageunderstandingsoftbots

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

The paper tries to establish that a planning system can take office requests in plain English and carry them out with the formal guarantees classical planning provides, without losing the language understanding only an LLM offers. Its central claim is that GenPlanX—which sends the request through entity extraction, uses an LLM to return a structured problem dictionary, compiles that dictionary into PDDL, solves it with a classical planner, and executes the plan with monitoring and replanning—delivers end-to-end office workflows such as generating a PowerPoint, choosing a cost-minimal database path, and training a decision-tree model. The intended significance is that LLMs alone cannot be trusted for plan correctness or optimality, while classical planners cannot read natural language; the hybrid puts each component where it is strong.

What carries the argument

The load-bearing object is the 'assistant' PDDL domain, a curated set of typed objects, predicates, and actions representing office tools and data: types such as data-file, dataframe, graph, slide, presentation, model, and response; predicates such as (in ?c ?c1), (available ?o), (done-query ?q), and (in-graph ?c ?c1 ?g); and actions like read-data, query-data-basic, query-data-optimized, create-graph, add-to-graph, create-slide, generate-presentation, learn-supervised, and predict-using-learned-model. Each action has a Python function of the same name that actually performs the operation on the execution state, so the plan is directly executable. The second load-bearing piece is the LLM-produced dictionary, the only place natural language enters: it specifies the objects, initial state, and goals, and the compiler turns it into PDDL. The separation is what lets the planner's validity guarantee apply to a problem that was authored in English.

What would settle it

Give GenPlanX a request with a known ground truth—'Read annual-report.csv, make a bar chart of balance by year, and email the result to alice@example.com'—and inspect whether the LLM's dictionary contains the email object and the sent-email goal; a well-formed dictionary that omits a stated goal or misrepresents the initial file would lead to an artifact that misses the request while the system reports success. A broader falsifier is a batch evaluation of such requests where the generated problem specification is compared to gold-standard extractions; any miss that the execution loop does not catch would signal that the classical planner's guarantee does not reach the user.

Watch

Extended reading notes

Core claim

The central discovery is the architecture itself and the division of labor inside it. The LLM never proposes a plan; it only proposes the planning problem, in the form of a Python dictionary whose keys are objects, 'init_state', and 'goals', and whose values carry both PDDL types and execution-time values such as file paths and query strings. A compiler converts that dictionary into a PDDL problem, and a classical planner returns a cost-optimal action sequence. The execution module maps each PDDL action to a same-named Python function, checks after each action whether the real-world effect matched, and invokes replanning on failure or when an action (such as reading an email) reveals new goals. The paper's evidence is three worked end-to-end examples plus a small comparison in which two recent LLMs, when asked to plan directly without hints, produced at most 4/5 correct and 0/5 optimal plans, while both reached 5/5 with hints, supporting the claim that the planner is what supplies correctness.

Load-bearing premise

The load-bearing premise is that the LLM will convert the user's request into a complete and accurate dictionary of objects, initial state, and goals; if it misses a goal or misstates the initial configuration, the planner will efficiently solve the wrong problem and the execution loop will carry out that wrong plan with confidence.

Editorial extensions

If this is right

Plans produced by GenPlanX are provably valid with respect to the problem the LLM generated, so any end-to-end failure can be traced to translation rather than plan search.
Because costs are encoded in domain actions, the planner can optimize over alternatives that LLM-only planning gets wrong, as in Example 1.2 where the lower-cost database route is chosen.
The architecture supports adding new tools by defining a PDDL action and a same-named Python function, making it adaptable to new office applications.
Execution monitoring plus replanning lets the system react to failures and to goals only discovered during execution, such as actions that read email contents and create new intents.
Any PDDL planner can be substituted, making the system planner-independent.

Reading between the lines

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

A fair evaluation would separate translation fidelity from planning: measure how often the LLM-produced dictionary matches a validated extraction of objects, initial state, and goals on a held-out set of office requests, because the planner's guarantee cannot compensate for translation error.
Using a Python-dictionary intermediate representation, rather than asking the LLM to write PDDL directly, likely makes the translation step easier for the model and also lets execution values such as file paths and email addresses ride along as object values, a design choice worth adopting in similar systems.
The replanning-on-new-goals behavior invites a goal-driven autonomy extension: the monitoring layer could detect opportunities in the environment and propose new goals, not just wait for an action to fail or explicitly announce them.

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.

Desk Editor's Note

read the letter

Straight to the point: this is a system paper describing GenPlanX, an integration of an LLM that turns natural-language office requests into a structured dictionary, a compiler that turns that dictionary into PDDL, a classical planner, and an execution-and-monitoring layer with replanning. The components are known—the paper itself cites LLM+P, TRIP-PAL, and the LLM-modulo framework—and the architecture is a reasonable instantiation with two small twists: the LLM outputs a dictionary instead of raw PDDL, and the domain is a realistic office assistant with actions for files, databases, presentations, and ML. The paper does a few things well. It gives a complete prompt in the appendix, which is genuinely useful for anyone building a similar copilot. It explains the execution mapping from PDDL actions to Python functions. And it includes a small sanity check showing that GPT-4o and o3-mini alone struggle to produce optimal plans for the same problem, which motivates the hybrid design.

The soft spot is the evidence. The end-to-end claim—that GenPlanX can reliably translate user intent into a planning problem that matches that intent—is supported only by a handful of curated examples. No translation accuracy, no plan-validity statistics, no success rates, no comparison against an end-to-end baseline. That matters because the planner's guarantee is conditional: it guarantees the plan is valid for the PDDL problem it receives, not for the user's request. If the LLM misstates a cost or a fact, the planner confidently solves the wrong problem.

And the paper contains a concrete instance of exactly that failure. In Example 1.2, the user request says db2 has reading cost 3. The LLM-generated initial state in Figure 7 sets (database-cost database2) to 2, and Table 2 and the surrounding text also use 2. Figure 7 also has a malformed literal and mixes db2 with database2. If that output came from the real pipeline, the 'optimal' plan the authors highlight is optimal for a problem the user didn't specify. That is not a minor typo; it's the load-bearing step in action. The stress-test note is correct about this.

The architecture itself is sound, and the paper is not circular or incoherent. But the central efficacy claim is not supported as written. The fix is straightforward: add a modest evaluation—say, fifty varied requests, report how often the LLM produces the right objects/init/goals, how often the compiled plan is executable, and what happens on failure—and fix the Example 1.2 discrepancy. If the translation reliability isn't there, scale back the claim to 'architecture and examples' rather than 'demonstrate efficacy.'

Who is this for? Researchers or engineers building LLM-plus-planner copilots; they'll get a useful prompt design and a concrete domain. It deserves peer review—the system is real and the translation step is exactly what a reviewer should pressure-test—but I'd expect major revision, not acceptance on the current evidence.

Referee Report

3 major / 7 minor

Summary. The paper presents GenPlanX, a hybrid system for office tasks in which an LLM turns a natural-language request into a Python dictionary containing entities, an initial state, and goals; a compiler converts that dictionary into PDDL; a classical planner (Fast Downward via the Unified Planning library) returns a plan; and an execution module maps PDDL actions to Python functions with monitoring and replanning on failure. The authors describe the architecture, the 'assistant' domain, the full prompt, several worked end-to-end examples (annual-report charts, decision-tree training and prediction, appointments filtering), and a small comparison showing that GPT-4o and o3-mini produce incorrect or suboptimal plans when asked to plan directly. The scientific claims are that this design yields plans that are valid and optimal with respect to the compiled planning problem, and that GenPlanX is effective end-to-end for office-related tasks.

Significance. The architecture is sensible and its division of labor—LLM for translation, classical planner for planning, Python functions for execution—is a credible way to avoid the lack of guarantees in LLM-only planning. The paper is transparent: it provides the full prompt, domain fragments, action implementations, and the raw LLM-planning comparison, which is a strength. If the translation step were shown to be reliable, the system would be a useful contribution in the line of LLM-plus-planner frameworks. However, the central efficacy claim currently rests on a handful of curated examples and no quantitative measurement of the LLM-to-dictionary step. The stress-test concern raised by the reader is real and is borne out by the inconsistencies in Example 1.2. The result is a promising system description rather than a demonstrated end-to-end solution.

major comments (3)

[§8.1, Example 1.2 / Figure 7 / Table 2] The user request states that db2 has reading cost 3, but the LLM-generated initial state in Figure 7 contains '(= (database-cost database2) 2)' and Table 2 and the surrounding text use cost 2 for read-data(db2). Figure 7 also contains a syntactically malformed literal '(in dataframe1 db1) in dataframe1 db2)' and mixes the object names 'db1'/'db2' with 'database1'/'database2'. Because the initial state is the input to the planner, the planner's 'optimal' plan minimizes cost with respect to a problem that does not match the user's stated request. If this is pipeline output, it is a concrete instance of the failure mode the paper must rule out; if it is a typo, it must be corrected and the example re-verified. As written, this section undermines rather than supports the claim that GenPlanX produces the optimal plan for the user's actual request.
[§6, §8, §10.1] The paper contains no quantitative evaluation of the LLM's dictionary generation, which is the load-bearing step of the system. There are no reported statistics on translation accuracy, goal completeness, initial-state faithfulness, type correctness, entity-extraction rates, or plan validity of the compiled problems. Section 11 compares LLM-only planning against a classical planner, but that comparison does not test GenPlanX's translation module; if the LLM produces a wrong dictionary, the planner still solves the wrong problem and the error is invisible in that comparison. The single handwritten prompt in Appendix 10.1 and the three curated examples do not establish reliability across the intent list of Section 6.2. A systematic evaluation, for example N requests per intent with human-verified dictionaries and plans, is needed before the end-to-end efficacy claim can be accepted.
[§7.2] The paper claims that GenPlanX monitors execution and replans on failure, but it reports no experiments in which an action fails or a new goal is generated during execution; all worked examples show successful first-pass execution. The success predicates in Listing 5 are essentially type checks, and the text does not specify how a detected failure is translated back into a PDDL state or what guarantees the replanning procedure provides. As written, the execution-monitoring and replanning claim is architectural rather than demonstrated. At minimum, the paper should define the replanning protocol precisely or present a failure-injection study.

minor comments (7)

[§2.1 and References] The text cites 'Etizioni et al.', but the reference list entry [8] is '0. Etzioni'; the '0.' appears to be a typo for 'O.', and the author name should be harmonized.
[References] References [41] and [42] are identical; one of them should be removed or replaced with the intended distinct source.
[§7.1] The phrase 'problem stated in 2' should read 'problem stated in Listing 2'.
[§11.2] The text says 'shown in Figure 8' when referring to the results table; the correct reference is Table 8.
[§11.1] The sentence 'We also tested with providing the action descriptions in PDDL and got simular results' contains a typo: 'simular' should be 'similar'.
[Appendix 10.1] The instruction 'Do not return Output: in the output' is ambiguous; it should be rephrased to make clear whether it means 'do not prefix your response with Output:'.
[Abstract and headings] Several instances of 'GenPlanX' are missing the space before the word (e.g., 'introducesGenPlanX' in the abstract and 'handle file and data management tasks' formatting issues); these should be fixed in the final formatting pass.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the LLM-to-PDDL-to-planner pipeline is a forward derivation, not a reduction to its inputs.

full rationale

GenPlanX's derivation chain is a forward pipeline: natural-language request to an LLM-produced dictionary, compilation to a PDDL problem, classical planning, and execution. The planner's guarantee that the plan achieves the goals from the initial state is with respect to the PDDL problem actually constructed; the LLM output is an input to that computation and is not itself derived from the plan. No parameter is fitted to output data and then renamed as a prediction, and no uniqueness theorem from prior work by the same authors is invoked to force a choice. The paper's self-citations, such as [21] for entity extraction and [25] for future work on learning action costs, are background or forward-looking and are not load-bearing for the central claim. The Appendix 11 comparison against GPT-4o and o3-mini is an independent empirical comparison, and the optimality contrast in Example 1.2 is computed by the classical planner over the stated planning problem. There is a factual inconsistency in Example 1.2 (the user request says db2 has reading cost 3, while Figure 7 sets database-cost to 2 and Table 2 uses 2+2=4), which is a fidelity/correctness risk but not circularity: the planner still optimizes whatever problem model it is given. Overall, no claimed result reduces by construction to its own inputs, so the circularity score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 1 invented entities

The central claim relies on the handcrafted PDDL domain, the handcrafted prompt, and the assumption that the LLM translates accurately into PDDL. The only numeric free parameters are the action costs, which are set manually. The paper does not fit parameters to data or derive constants from first principles.

free parameters (2)

Cost values in the domain model = 1 for most actions, 5 for query-data-basic, 2 for query-data-optimized
These costs are chosen by hand for the examples to make the optimal plan illustration work. They are not learned or derived from data. The planner's optimality guarantee is relative to these arbitrary costs, so the claim that the system produces optimal plans is only meaningful given these hand-set numbers.
Prompt structure and few-shot examples = Handwritten prompt in Appendix 10.1
The entire LLM translation performance depends on the specific prompt, which is handcrafted. The paper does not tune or evaluate the prompt against a benchmark. It is a free parameter of the system, though not a numeric one.

assumptions (4)

standard math Classical planners such as Fast Downward are sound and complete for the PDDL fragments used.
The paper relies on the planner's guarantee that any returned plan achieves the goals from the initial state. This is a standard property of classical planning when the problem is correctly specified.
domain assumption The LLM will reliably produce well-formed dictionaries that match the prompt's constraints.
The entire pipeline depends on the LLM following the prompt's instructions: lowercased keys, correct types, literals separated by spaces, and no reference to undefined predicates. The paper provides no quantitative evidence that this holds across varied requests.
domain assumption The world is fully described by the PDDL state, and action preconditions and effects accurately capture real-world action outcomes.
The planner's guarantee is about the PDDL model, not the real office environment. The paper acknowledges that execution monitoring only checks low-level boolean conditions, not the high-level PDDL state, so deviations from the model may go undetected.
domain assumption Entity extraction correctly identifies the domain-specific entities needed by the LLM.
The paper describes the entity extraction module as an ensemble of deep learning, patterns, and domain expertise, citing prior work. No evaluation of this module is given in this paper, yet it is an input to the prompt.

invented entities (1)

The assistant PDDL domain
purpose: Defines the actions, types, and predicates that the planner uses to generate office-task plans.
The domain model is handcrafted for this paper. It is not a discovered structure or a new theoretical entity. Its validity is measured only by the examples in the paper.

how reviews work

0 comments

Cite this review

Pith. "Pith review of GenPlanX. Generation of Plans and Execution." pith.science (2026). https://pith.science/paper/TACTGM2Q

@misc{pith2026250610897,
  author       = {Pith},
  title        = {Pith review of: GenPlanX. Generation of Plans and Execution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TACTGM2Q}},
  note         = {Machine review of arXiv:2506.10897}
}

read the original abstract

Classical AI Planning techniques generate sequences of actions for complex tasks. However, they lack the ability to understand planning tasks when provided using natural language. The advent of Large Language Models (LLMs) has introduced novel capabilities in human-computer interaction. In the context of planning tasks, LLMs have shown to be particularly good in interpreting human intents among other uses. This paper introduces GenPlanX that integrates LLMs for natural language-based description of planning tasks, with a classical AI planning engine, alongside an execution and monitoring framework. We demonstrate the efficacy of GenPlanX in assisting users with office-related tasks, highlighting its potential to streamline workflows and enhance productivity through seamless human-AI collaboration.

Figures

Figures reproduced from arXiv: 2506.10897 by the authors.

**Figure 1.** Architecture of GenPlanX. Boxes in green are the unique implementations for our approach. Boxes in white are components integrated from existing AI tools. • Entity Extraction: This module analyzes the incoming request to identify relevant entities and their types. These entities and entity types are part of the organizational knowledge and are usually not known by general-purpose LLMs. This step is essential for und… view at source ↗

**Figure 2.** Example of a structured output dictionary for getting the status of a trade. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

**Figure 3.** The initial state and goals for Example 1.1. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

Figures from the paper (16 more)

**Figure 4.** Figure 4: Initial state, goals and plan for Example 1.1 in PDDL. The initial state and the goals are generated [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

**Figure 5.** Figure 5: Step by step plan generated and executed by [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

**Figure 6.** Figure 6: The data file (left) and the final result, Powerpoint Slide (right) created for Example 1.1 and 1.2. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]

**Figure 7.** Figure 7: Initial state and goals for Example1.2 We can observe that the optimal plan should include read-data(db2) and query-data-optimized. With the help of a classical planner, GenPlanX is able to generate an optimal plan (as shown in [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

**Figure 8.** Figure 8: The initial state, goals and plan for Example 1.2 in PDDL. The initial state and the goals are [PITH_FULL_IMAGE:figures/full_fig_p015_8.png]

**Figure 9.** Figure 9: Step by step plan generated and executed by [PITH_FULL_IMAGE:figures/full_fig_p016_9.png]

**Figure 10.** Figure 10: The initial state and goals in Example 2. [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]

**Figure 11.** Figure 11: The initial state, goals and plan for Example 2in PDDL. The initial state and the goals are [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]

**Figure 12.** Figure 12: Plan generation and execution for User Request 2. [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]

**Figure 13.** Figure 13: The input file (top) and the final result, prediction of the ‘WillPurchase’ column using learned [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]

**Figure 14.** Figure 14: The initial state, goals, and plan for the appointments data analysis. The initial state and the [PITH_FULL_IMAGE:figures/full_fig_p024_14.png]

**Figure 15.** Figure 15: Step by step plan generated and executed by [PITH_FULL_IMAGE:figures/full_fig_p024_15.png]

**Figure 16.** Figure 16: Appoinments file created by GenPlanX 24 [PITH_FULL_IMAGE:figures/full_fig_p024_16.png]

**Figure 17.** Figure 17: Connecting to Outlook and saving the appointments. [PITH_FULL_IMAGE:figures/full_fig_p025_17.png]

**Figure 18.** Figure 18: Appoinments counts created by GenPlanX 11 Plan Generation Using LLMs In this section, we provide more information on how we use LLM directly to obtain plans in the office domain. 11.1 Prompting To generate a plan using LLM, we provide a prompt describing the actions o…

**Figure 19.** Figure 19: The initial state, goals, and plan for filtering appointments and counting specific types. The [PITH_FULL_IMAGE:figures/full_fig_p029_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 35 canonical work pages

[1]

Ai-Chang, J

M. Ai-Chang, J. Bresina, L. Charest, A. Chase, J.C.-J. Hsu, A. Jonsson, B. Kanefsky, P. Morris, K. Rajan, J. Yglesias, B.G. Chafin, W.C. Dias, and P.F. Maldague. MAPGEN: Mixed-initiative planning and scheduling for the Mars Exploration Rover mission.IEEE Intelligent Systems, 19(1):8–12, feb 2004

work page 2004
[2]

Learning action models with minimal ob- servability.Artificial Intelligence, 275:104–137, 2019

Diego Aineto, Sergio Jim´ enez Celorrio, and Eva Onaindia. Learning action models with minimal ob- servability.Artificial Intelligence, 275:104–137, 2019

work page 2019
[3]

Daniel Borrajo and Manuela M. Veloso. Computing opportunities to augment plans for novel replanning during execution. InProceedings of ICAPS, Guangzhou (China), 2021

work page 2021
[4]

Language models are few- shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

Tom B Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few- shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

work page 1901
[5]

A multi-agent architecture for intelligent gathering systems.AI Communications, 18(1):15–32, 2005

David Camacho, Ricardo Aler, Daniel Borrajo, and Jos´ e Manuel Molina. A multi-agent architecture for intelligent gathering systems.AI Communications, 18(1):15–32, 2005

work page 2005
[6]

Planning for tourism routes using social networks.Expert Systems with Applications, 69:1–9, 2017

Isabel Cenamor, Sergio N´ u˜ nez, Tom´ as de la Rosa, and Daniel Borrajo. Planning for tourism routes using social networks.Expert Systems with Applications, 69:1–9, 2017

work page 2017
[7]

TRIP- PAL: Travel planning with guarantees by combining large language models and automated planners

Tom´ as de la Rosa, Sriram Gopalakrishnan, Alberto Pozanco, Zhen Zeng, and Daniel Borrajo. TRIP- PAL: Travel planning with guarantees by combining large language models and automated planners. arXiv e-prints, abs/2406.10196, 2024

arXiv 2024
[8]

Etzioni, S

0. Etzioni, S. Hanks, D. Weld, D. Draper, N. Lesh, , and M. Williamson. An approach to planning with incomplete information. InProc. 3rd Int. Conf. on Principles of Knowledge Representation and Reasoning, 1992. 17 Initial State: (in dataframe1 data-file1) (in dataframe2 data-file2) (= (database-cost database1) 1) read-data(ai, dataframe1, data-file1, data...

work page 1992

Show all 45 references

[9]

Assisting data mining through automated planning

Fernando Fern´ andez, Daniel Borrajo, Susana Fern´ andez, and David Manzano. Assisting data mining through automated planning. In P. Perner, editor,Proceedings of the International Conference on Ma- chine Learning and Data Mining (MLDM 2009), volume 5632 ofLecture Notes in Art...

2009
[10]

Anticipation of goals in automated planning

Raquel Fuentetaja, Daniel Borrajo, and Tom´ as de la Rosa. Anticipation of goals in automated planning. AI Communications, 31(2):117–135, March 2018

2018
[11]

Florez,´Alvaro Torralba, Daniel Borrajo, Carlos Linares-L´ opez, ´Angel Garc ´ ıa- Olaya, and Juan S´ aenz

Javier Garc ´ ıa, Jos´ e E. Florez,´Alvaro Torralba, Daniel Borrajo, Carlos Linares-L´ opez, ´Angel Garc ´ ıa- Olaya, and Juan S´ aenz. Combining linear programming and automated planning to solve intermodal transportation problems.European Journal of Operations Research, 227(...

2013
[12]

Ghallab, A

M. Ghallab, A. Howe, C. Knoblock, D. McDermott, A. Ram, M. Veloso, D. Weld, and D. Wilkins. PDDL - the planning domain definition language. Technical Report CVC TR-98-003/DCS TR-1165, Yale Center for Computational Vision and Control, 1998

1998
[13]

Access Online via Elsevier, 2004

Malik Ghallab, Dana Nau, and Paolo Traverso.Automated planning: theory & practice. Access Online via Elsevier, 2004. 18 Figure 13: The input file (top) and the final result, prediction of the ‘WillPurchase’ column using learned model for the (bottom) Example 2

2004
[14]

A planning approach to repair domains with incomplete action effects

Alba Gragera, Raquel Fuentetaja, ´Angel Garc ´ ıa-Olaya, and Fernando Fern´ andez. A planning approach to repair domains with incomplete action effects. InProceedings of the International Conference on Automated Planning and Scheduling, volume 33, pages 153–161, 2023

2023
[15]

Leveraging pre- trained large language models to construct and utilize world models for model-based task planning

Lin Guan, Karthik Valmeekam, Sarath Sreedharan, and Subbarao Kambhampati. Leveraging pre- trained large language models to construct and utilize world models for model-based task planning. Advances in Neural Information Processing Systems, 36:79081–79094, 2023

2023
[16]

Transformers for natural language to structured planning: Integrating domain knowledge

Rajesh Gupta, Suresh Patel, and Ming Lee. Transformers for natural language to structured planning: Integrating domain knowledge. InProceedings of the AAAI Conference on Artificial Intelligence, pages 4567–4574, 2021

2021
[17]

Building a domain-independent architecture for planning, learning and execution.pelea

C´ esar Guzm´ an, Vidal Alc´ azar, David Prior, Eva Onaind ´ ıa, Daniel Borrajo, and Juan Fdez-Olivares. Building a domain-independent architecture for planning, learning and execution.pelea. InProceedings of the ICAPS 2011 System Demonstrations, Freiburg (Germany), 2011. AAAI Press

2011
[18]

The fast downward planning system.Journal of Artificial Intelligence Research, 26:191– 246, 2006

Malte Helmert. The fast downward planning system.Journal of Artificial Intelligence Research, 26:191– 246, 2006

2006
[19]

Language models as zero-shot planners: Extracting actionable knowledge for embodied agents.arXiv preprint arXiv:2201.07207, 2022

Wenlong Huang, Pieter Abbeel, Deepak Pathak, and Roozbeh Mottaghi. Language models as zero-shot planners: Extracting actionable knowledge for embodied agents.arXiv preprint arXiv:2201.07207, 2022

2022 arXiv
[20]

Llms can’t plan, but can help planning in llm-modulo frameworks

Subbarao Kambhampati, Karthik Valmeekam, Lin Guan, Kaya Stechly, Mudit Verma, Siddhant Bham- bri, Lucas Saldyt, and Anil Murthy. Llms can’t plan, but can help planning in llm-modulo frameworks. arXiv preprint arXiv:2402.01817, 2024

2024 arXiv
[21]

Advanced messaging platform (amp): Pipeline for automated enterprise email processing

Simerjot Kaur, Charese Smiley, Keshav Ramani, Elena Kochkina, Mathieu Sibue, Samuel Mensah, Pietro Totis, Cecilia Tilli, Toyin Aguda, Daniel Borrajo, and Manuela Veloso. Advanced messaging platform (amp): Pipeline for automated enterprise email processing. InProceedings of the...

2025
[22]

Elevator control as a planning problem

Jana Koehler and Kilian Schuster. Elevator control as a planning problem. InProceedings of the Fifth International Conference on Artificial Intelligence Planning Systems, pages 331–338, 2000. 19

2000
[23]

Llm+ p: Empowering large language models with optimal planning proficiency.arXiv preprint arXiv:2304.11477, 2023

Bo Liu, Yuqian Jiang, Xiaohan Zhang, Qiang Liu, Shiqi Zhang, Joydeep Biswas, and Peter Stone. Llm+ p: Empowering large language models with optimal planning proficiency.arXiv preprint arXiv:2304.11477, 2023

2023 arXiv
[24]

Unified planning: Modeling, manipulating and solving ai planning problems in python.SoftwareX, 29:102012, 2025

Andrea Micheli, Arthur Bit-Monnot, Gabriele R¨ oger, Enrico Scala, Alessandro Valentini, Luca Framba, Alberto Rovetta, Alessandro Trapasso, Luigi Bonassi, Alfonso Emilio Gerevini, Luca Iocchi, Felix In- grand, Uwe K¨ ockemann, Fabio Patrizi, Alessandro Saetti, Ivan Serina, and...

2025
[25]

On learning action costs from input plans.arXiv preprint arXiv:2408.10889, 2024

Marianela Morales, Alberto Pozanco, Giuseppe Canonaco, Sriram Gopalakrishnan, Daniel Borrajo, and Manuela Veloso. On learning action costs from input plans.arXiv preprint arXiv:2408.10889, 2024

2024 arXiv
[26]

Nocturne: A scalable driving benchmark for bringing multi-agent learning one step closer to the real world

Roozbeh Mottaghi, Chuang Gan Wu, and Ali Farhadi. Nocturne: A scalable driving benchmark for bringing multi-agent learning one step closer to the real world. InConference on Robot Learning, pages 647–656. PMLR, 2020

2020
[27]

Aha, and Elizabeth Carter

H´ ector Mu˜ noz-Avila, Ulit Jaidee, David W. Aha, and Elizabeth Carter. Goal-driven autonomy with case-based reasoning. InProceedings of Case-Based Reasoning. Research and Development, 18th In- ternational Conference on Case-Based Reasoning (ICCBR) 2010, volume 6176 ofLecture...

2010
[28]

Pandurang Nayak, Barney Pell, and Brian C

Nicola Muscettola, P. Pandurang Nayak, Barney Pell, and Brian C. Williams. Remote agent: To boldly go where no AI system has gone before.Artificial Intelligence, 103(1-2):5–47, 1998

1998
[29]

Using copilot in microsoft 365

Anand Narayanaswamy. Using copilot in microsoft 365. InMicrosoft Copilot for Windows 11: Under- standing the AI-Powered Features in Windows 11, pages 205–233. Springer, 2024

2024
[30]

Large language models as planning domain generators

James Oswald, Kavitha Srinivas, Harsha Kokel, Junkyu Lee, Michael Katz, and Shirin Sohrabi. Large language models as planning domain generators. In34th International Conference on Automated Plan- ning and Scheduling, 2024

2024
[31]

On the prospects of incorporating large language models (llms) in automated planning and scheduling (aps).arXiv preprint arXiv:2401.02500, 2024

Vishal Pallagani, Kaushik Roy, Bharath Muppasani, Francesco Fabiano, Andrea Loreggia, Keerthiram Murugesan, Biplav Srivastava, Francesca Rossi, Lior Horesh, and Amit Sheth. On the prospects of incorporating large language models (llms) in automated planning and scheduling (aps...

2024 arXiv
[32]

Using online planning and acting to recover from cyberattacks on software-defined networks

Sunandita Patra, Alex Velazquez, Myong Kang, and Dana Nau. Using online planning and acting to recover from cyberattacks on software-defined networks. InProceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 15377–15384, 2021

2021
[33]

Generating replanning goals through multi- objective optimization in response to execution observation

Alberto Pozanco, Daniel Borrajo, and Manuela Veloso. Generating replanning goals through multi- objective optimization in response to execution observation. InProceedings of ECAI, Krak´ ow (Poland), 2023

2023
[34]

Learning-driven goal generation.AI Commu- nications, 31(2):137–150, 2018

Alberto Pozanco, Susana Fern´ andez, and Daniel Borrajo. Learning-driven goal generation.AI Commu- nications, 31(2):137–150, 2018. https://content.iospress.com/articles/ai-communications/aic754, DOI: 10.3233/AIC-180754

2018 doi
[35]

Computing planning centroids and minimum covering states using symbolic bidirectional search

Alberto Pozanco, ´Alvaro Torralba, and Daniel Borrajo. Computing planning centroids and minimum covering states using symbolic bidirectional search. InProceedings of ICAPS, Banff (Canada), 2024

2024
[36]

Adapt: As-needed decomposition and planning with language models, 2024

Archiki Prasad, Alexander Koller, Mareike Hartmann, Peter Clark, Ashish Sabharwal, Mohit Bansal, and Tushar Khot. Adapt: As-needed decomposition and planning with language models, 2024

2024
[37]

Integrating plan- ning and scheduling in workflow domains.Expert Systems with Applications, 33(2):389–406, October 2007

Mar ´ ıa Dolores Rodr ´ ıguez-Moreno, Daniel Borrajo, Amedeo Cesta, and Angelo Oddi. Integrating plan- ning and scheduling in workflow domains.Expert Systems with Applications, 33(2):389–406, October 2007. 20

2007
[38]

An AI planning- based tool for scheduling satellite nominal operations.AI Magazine, 25(4):9–27, Winter 2004

Mar ´ ıa Dolores Rodr ´ ıguez-Moreno, Daniel Borrajo, and Daniel Meziat. An AI planning- based tool for scheduling satellite nominal operations.AI Magazine, 25(4):9–27, Winter 2004. http://www.aaai.org/ojs/index.php/aimagazine/article/viewArticle/1782

2004
[39]

Twostep: Multi-agent task planning using classical planners and large language models, 2024

Ishika Singh, David Traum, and Jesse Thomason. Twostep: Multi-agent task planning using classical planners and large language models, 2024

2024
[40]

Automated composition of semantic web services into executable processes

Paolo Traverso and Marco Pistore. Automated composition of semantic web services into executable processes. InInternational Semantic Web Conference, pages 380–394. Springer, 2004

2004
[42]

Large language models still can’t plan (a benchmark for llms on planning and reasoning about change).arXiv preprint arXiv:2206.10498, 2022

Karthik Valmeekam, Alberto Olmo, Sarath Sreedharan, and Subbarao Kambhampati. Large language models still can’t plan (a benchmark for llms on planning and reasoning about change).arXiv preprint arXiv:2206.10498, 2022

2022 arXiv
[43]

Chain-of-thought prompting elicits reasoning in large language models, 2023

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models, 2023

2023
[44]

Griffiths, Yuan Cao, and Karthik Narasimhan

Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models, 2023

2023
[45]

React: Synergizing reasoning and acting in language models, 2023

Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models, 2023

2023
[46]

type": "text

Li Zhang, Mei Wang, and Hui Chen. Structured representations for natural language to planning problem translation.IEEE Transactions on Neural Networks and Learning Systems, 33(5):2003–2015, 2022. 10 Appendix 1 10.1 Complete prompt used to call the LLM We include here an exampl...

2003

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Ai-Chang, J

M. Ai-Chang, J. Bresina, L. Charest, A. Chase, J.C.-J. Hsu, A. Jonsson, B. Kanefsky, P. Morris, K. Rajan, J. Yglesias, B.G. Chafin, W.C. Dias, and P.F. Maldague. MAPGEN: Mixed-initiative planning and scheduling for the Mars Exploration Rover mission.IEEE Intelligent Systems, 19(1):8–12, feb 2004

work page 2004

[2] [2]

Learning action models with minimal ob- servability.Artificial Intelligence, 275:104–137, 2019

Diego Aineto, Sergio Jim´ enez Celorrio, and Eva Onaindia. Learning action models with minimal ob- servability.Artificial Intelligence, 275:104–137, 2019

work page 2019

[3] [3]

Daniel Borrajo and Manuela M. Veloso. Computing opportunities to augment plans for novel replanning during execution. InProceedings of ICAPS, Guangzhou (China), 2021

work page 2021

[4] [4]

Language models are few- shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

Tom B Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few- shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

work page 1901

[5] [5]

A multi-agent architecture for intelligent gathering systems.AI Communications, 18(1):15–32, 2005

David Camacho, Ricardo Aler, Daniel Borrajo, and Jos´ e Manuel Molina. A multi-agent architecture for intelligent gathering systems.AI Communications, 18(1):15–32, 2005

work page 2005

[6] [6]

Planning for tourism routes using social networks.Expert Systems with Applications, 69:1–9, 2017

Isabel Cenamor, Sergio N´ u˜ nez, Tom´ as de la Rosa, and Daniel Borrajo. Planning for tourism routes using social networks.Expert Systems with Applications, 69:1–9, 2017

work page 2017

[7] [7]

TRIP- PAL: Travel planning with guarantees by combining large language models and automated planners

Tom´ as de la Rosa, Sriram Gopalakrishnan, Alberto Pozanco, Zhen Zeng, and Daniel Borrajo. TRIP- PAL: Travel planning with guarantees by combining large language models and automated planners. arXiv e-prints, abs/2406.10196, 2024

arXiv 2024

[8] [8]

Etzioni, S

0. Etzioni, S. Hanks, D. Weld, D. Draper, N. Lesh, , and M. Williamson. An approach to planning with incomplete information. InProc. 3rd Int. Conf. on Principles of Knowledge Representation and Reasoning, 1992. 17 Initial State: (in dataframe1 data-file1) (in dataframe2 data-file2) (= (database-cost database1) 1) read-data(ai, dataframe1, data-file1, data...

work page 1992