Pith. sign in

REVIEW 2 major objections 4 minor 34 references

A trained planner turns terse household instructions into symbolic subgoals and reaches 89% success on abstract tasks, sharply ahead of prior methods.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 06:00 UTC pith:53N4BT7I

load-bearing objection Solid new training recipe for LLM+PDDL, but the headline success rates are computed with the same LLM judge that supplies the training reward, so the numbers are not trustworthy as reported. the 2 major comments →

arxiv 2602.12244 v2 pith:53N4BT7I submitted 2026-02-12 cs.RO

Any House Any Task: Scalable Long-Horizon Planning for Abstract Human Tasks

Zhihong Liu , Yang Li , Rengming Huang , Cewu Lu , Panpan Cai This is my paper
classification cs.RO
keywords long-horizon planninghousehold roboticsPDDLsubgoal decompositionreinforcement learningabstract instructionsscene graphsGRPO
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that long-horizon household planning fails when language models try to generate entire action plans directly, because intention understanding, constraint satisfaction, and long-horizon reasoning get entangled. Instead, the authors propose AHAT, a model trained via reinforcement learning to decompose an abstract instruction into a sequence of PDDL subgoals, which are then solved by an off-the-shelf symbolic planner. The central claim is that this split, plus a new RL algorithm called TGPO that externally corrects the intermediate decomposition trace, lets planning scale to large scenes, long plans, and highly ambiguous instructions. Reported success rates on an abstract task suite are 89.0% versus 20.1% for the strongest prompting baseline. If true, it means a relatively small trained model plus a symbolic planner can handle human-level housework requests without end-to-end action generation.

Core claim

AHAT establishes that a language model trained to emit PDDL subgoals, rather than low-level actions, can plan household tasks from terse instructions with high success. The core discovery is that reinforcement learning benefits materially from externally fixing the intermediate decomposition trace while letting the policy propose the subgoals: TGPO's two-pass design yields 89.0% success on abstract tasks versus 20.1% for the strongest prompting baseline. The paper further reports that performance degrades only modestly as scene graphs grow, plans lengthen, instructions become more abstract, and constraints become denser, whereas baselines degrade substantially along one or more of these axes

What carries the argument

The central object is TGPO, a two-pass reinforcement learning algorithm built on GRPO. In the first pass, the policy generates candidate decomposition traces and PDDL subgoals; failed candidates are sent to an external trace improver that rewrites the natural-language decomposition. In the second pass, the corrected trace is forced into the rollout via token-level constrained sampling while the policy continues to generate subgoals autoregressively. Group-relative advantages are computed over the augmented candidate group. This gives the policy credit for producing subgoals that the improved trace makes solvable, while preserving on-policy optimization. The second key piece is the PDDL subgo

Load-bearing premise

The reported success rates are scored by the same LLM reviewer that provides the reward during training; if that reviewer is biased toward the policy's style or overly lenient, the numbers are not a reliable measure of how well a real user's request is fulfilled.

What would settle it

Run AHAT on a held-out set of abstract instructions and have human annotators judge whether each generated plan would satisfy the request, then compare those human labels with the reviewer model's scores; a large mismatch would indicate that the reported success rates reflect reviewer leniency rather than genuine task understanding.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If AHAT's central claim holds, household robots can handle natural, terse instructions by combining a small trained LLM with a symbolic planner, reducing the need for large end-to-end action-generating models.
  • Robustness to plan lengths over 30 actions and scene graphs of roughly 150 nodes suggests the decompose-then-solve design avoids the error accumulation that plagues monolithic action-sequence generation.
  • Because the policy only outputs subgoals, the same framework could transfer to new PDDL domains with additional domain-specific annotations and retraining.
  • TGPO provides a template for injecting external feedback into the reasoning traces of any LLM whose outputs are evaluated by a verifier, which may extend beyond planning to other structured reasoning tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The reported success rates rely on the same LLM reviewer that provides training rewards; if that reviewer is lenient toward the policy's output style, the absolute numbers may overstate real-world performance, and an independent human evaluation would be needed to confirm the gains.
  • The method assumes a closed, predefined PDDL domain and full observability, as the paper acknowledges; a natural extension is to learn domain constraints from raw sensor data, which could open partially observable settings.
  • TGPO's trace correction could be applied beyond planning: any task where an intermediate chain-of-thought can be improved by an external checker (for example, proof steps or code reasoning) might benefit from the same constrained-sampling RL loop.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper proposes AHAT, a long-horizon household task planner built on a Qwen2.5-7B LLM that first generates a natural-language decomposition trace and then grounds each subtask into PDDL subgoals, which are solved by an off-the-shelf PDDL planner. The model is trained with SFT on a large synthetic dataset and then with TGPO, an extension of GRPO that uses an external 'trace improver' to correct failed decomposition traces and regenerates subgoals under the corrected trace. The authors claim large gains over prompting-based, general-purpose LLM, and learning-based baselines on a held-out synthetic benchmark, human-authored tasks, and two public benchmarks (PARTNR and Behavior-1K), and report scalability in environment size, plan length, abstractness, and constraints. The paper's central quantitative claims rest on an evaluation metric SR = (1/N) Σ I_feas·R_comp, where R_comp is produced by the same Qwen3-235B 'reviewer' that supplies the completion reward in TGPO training. This identity raises serious concerns about the validity of the reported success-rate figures.

Significance. If the reported numbers are trustworthy, the paper would be a significant advance: AHAT's decompose-then-solve architecture cleanly combines LLM-based intention understanding with symbolic planning, and TGPO is an interesting RL variant that incorporates external trace correction. The 50k-task dataset with persona-driven instruction diversity is a valuable resource, and the paper is transparent about its PDDL feasibility checking and planning pipeline, which are machine-checkable and reproducible in principle. However, the evaluation's dependence on the same judge used for training is a major threat to the internal validity of all reported gains. The paper's strengths in method design and dataset construction cannot be fully credited until an independent success criterion—human judgments or simulator-based ground truth—is provided.

major comments (2)
  1. [§V-A, §VI-A2] The training reward (§V-A) and evaluation metric (§VI-A2) are both R(o)=I_feas·R_comp with the same Qwen3-235B 'reviewer' producing R_comp. Since TGPO optimizes the policy against this judge and the SFT data are filtered/reviewed by the same model (§IV-B), the reported success rate is a measure of fit to this judge, not of task completion. Feasibility is objective, but completion dominates on abstract tasks (Table I: 89.0% vs 20.1% baseline). No human or simulator ground truth is reported for any benchmark. This affects every claimed gain (Tables I–IV, Fig. 3). The authors should add human-rated subsets or use the official success criteria of PARTNR/Behavior-1K.
  2. [§VI-A1] The 'Human Tasks' test set is not human-evaluated: the instructions are authored by humans, but the completion score is calculated with the same Qwen3-235B reviewer used in training. The 82.2% success rate therefore reflects the reviewer's opinion, not human satisfaction. This label is misleading and should be changed (e.g., to 'human-authored tasks'), and a genuine human evaluation should be provided to support the claim of performance on real user tasks.
minor comments (4)
  1. [Table III] The name 'E2E-TRPO' appears in Table III, while the text and Fig. 3 refer to 'E2E-TGPO'. Please fix this inconsistency.
  2. [§V-B] The trace improver model is not identified. Please state which model is used and whether it is the same as the reviewer, as this affects the interpretation of the training signal.
  3. [§VI-A2] The paper does not report how the baseline methods are evaluated with respect to the reviewer; e.g., whether the same PDDL problem construction and reviewer prompt are used for SayPlan and DELTA. Please provide details for reproducibility.
  4. [Fig. 3] The x-axis labels 'Task Abstractness' and 'Constraint Complexity' are vague; please define the precise metrics used to compute these axes.

Circularity Check

1 steps flagged

Reported success rate is the TGPO reward itself: the same Qwen3-235B reviewer supplies R_comp in both the training reward and the evaluation metric, making the completion component of every headline gain self-referential.

specific steps
  1. self definitional [Section V-A (Reward definition) and Section VI-A2 (Evaluation Metrics)]
    "For task completion, we employ an auxiliary LLM as a reviewer to assess whether the resulting action sequence fulfills the instruction. Concretely, we use QWEN3-235B [29] to review the solver-composed plan by summarizing and critiquing it w.r.t. q, and output a discrete label c∈{Bad,Normal,Good}, which we map to a scalar completion score R_comp(o)∈{0,α,1} with α=0.5. ... We define the final reward by combining the above scores: R(o)=I_feas(o)·R_comp(o). ... SR = 1/N Σ I_feas(o_i)·R_comp(o_i)."

    The completion term R_comp is the same Qwen3-235B judgment in both the TGPO training reward and the reported success-rate metric. TGPO explicitly optimizes the policy against R(o)=I_feas(o)·R_comp(o), while the paper reports success as the average of the same product with the same reviewer. The policy is therefore trained to maximize the exact scalar later presented as 'success.' Any systematic leniency or stylistic preference of Qwen3-235B toward AHAT's outputs inflates every reported success number, including public-benchmark results in Table II. Only the PDDL feasibility factor is objective; the completion half of the success rate is self-referential. This is a partial circularity because I_feas is independently verified by the symbolic planner.

full rationale

The derivation chain of AHAT is mostly self-contained: subgoal feasibility is objectively checked by a PDDL solver, the SFT/RL training loop is specified, and no load-bearing argument reduces to a self-citation. The one genuine circular step is the identity between the optimization target and the evaluation metric. In Section V-A, the TGPO reward is R(o)=I_feas(o)·R_comp(o), with R_comp produced by QWEN3-235B as the reviewer; in Section VI-A2, the reported success rate is SR=(1/N)Σ I_feas(o_i)·R_comp(o_i), using the same R_comp function. The policy is therefore explicitly trained to maximize the same scalar that is later reported as 'success.' Because the reviewer also supplies trace-improvement feedback and filters SFT annotations, AHAT is shaped to satisfy this specific judge's preferences. Consequently, the completion half of the success metric is circular: any systematic leniency or output-style bias in Qwen3-235B directly inflates all reported gains, including the public-benchmark numbers in Table II, which use the same proxy rather than simulator or human ground truth. The feasibility factor is objective, which is why this is partial rather than total circularity, and the method itself may still be sound. No other enumerated circularity pattern is present.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 0 invented entities

The method adds no new physical entities. The central assumptions are the reliability of an LLM judge that is shared between training and evaluation, and the fidelity of the synthetic task distribution; both are load-bearing and unverified.

free parameters (5)
  • alpha (completion reward weight) = 0.5
    Maps reviewer labels {Bad, Normal, Good} to {0, 0.5, 1} in R_comp (§V-A); chosen by hand.
  • tau_fail (failure threshold) = not specified
    Threshold for selecting failed candidates for trace improvement (§V-B); value not reported, affects training dynamics.
  • GRPO clipping/KL coefficients epsilon, beta = not specified
    Hyperparameters in TGPO objective (§V-B) not reported.
  • candidate group size N / N' = not specified
    Number of rollouts per group in TGPO is not reported; affects advantage estimation.
  • SFT epochs before RL = 3
    Empirically chosen warm-up duration (§III-B).
axioms (4)
  • domain assumption Qwen3-235B reviewer judgements are a valid proxy for human task completion
    Used both in RL reward (§V-A) and evaluation metric (§VI-A2); if biased, headline numbers are inflated.
  • domain assumption Scene graphs and fixed PDDL domain provide a complete, closed-world model of the environment
    The planner only sees the textual scene graph; limitations admit full observability and closed-domain assumptions (§VII).
  • domain assumption LLM-generated synthetic instructions with persona prompts are representative of real human instructions
    The 50k training set and the held-out AHAT benchmark share this distribution (§IV-A); human-task set is small (50 tasks).
  • domain assumption SFT annotations from GPT-5 Hi-PDDL with retry and scene-repair are high-quality supervision
    Used as warm-up supervision (§IV-B); scene repair injects objects, changing the data distribution.

pith-pipeline@v1.3.0-alltime-deepseek · 12265 in / 11468 out tokens · 115418 ms · 2026-08-04T06:00:13.705273+00:00 · methodology

0 comments
read the original abstract

Open world language conditioned task planning is crucial for robots operating in large-scale household environments. While many recent works attempt to address this problem using Large Language Models (LLMs) via prompting or training, a key challenge remains scalability. Performance often degrades rapidly with increasing environment size, plan length, instruction ambiguity, and constraint complexity. In this work, we propose Any House Any Task (AHAT), a household task planner optimized for long-horizon planning in large environments given ambiguous human instructions. At its core, AHAT utilizes an LLM trained to map task instructions and textual scene graphs into grounded subgoals defined in the Planning Domain Definition Language (PDDL). These subgoals are subsequently solved to generate feasible and optimal long-horizon plans through explicit symbolic reasoning. To enhance the model's ability to decompose complex and ambiguous intentions, we introduce TGPO, a novel reinforcement learning algorithm that integrates external correction of intermediate reasoning traces into Group Relative Policy Optimization (GRPO). Experiments demonstrate that AHAT achieves significant performance gains over state-of-the-art prompting, planning, and learning methods, particularly in human-style household tasks characterized by brief instructions but requiring complex execution plans.

Figures

Figures reproduced from arXiv: 2602.12244 by Cewu Lu, Panpan Cai, Rengming Huang, Yang Li, Zhihong Liu.

Figure 1
Figure 1. Figure 1: In large-scale environments, AHAT receives abstract [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of AHAT. (a) Data Generation: Task synthesis and annotation. (b) Policy Supervision: Supervised fine-tuning (SFT) on the constructed long-horizon household planning dataset. (c) Trace-Guided Policy Optimization: The reinforcement learning loop that integrates external correction of intermediate reasoning traces, improving subgoal generation and task decomposition through constrained sampling, and … view at source ↗
Figure 3
Figure 3. Figure 3: Scalability Test Results showing the variation of [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 17 linked inside Pith

  1. [1]

    Pddl—the planning domain definition language.Technical Report, Tech

    Constructions Aeronautiques, Adele Howe, Craig Knoblock, ISI Drew McDermott, Ashwin Ram, Manuela Veloso, Daniel Weld, David Wilkins Sri, Anthony Barrett, Dave Christianson, et al. Pddl—the planning domain definition language.Technical Report, Tech. Rep., 1998

    1998

  2. [2]

    Autogpt+ p: Affordance-based task planning with large language models.arXiv preprint arXiv:2402.10778, 2024

    Timo Birr, Christoph Pohl, Abdelrahman Younes, and Tamim Asfour. Autogpt+ p: Affordance-based task planning with large language models.arXiv preprint arXiv:2402.10778, 2024

    Pith/arXiv arXiv 2024

  3. [3]

    Do as i can, not as i say: Grounding language in robotic affordances

    Anthony Brohan, Yevgen Chebotar, Chelsea Finn, Karol Hausman, Alexander Herzog, Daniel Ho, Julian Ibarz, Alex Irpan, Eric Jang, Ryan Julian, et al. Do as i can, not as i say: Grounding language in robotic affordances. InConference on robot learning, pages 287–318. PMLR, 2023

    2023

  4. [4]

    Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks.arXiv preprint arXiv:2411.00081, 2024

    Matthew Chang, Gunjan Chhablani, Alexander Clegg, Mikael Dallaire Cote, Ruta Desai, Michal Hlavac, Vladimir Karashchuk, Jacob Krantz, Roozbeh Mottaghi, Priyam Parashar, et al. Partnr: A benchmark for planning and reasoning in embodied multi-agent tasks.arXiv preprint arXiv:2411.00081, 2024

    Pith/arXiv arXiv 2024

  5. [5]

    Language-augmented symbolic planner for open-world task planning.arXiv preprint arXiv:2407.09792, 2024

    Guanqi Chen, Lei Yang, Ruixing Jia, Zhe Hu, Yizhou Chen, Wei Zhang, Wenping Wang, and Jia Pan. Language-augmented symbolic planner for open-world task planning.arXiv preprint arXiv:2407.09792, 2024

    Pith/arXiv arXiv 2024

  6. [6]

    Llm-state: Open world state representation for long-horizon task planning with large language model.arXiv preprint arXiv:2311.17406, 2023

    Siwei Chen, Anxing Xiao, and David Hsu. Llm-state: Open world state representation for long-horizon task planning with large language model.arXiv preprint arXiv:2311.17406, 2023

    Pith/arXiv arXiv 2023

  7. [7]

    Dynamic planning with a llm.arXiv preprint arXiv:2308.06391, 2023

    Gautier Dagan, Frank Keller, and Alex Lascarides. Dynamic planning with a llm.arXiv preprint arXiv:2308.06391, 2023. 9

    Pith/arXiv arXiv 2023

  8. [8]

    Nl2plan: Robust llm-driven planning from minimal text descriptions.arXiv preprint arXiv:2405.04215, 2024

    Elliot Gestrin, Marco Kuhlmann, and Jendrik Seipp. Nl2plan: Robust llm-driven planning from minimal text descriptions.arXiv preprint arXiv:2405.04215, 2024

    arXiv 2024

  9. [9]

    Llm-empowered embodied agent for memory-augmented task planning in household robotics

    Marc Glocker, Peter H ¨onig, Matthias Hirschmanner, and Markus Vincze. Llm-empowered embodied agent for memory-augmented task planning in household robotics. arXiv preprint arXiv:2504.21716, 2025

    Pith/arXiv arXiv 2025

  10. [10]

    Llm- personalize: Aligning llm planners with human preferences via reinforced self-training for housekeeping robots

    Dongge Han, Trevor McInroe, Adam Jelley, Stefano V Albrecht, Peter Bell, and Amos Storkey. Llm- personalize: Aligning llm planners with human preferences via reinforced self-training for housekeeping robots. InProceedings of the 31st International Conference on Computational Linguistics, pages 1465–1474, 2025

    2025

  11. [11]

    Interpret: Interactive predicate learning from language feedback for generalizable task planning

    Muzhi Han, Yifeng Zhu, Song-Chun Zhu, Ying Nian Wu, and Yuke Zhu. Interpret: Interactive predicate learning from language feedback for generalizable task planning. arXiv preprint arXiv:2405.19758, 2024

    Pith/arXiv arXiv 2024

  12. [12]

    The fast downward planning system

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

    2006

  13. [13]

    Reinforce++: Stabilizing critic-free policy optimization with global advantage normalization, 2025.URL https://arxiv

    Jian Hu, Jason Klein Liu, Haotian Xu, and Wei Shen. Reinforce++: Stabilizing critic-free policy optimization with global advantage normalization, 2025.URL https://arxiv. org/abs/2501.03262

    Pith/arXiv arXiv 2025

  14. [14]

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

    Wenlong Huang, Pieter Abbeel, Deepak Pathak, and Igor Mordatch. Language models as zero-shot planners: Extracting actionable knowledge for embodied agents. InInternational conference on machine learning, pages 9118–9147. PMLR, 2022

    2022

  15. [15]

    Inner monologue: Embodied reasoning through planning with language models.arXiv preprint arXiv:2207.05608, 2022

    Wenlong Huang, Fei Xia, Ted Xiao, Harris Chan, Jacky Liang, Pete Florence, Andy Zeng, Jonathan Tompson, Igor Mordatch, Yevgen Chebotar, et al. Inner monologue: Embodied reasoning through planning with language models.arXiv preprint arXiv:2207.05608, 2022

    Pith/arXiv arXiv 2022

  16. [16]

    Grounded decoding: Guiding text generation with grounded models for embodied agents.Advances in Neural Information Processing Systems, 36:59636–59661, 2023

    Wenlong Huang, Fei Xia, Dhruv Shah, Danny Driess, Andy Zeng, Yao Lu, Pete Florence, Igor Mordatch, Sergey Levine, Karol Hausman, et al. Grounded decoding: Guiding text generation with grounded models for embodied agents.Advances in Neural Information Processing Systems, 36:59636–59661, 2023

    2023

  17. [17]

    Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation

    Mukul Khanna, Yongsen Mao, Hanxiao Jiang, Sanjay Haresh, Brennan Shacklett, Dhruv Batra, Alexander Clegg, Eric Undersander, Angel X Chang, and Manolis Savva. Habitat synthetic scenes dataset (hssd-200): An analysis of 3d scene scale and realism tradeoffs for objectgoal navigation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

    2024

  18. [18]

    Embomatrix: A scalable training-ground for embodied decision-making.arXiv preprint arXiv:2510.12072, 2025

    Zixing Lei, Sheng Yin, Yichen Xiong, Yuanzhuo Ding, Wenhao Huang, Yuxi Wei, Qingyao Xu, Yiming Li, Weixin Li, Yunhong Wang, et al. Embomatrix: A scalable training-ground for embodied decision-making.arXiv preprint arXiv:2510.12072, 2025

    arXiv 2025

  19. [19]

    Behavior-1k: A benchmark for embodied ai with 1,000 everyday activities and realistic simulation

    Chengshu Li, Ruohan Zhang, Josiah Wong, Cem Gokmen, Sanjana Srivastava, Roberto Mart ´ın-Mart´ın, Chen Wang, Gabrael Levine, Michael Lingelbach, Jiankai Sun, et al. Behavior-1k: A benchmark for embodied ai with 1,000 everyday activities and realistic simulation. InConference on Robot Learning, pages 80–

  20. [20]

    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

    Pith/arXiv arXiv 2023

  21. [21]

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

    Yuchen Liu, Luigi Palmieri, Sebastian Koch, Ilche Georgievski, and Marco Aiello. Delta: Decomposed efficient long-term robot task planning using large language models. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 10995– 11001. IEEE, 2025

    2025

  22. [22]

    Sayplan: Grounding large language models using 3d scene graphs for scalable robot task planning

    Krishan Rana, Jesse Haviland, Sourav Garg, Jad Abou- Chakra, Ian Reid, and Niko Suenderhauf. Sayplan: Grounding large language models using 3d scene graphs for scalable robot task planning. In7th Annual Conference on Robot Learning

  23. [23]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    Pith/arXiv arXiv 2024

  24. [24]

    Vision-language interpreter for robot task planning

    Keisuke Shirai, Cristian C Beltran-Hernandez, Masashi Hamaya, Atsushi Hashimoto, Shohei Tanaka, Kento Kawaharazuka, Kazutoshi Tanaka, Yoshitaka Ushiku, and Shinsuke Mori. Vision-language interpreter for robot task planning. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 2051–2058. IEEE, 2024

    2051

  25. [25]

    Progprompt: Generating situated robot task plans using large language models

    Ishika Singh, Valts Blukis, Arsalan Mousavian, Ankit Goyal, Danfei Xu, Jonathan Tremblay, Dieter Fox, Jesse Thomason, and Animesh Garg. Progprompt: Generating situated robot task plans using large language models. arXiv preprint arXiv:2209.11302, 2022

    Pith/arXiv arXiv 2022

  26. [26]

    Llm- planner: Few-shot grounded planning for embodied agents with large language models

    Chan Hee Song, Jiaman Wu, Clayton Washington, Brian M Sadler, Wei-Lun Chao, and Yu Su. Llm- planner: Few-shot grounded planning for embodied agents with large language models. InProceedings of the IEEE/CVF international conference on computer vision, pages 2998–3009, 2023

    2023

  27. [27]

    Embodied task planning with large language models.arXiv preprint arXiv:2307.01848, 2023

    Zhenyu Wu, Ziwei Wang, Xiuwei Xu, Jiwen Lu, and Haibin Yan. Embodied task planning with large language models.arXiv preprint arXiv:2307.01848, 2023

    Pith/arXiv arXiv 2023

  28. [28]

    Zamir, Zhi-Yang He, Alexander Sax, Jitendra Malik, and Silvio Savarese

    Fei Xia, Amir R. Zamir, Zhi-Yang He, Alexander Sax, Jitendra Malik, and Silvio Savarese. Gibson env: real- world perception for embodied agents. InComputer Vision and Pattern Recognition (CVPR), 2018 IEEE Conference on. IEEE, 2018

    2018

  29. [29]

    Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    Pith/arXiv arXiv 2025

  30. [30]

    Unidomain: Pretraining a unified pddl domain from real-world demonstrations for generalizable robot task planning

    Haoming Ye, Yunxiao Xiao, Cewu Lu, and Panpan Cai. Unidomain: Pretraining a unified pddl domain from real-world demonstrations for generalizable robot task planning. 10

  31. [31]

    Generating symbolic world models via test- time scaling of large language models.arXiv preprint arXiv:2502.04728, 2025

    Zhouliang Yu, Yuhuan Yuan, Tim Z Xiao, Fuxiang Frank Xia, Jie Fu, Ge Zhang, Ge Lin, and Weiyang Liu. Generating symbolic world models via test- time scaling of large language models.arXiv preprint arXiv:2502.04728, 2025

    Pith/arXiv arXiv 2025

  32. [32]

    Pddlego: Iterative planning in textual environments.arXiv preprint arXiv:2405.19793, 2024

    Li Zhang, Peter Jansen, Tianyi Zhang, Peter Clark, Chris Callison-Burch, and Niket Tandon. Pddlego: Iterative planning in textual environments.arXiv preprint arXiv:2405.19793, 2024

    Pith/arXiv arXiv 2024

  33. [33]

    Embodied- reasoner: Synergizing visual search, reasoning, and action for embodied interactive tasks.arXiv preprint arXiv:2503.21696, 2025

    Wenqi Zhang, Mengna Wang, Gangao Liu, Xu Huixin, Yiwei Jiang, Yongliang Shen, Guiyang Hou, Zhe Zheng, Hang Zhang, Xin Li, et al. Embodied- reasoner: Synergizing visual search, reasoning, and action for embodied interactive tasks.arXiv preprint arXiv:2503.21696, 2025

    Pith/arXiv arXiv 2025

  34. [34]

    Isr-llm: Iterative self-refined large language model for long-horizon sequential task planning

    Zhehua Zhou, Jiayang Song, Kunpeng Yao, Zhan Shu, and Lei Ma. Isr-llm: Iterative self-refined large language model for long-horizon sequential task planning. In 2024 IEEE International Conference on Robotics and Automation (ICRA), pages 2081–2088. IEEE, 2024

    2024