Pith. sign in

REVIEW 4 major objections 6 minor 34 references

Achieving Scalable Robot Autonomy via neurosymbolic planning using lightweight local LLM

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a 1.5-billion-parameter local language model, fine-tuned on systematically generated PDDL training data, produces formally valid plans in 66.1% of single-domain tests and 70.6% of multi-domain tests.

desk verdict A credible feasibility study with a genuinely useful problem generator; the planning validity numbers are real but only demonstrate in-distribution imitation, not yet scalable general planning. read the letter →

arxiv 2505.08492 v2 pith:WJJSA54F submitted 2025-05-13 cs.AI cs.LGcs.RO

Nicholas Attolino , Alessio Capitanelli , Fulvio Mastrogiovanni This is my paper · ORCID
classification cs.AIcs.LGcs.RO
keywords neurosymbolicplanningPDDLlargelanguagemodelstaskdomain-problem-plangenerationhuman-robotcollaborationfine-tuninglocalLLM
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 argues that robot task planning can be pushed onto small language models that run on local hardware, removing dependence on cloud APIs. The Gideon framework generates large volumes of realistic symbolic planning problems from a PDDL domain description, fine-tunes a 1.5-billion-parameter model on the resulting domain-problem-plan triples, and reports that the model produces formally valid plans in 66.1% of single-domain test cases and 70.6% of mixed-domain test cases. The authors frame the result as proof of feasibility and scalability for on-device neurosymbolic planning, not as a claim of superiority over traditional or larger planners.

What carries the argument

The load-bearing mechanism is the Domain-Problem Generation Configuration (DPGC), a JSON schema that controls random generation of solvable PDDL problem instances from a domain, together with the pipeline that converts generated problems into plans with a classical planner, validates them with the VAL validator, and packs them into instruction-tuning examples where the domain is the instruction, the problem is the input, and the plan is the output. This machinery is what makes large-scale multi-domain training data cheap and unique, and the extended context window of the local model is what allows the full domain and problem to appear together in each example, enabling multi-domain generalization.

What would settle it

Run the same trained model on hand-written test problems taken from a real human-robot collaboration, or on a previously unseen PDDL domain the DPGC was not configured for, and compare plan validity; if validity falls well below the 66-71% range while generated-problem validity stays high, the generator's difficulty distribution is not representative and the scalability claim does not transfer.

Watch

Extended reading notes

Core claim

The central claim is that a 1.5-billion-parameter, locally deployable language model can learn to produce formally valid PDDL plans when fine-tuned on large quantities of systematically generated domain-problem-plan triples. Concretely, the paper reports 66.1% valid plans for the best single-domain model trained on 32,000 samples, and 70.6% valid plans for a multi-domain model trained on 16,000 samples from two related articulated-object manipulation domains, with 83.0% validity on one domain alone. The framework is not claimed to beat traditional planners or larger models on speed or sample efficiency; rather, the paper claims it demonstrates feasibility and scalability of on-device neurosymbolic planning with a model about 120 times smaller than the cloud-hosted baseline.

Load-bearing premise

The load-bearing premise is that the user-configured problem generator produces problems whose difficulty and structure match real robot planning tasks, so the validity measured on generated test sets transfers to actual human-robot collaboration.

Editorial extensions

If this is right

  • A model roughly 120 times smaller than the older cloud baseline can produce valid plans on this class of articulated-object manipulation tasks, making on-device planning practical where network access is unreliable.
  • Because validity jumped from 38.0% at 16k samples to 66.1% at 32k samples, scaling the generated dataset further should continue to raise the validity ceiling.
  • Multi-domain training appears to act as a regularizer: the 16k multi-domain model reached 83.0% on the single NO-MACRO test set, higher than any single-domain model, so generalist planning models may be easier to train than specialists.
  • Including the full PDDL domain in the context, which was not possible with older 2048-token models, is what enables the multi-domain setup, and the same mechanism should extend to more domains as context windows grow.
  • An action-by-action generation strategy would cut the 32k model's estimated planning time to around 1.6 seconds, beating the traditional planner on responsiveness.

Reading between the lines

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

  • Because the multi-domain model beat every single-domain model on NO-MACRO despite seeing only half as much NO-MACRO data, data variety may be a cheaper scaling axis than data volume; testing on domains that share no actions or predicates would show whether the effect is general.
  • The DPGC's probability and mutex knobs make difficulty an input parameter, so Gideon could be used to study curriculum learning for planning: training on easy problems first and increasing plan length or mutex density may cut the data needed to reach valid planning.
  • The reported validity measures formal plan correctness, not interaction fluency; an end-to-end test with a real robot and a human would show whether faster first-action availability actually improves human-robot collaboration.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper presents Gideon, a framework for neurosymbolic PDDL planning using locally deployed lightweight LLMs. It contributes a DPGC-based problem generator that produces domain-problem-plan tuples for PDDL 2.1 domains, a pipeline that fine-tunes Qwen-2.5 1.5B on Alpaca-formatted tuples, and evaluations on two articulated-object manipulation domains (MACRO and NO-MACRO). Single-domain models fine-tuned on 8k, 16k, and 32k samples achieve 39.9%, 38.0%, and 66.1% VAL-valid plans on a held-out 1,000-problem test set; a multi-domain 16k model reaches 70.6% validity on a mixed test set and 83.0% on the NO-MACRO subset. The authors position these results as evidence that lightweight local LLMs can support scalable multi-domain planning, using Teriyaki (GPT-3) and Probe as baselines.

Significance. If the transfer assumption holds, the paper's main contribution is a practical, open pipeline for generating large PDDL training corpora and evidence that a 1.5B-parameter model can be fine-tuned to produce mostly valid plans, roughly 120x smaller than the GPT-3 baseline. Strengths include the public code and data, external VAL validation, held-out and deduplicated test sets, and honest acknowledgment that comparison with Teriyaki is against published numbers. The significance is currently limited by the fact that all test problems are generated from the same DPGC configuration as training, and by results being point estimates from single runs; the broader human-robot collaboration scalability claim therefore remains conditional.

major comments (4)
  1. [Section 3.3 / Section 4.1 (Tables 1-2)] The central transfer claim is not yet empirically supported. Section 3.3 asserts that 'the planner cannot distinguish between generated and real-world data,' but this is a representational statement, not evidence about difficulty distribution. All evaluation problems are sampled from the same DPGC configuration used for training. Table 1 already contradicts full generality: Gideon plans never have fewer than 15 steps, while Probe solves trivial instances in 2 steps. To support the title's 'Scalable Robot Autonomy' and the human-robot collaboration motivation, the authors should evaluate on out-of-distribution problems such as different object counts, different mutex and invariant configurations, IPC-style instances, or hand-designed minimal problems, and report validity there. Without such tests, the headline 66.1% and 70.6% figures may reflect imitation of the generated distribution rather than planning competence.
  2. [Section 4.1 (Figure 2, Table 1)] All validity numbers are point estimates from a single fine-tuned model and a single test set. No confidence intervals, multiple seeds, or significance tests are reported. The scaling claim is weakened by the non-monotonic sequence 39.9% (8k), 38.0% (16k), 66.1% (32k); the 32k improvement rests on one run. Please provide at least three seeds per configuration, or bootstrap confidence intervals on plan validity, and a test of whether the observed differences are significant given the 1,000-sample test set.
  3. [Section 4.2 (Table 3)] The claimed benefit of multi-domain training is confounded. Gideon-MD-16k is trained on 8,000 NO-MACRO and 8,000 MACRO samples, but no single-domain MACRO model is trained. Its 83.0% NO-MACRO validity is compared against single-domain NO-MACRO models, which does not isolate the effect of 'regularization by data variety' from the effect of including MACRO data or from domain-specific difficulty. Without a single-domain 16k MACRO baseline and a matched single-domain NO-MACRO run, the multi-domain improvement cannot be attributed to data variety; please either add the missing baseline or explicitly frame the multi-domain result as a proof-of-concept without causal attribution.
  4. [Section 3.3] The DPGC format is a central contribution, but the paper explicitly leaves 'an overly detailed description' out of scope and only sketches probabilities, pools, mutexes, and tagged references. Because the full schema is only in a repository, the exact difficulty distribution of the generated problems cannot be assessed from the manuscript. At a minimum, include the DPGC files used for MACRO and NO-MACRO as supplementary material in the paper, or provide a complete specification in an appendix, so that reviewers and readers can reproduce the data distribution and judge its representativeness.
minor comments (6)
  1. [Section 2] In the paragraph beginning 'Neurosymbolic approaches are an alternative class,' the phrase 'the promise to to address' contains a duplicated 'to'.
  2. [Section 3.2 (Figure 1 caption)] The caption refers to 'DPCG file' while the text consistently uses 'DPGC'; please unify the acronym.
  3. [Section 4.2 (Tables 3 and 4)] Decimal commas are used inconsistently (e.g., '70,6', '39,9' versus '70.6', '39.9' in the text); please use a single convention throughout.
  4. [Section 4.1] The statement that 66.1% represents a '75% increase' over the 16k model is slightly off; 66.1/38.0 is approximately 1.74. Please write 'approximately 74%' or provide the exact computation.
  5. [Section 5 / overall structure] The introduction promises a Section V titled 'Limitations,' but no such section exists; limitations appear only as a brief paragraph in the Conclusion. Please add a dedicated limitations section or correct the roadmap.
  6. [Section 3.4 / Section 4.1 (Table 2)] The hardware description lists an NVIDIA RTX A6000, yet average planning times exceed 60 seconds for a 1.5B model; please state explicitly whether inference ran on GPU or CPU and whether any quantization was used, as this affects the timing interpretation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: results are measured accuracies on held-out generated problems, and self-citations are baselines rather than derivation inputs.

full rationale

The central numbers (39.9%, 38.0%, 66.1%, 70.6%, 83.0%, 58.2%) are empirical planning-validity rates obtained by fine-tuning Qwen-2.5 1.5B on generated domain-problem-plan tuples and then testing on held-out instances, with the paper stating that 'samples were equally split between domains' and that leakage was avoided because 'it is guaranteed that all data points are unique.' The test accuracy is not a fitted parameter renamed as a prediction; it is a measured generalization outcome. The test problems come from the same DPGC configuration as training, so the results may overstate out-of-distribution transfer, but that is a generalization limitation rather than a circular reduction. The Section 3.3 assertion that 'the planner cannot distinguish between generated and real-world data' is a stated modeling premise about symbolic planning with assumed perfect perception, not a conclusion forced by the framework's own definitions. Self-citations to Teriyaki [5] and the articulated-object domains [4] are used as published baselines and benchmark definitions, and the projected 1.6 s planning-time estimate cites a previously published linear-scaling relation that is externally testable rather than being derived from the present fitted values. No self-definitional step, no renamed fit, no imported uniqueness theorem, and no ansatz smuggled via citation were found.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The loads are modest: standard PDDL planning assumptions, user-configured generation rules, and the assumption that generated data matches real planning conditions. No new physical or metaphysical entities are introduced.

free parameters (3)
  • Training dataset size = 8k, 16k, 32k samples
    The number of generated examples per model is chosen by hand and is the main axis of the scaling claim; the paper reports only three discrete points, with a non-monotonic trend.
  • DPGC predicate probabilities and object pool sizes = Not specified in text
    The generator's randomness is governed by user-set probabilities and pool sizes, which shape the data distribution and hence the measured validity; the paper points to GitHub rather than listing them.
  • Hyperparameters (learning rate, epochs, temperature, token cutoff) = 5e-5, 4 epochs, 0.01, 3096 tokens
    These are hand-chosen and standard, but the non-monotonic 8k/16k results suggest sensitivity to them that is not explored.
assumptions (4)
  • domain assumption Generated problems are equivalent to real problems for planning purposes (Section 3.3).
    Used to justify training and testing only on generated data, so that in-distribution validity is assumed to reflect real-world validity.
  • standard math VAL is a correct validator of PDDL 2.1 plan validity (Section 3.2).
    All validity numbers depend on VAL, a standard tool treated as ground truth.
  • domain assumption The two domains used, MACRO and NO-MACRO, share enough structure that multi-domain training effects can be attributed to data variety (Sections 3.1 and 4.2).
    The authors note 'some transfer learning occurs' and call the multi-domain test a feasibility proof, so the 'data variety helps' claim rests on this similarity.
  • domain assumption Probe planner provides sound training plans (Section 3.1).
    Training targets are generated by Probe; if Probe produced invalid or suboptimal plans, fine-tuning would learn those, though VAL validation filters invalid ones.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Achieving Scalable Robot Autonomy via neurosymbolic planning using lightweight local LLM." pith.science (2026). https://pith.science/paper/WJJSA54F

@misc{pith2026250508492,
  author       = {Pith},
  title        = {Pith review of: Achieving Scalable Robot Autonomy via neurosymbolic planning using lightweight local LLM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WJJSA54F}},
  note         = {Machine review of arXiv:2505.08492}
}
read the original abstract

PDDL-based symbolic task planning remains pivotal for robot autonomy yet struggles with dynamic human-robot collaboration due to scalability, re-planning demands, and delayed plan availability. Although a few neurosymbolic frameworks have previously leveraged LLMs such as GPT-3 to address these challenges, reliance on closed-source, remote models with limited context introduced critical constraints: third-party dependency, inconsistent response times, restricted plan length and complexity, and multi-domain scalability issues. We present Gideon, a novel framework that enables the transition to modern, smaller, local LLMs with extended context length. Gideon integrates a novel problem generator to systematically generate large-scale datasets of realistic domain-problem-plan tuples for any domain, and adapts neurosymbolic planning for local LLMs, enabling on-device execution and extended context for multi-domain support. Preliminary experiments in single-domain scenarios performed on Qwen-2.5 1.5B and trained on 8k-32k samples, demonstrate a valid plan percentage of 66.1% (32k model) and show that the figure can be further scaled through additional data. Multi-domain tests on 16k samples yield an even higher 70.6% planning validity rate, proving extensibility across domains and signaling that data variety can have a positive effect on learning efficiency. Although long-horizon planning and reduced model size make Gideon training much less efficient than baseline models based on larger LLMs, the results are still significant considering that the trained model is about 120x smaller than baseline and that significant advantages can be achieved in inference efficiency, scalability, and multi-domain adaptability, all critical factors in human-robot collaboration. Training inefficiency can be mitigated by Gideon's streamlined data generation pipeline.

Figures

Figures reproduced from arXiv: 2505.08492 by the authors.

Figure 1
Figure 1. A diagram of Gideon’s training pipeline. In order, selected domains are pro￾cessed in accordance to the rules specified in the DPCG file to generate a set amount of problem instance; then plans are generated by the Plan Generator module with the se￾lected planner; finally, domain-problem-plans tuples are assembled and formatted into the training, validation and test set files. The core of the Gideon’s framework is t… view at source ↗
Figure 2
Figure 2. Planning validity achieved by Gideon-NO-MACRO and Gideon-MD with re￾spect to increasing training steps. Gideon-NO-MACRO-8k, -16k, and -32k are reported in orange. For Gideon MD, both results on the multi-domain test set (in green) and on a second single-domain test set (NO-MACRO, in blue) are reported to facilitate comparison with single-domain models. The red line above represent the success rate of the baseline tr… view at source ↗
Figure 3
Figure 3. Box plots of the planning times of Gideon-NO-MACRO-16k, Gideon-MD-16k tested on 500 NO-MACRO samples, Gideon-MD-16k tested on the complete multi￾domain test set, and the baseline traditional planner Probe. planning. Our work was motivated by the critical need for interaction fluency in dynamic environments while overcoming the constraints of traditional symbolic planners and cloud-based neurosymbolic solutions. At i… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 20 canonical work pages

  1. [1]

    arXiv preprint arXiv:2404.14219 (2024)

    Abdin, M., Aneja, J., Awadalla, H., Awadallah, A., Awan, A.A., Bach, N., Bahree, A., Bakhtiari, A., Bao, J., Behl, H., et al.: Phi-3 technical report: A highly capable language model locally on your phone. arXiv preprint arXiv:2404.14219 (2024)

    arXiv 2024

  2. [2]

    Transactions on Machine Learning Research (2024)

    Biderman, D., Portes, J., Ortiz, J.J.G., Paul, M., Greengard, P., Jennings, C., King, D., Havens, S., Chiley, V., Frankle, J., et al.: Lora learns less and forgets less. Transactions on Machine Learning Research (2024)

    work page 2024

  3. [3]

    Advances in neural information processing systems33, 1877–1901 (2020)

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language models are few-shot learners. Advances in neural information processing systems33, 1877–1901 (2020)

    2020

  4. [4]

    Robotics and Autonomous Systems 109, 139–155 (2018)

    Capitanelli, A., Maratea, M., Mastrogiovanni, F., Vallati, M.: On the manipula- tion of articulated objects in human–robot cooperation scenarios. Robotics and Autonomous Systems 109, 139–155 (2018)

    work page 2018

  5. [5]

    Frontiers in Neu- rorobotics 18 (Jun 2024)

    Capitanelli, A., Mastrogiovanni, F.: A framework for neurosymbolic robot action planning using large language models. Frontiers in Neu- rorobotics 18 (Jun 2024). https://doi.org/10.3389/fnbot.2024.1342786, http://dx.doi.org/10.3389/fnbot.2024.1342786 Achieving Scalable Robot Autonomy via neurosymbolic planning 17

    arXiv 2024

  6. [6]

    Data in Brief22, 119–117 (2019)

    Carfí, A., Foglino, F., Bruno, B., Mastrogiovanni, F.: A multi-sensor dataset for human-human handover. Data in Brief22, 119–117 (2019)

    work page 2019

  7. [7]

    arXiv preprint arXiv:2107.03374 (2021)

    Chen, M., Tworek, J., Jun, H., Yuan, Q., Pinto, H.P.d.O., Kaplan, J., Edwards, H., Burda, Y., Joseph, N., Brockman, G., et al.: Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374 (2021)

    arXiv 2021

  8. [8]

    IEEE Transactions on Robotics 37(2), 567–586 (2021)

    Darvish, K., Simetti, E., Mastrogiovanni, F., Casalino, G.: A hierarchical archi- tecture for human-robot cooperation processes. IEEE Transactions on Robotics 37(2), 567–586 (2021)

    work page 2021

Show all 34 references
  1. [9]

    Mecha- tronics 51, 97–115 (2018)

    Darvish, K., Wanderlingh, F., Bruno, B., Simetti, E., Mastrogiovanni, F., Casalino, G.: Flexible human-robot cooperation models for assisted shop-floor tasks. Mecha- tronics 51, 97–115 (2018)

    2018

  2. [10]

    Autonomous Robots 47(8), 981–997 (Aug 2023)

    Ding, Y., Zhang, X., Amiri, S., Cao, N., Yang, H., Kaminski, A., Esselink, C., Zhang, S.: Integrating action knowledge and llms for task planning and situation handling in open worlds. Autonomous Robots 47(8), 981–997 (Aug 2023). https://doi.org/10.1007/s10514-023-10133-5, htt...

    2023 doi

  3. [11]

    arXiv preprint arXiv:2012.05876 (2020)

    Garcez, A.d., Lamb, L.C.: Neurosymbolic ai: the 3rd wave. arXiv preprint arXiv:2012.05876 (2020)

    2020 arXiv

  4. [12]

    In: Proc

    Garrett, C.R., Lozano-Perez, T., Kaelbing, L.P.: PDDLstream: integrating sym- bolic planners and blackbox samplers with optimistic adaptive planning. In: Proc. 30th International Conference on Automated Planning and Scheduling (ICAPS). Anywhere on Earth (October 2020)

    2020

  5. [13]

    Robotica 42(4), 1094–1112 (Feb 2024)

    Ghafarian Tamizi, M., Honari, H., Nozdryn-Plotnicki, A., Naj- jaran, H.: End-to-end deep learning-based framework for path plan- ning and collision checking: bin-picking application. Robotica 42(4), 1094–1112 (Feb 2024). https://doi.org/10.1017/s0263574724000109, http://dx.doi...

    2024 doi

  6. [14]

    Ghallab, M., Knoblock, C., Wilkins, D., Barrett, A., Christianson, D., Friedman, M., Kwok, C., Golden, K., Penberthy, S., Smith, D., Sun, Y., Weld, D.: Pddl - the planning domain definition language (08 1998)

    1998

  7. [15]

    arXiv preprint arXiv:2103.03874 (2021)

    Hendrycks, D., Burns, C., Kadavath, S., Arora, A., Basart, S., Tang, E., Song, D., Steinhardt, J.: Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874 (2021)

    2021 arXiv

  8. [16]

    In: 2010 ieee/rsj international conference on intelligent robots and systems

    Heyer, C.: Human-robot interaction and future industrial robotics applications. In: 2010 ieee/rsj international conference on intelligent robots and systems. pp. 4749–4754. IEEE (2010)

    2010

  9. [17]

    IEEE Transactions on Human-Machine Systems49(3), 209–218 (2019)

    Hoffman, G.: Evaluating fluency in human–robot collaboration. IEEE Transactions on Human-Machine Systems49(3), 209–218 (2019)

    2019

  10. [18]

    In: 16th IEEE International Conference on Tools with Artificial Intelligence

    Howey, R., Long, D., Fox, M.: Val: Automatic plan validation, continuous effects and mixed initiative planning using pddl. In: 16th IEEE International Conference on Tools with Artificial Intelligence. pp. 294–301. IEEE (2004)

    2004

  11. [19]

    In: 2024 IEEE Interna- tional Conference on Robotics and Automation (ICRA)

    Joublin, F., Ceravola, A., Smirnov, P., Ocker, F., Deigmoeller, J., Belardinelli, A., Wang, C., Hasler, S., Tanneberg, D., Gienger, M.: Copal: Corrective plan- ning of robot actions with large language models. In: 2024 IEEE Interna- tional Conference on Robotics and Automation...

    2024

  12. [20]

    arXiv preprint arXiv:2411.19886 (2024)

    Khandelwal, V., Sheth, A., Agostinelli, F.: Pddlfuse: A tool for generating diverse planning domains. arXiv preprint arXiv:2411.19886 (2024)

    2024 arXiv

  13. [21]

    In: Proceedings of the International Conference on Automated Planning and Schedul- ing

    Lipovetzky, N., Geffner, H.: Searching for plans with carefully designed probes. In: Proceedings of the International Conference on Automated Planning and Schedul- ing. vol. 21, pp. 154–161 (2011) 18 N. Attolino et al

    2011

  14. [22]

    In: The Eleventh International Conference on Learning Representations (2023), https://openreview.net/forum?id=iOc57X9KM54

    Lu, Y., Feng, W., Zhu, W., Xu, W., Wang, X.E., Eckstein, M., Wang, W.Y.: Neuro-symbolic procedural planning with commonsense prompting. In: The Eleventh International Conference on Learning Representations (2023), https://openreview.net/forum?id=iOc57X9KM54

    2023

  15. [23]

    In: 8th International Conference on Intelligent Autonomous Systems (2004)

    Mastrogiovanni, F., Sgorbissa, A., Zaccaria, R.: A system for hierarchical plan- ning in service mobile robotics. In: 8th International Conference on Intelligent Autonomous Systems (2004)

    2004

  16. [24]

    Intelligent Service Robotics13, 439–457 (2020)

    Murali, P.K., Darvish, K., Mastrogiovanni, F.: Deployment and evaluation of a flex- ible human-robot collaboration model based on and/or graphs in a manufacturing environment. Intelligent Service Robotics13, 439–457 (2020)

    2020

  17. [25]

    Pallagani, V., Muppasani, B., Murugesan, K., Rossi, F., Horesh, L., Srivastava, B., Fabiano, F., Loreggia, A.: Plansformer: Generating symbolic plans using trans- formers (2022), https://arxiv.org/abs/2212.08681

    2022 arXiv

  18. [26]

    Proceedings of the International Conference on Automated Planning and Scheduling 34, 432–444 (May 2024)

    Pallagani, V., Muppasani, B.C., Roy, K., Fabiano, F., Loreggia, A., Murugesan, K., Srivastava, B., Rossi, F., Horesh, L., Sheth, A.: On the prospects of incor- porating large language models (llms) in automated planning and scheduling (aps). Proceedings of the International Co...

    2024 doi

  19. [27]

    Proceedings of the International Conference on Automated Planning and Scheduling 34(1), 500–508 (May 2024)

    Rossetti, N., Tummolo, M., Gerevini, A.E., Putelli, L., Serina, I., Chiari, M., Olivato, M.: Learning general policies for planning through gpt mod- els. Proceedings of the International Conference on Automated Planning and Scheduling 34(1), 500–508 (May 2024). https://doi.org...

    2024 doi

  20. [28]

    Song, C.H., Wu, J., Washington, C., Sadler, B.M., Chao, W.L., Su, Y.: Llm- planner: Few-shot grounded planning for embodied agents with large language models (2023), https://arxiv.org/abs/2212.04088

    2023 arXiv

  21. [29]

    arXiv preprint arXiv:2503.18971 (2025)

    Tantakoun, M., Zhu, X., Muise, C.: Llms as planning modelers: A survey for leveraging large language models to construct automated planning models. arXiv preprint arXiv:2503.18971 (2025)

    2025

  22. [30]

    Team, Q.: Qwen2.5: A party of foundation models (September 2024), https://qwenlm.github.io/blog/qwen2.5/

    2024

  23. [31]

    arXiv preprint arXiv:2307.09288 (2023)

    Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bash- lykov, N., Batra, S., Bhargava, P., Bhosale, S., et al.: Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288 (2023)

    2023 arXiv

  24. [32]

    In: NeurIPS 2022 Foundation Models for Decision Making Workshop (2022), https://openreview.net/forum?id=wUU-7XTL5XO

    Valmeekam, K., Olmo, A., Sreedharan, S., Kambhampati, S.: Large language mod- els still can’t plan (a benchmark for LLMs on planning and reasoning about change). In: NeurIPS 2022 Foundation Models for Decision Making Workshop (2022), https://openreview.net/forum?id=wUU-7XTL5XO

    2022

  25. [33]

    Wang, L., Xu, W., Lan, Y., Hu, Z., Lan, Y., Lee, R.K.W., Lim, E.P.: Plan-and- solve prompting: Improving zero-shot chain-of-thought reasoning by large language models (2023), https://arxiv.org/abs/2305.04091

    2023 arXiv

  26. [34]

    In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)

    Zheng, Y., Zhang, R., Zhang, J., Ye, Y., Luo, Z., Feng, Z., Ma, Y.: Llamafac- tory: Unified efficient fine-tuning of 100+ language models. In: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations). Association...

    2024 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.