Pith. sign in

REVIEW 4 major objections 4 minor 48 references

Grounded Vision-Language Interpreter for Long-Horizon Bimanual Task and Motion Planning

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Grounding vision-language output in symbolic-and-geometric planning with corrective failure feedback yields higher success than direct VLM planning on two-arm cooking tasks.

desk verdict Useful integration of VLM-to-PDDL generation with TAMP for bimanual cooking, but the main success-rate claim is not yet verifiable from the paper as written. read the letter →

arxiv 2506.03270 v3 pith:3Q4FEPXF submitted 2025-06-03 cs.RO cs.AI

classification cs.ROcs.AI
keywords bimanualmanipulationtaskandmotionplanningvision-languagemodelsPDDLproblemgenerationcorrectivelong-horizoncookingroboticsfailurefeedback
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

ViLaIn-TAMP tries to settle a practical question: can a robot turn a natural-language instruction and a camera image into a long-horizon, two-arm manipulation plan that is safe to execute? The paper's answer is to combine a vision-language model that writes structured symbolic planning problems with a planner that checks those problems for logical consistency and collision-free motions, and then closes the loop by feeding motion failures back to the model for revision. On five bimanual cooking tasks, the framework outperforms a direct vision-language-as-a-planner baseline by a reported 18 percentage points in mean success rate, with corrective planning adding a further 32-point boost. The reason this matters is that direct vision-language planners can hallucinate goals and produce plans that are physically infeasible; a symbolic check catches these before the robot moves.

What carries the argument

The load-bearing mechanism is a three-part cycle. First, a vision-language interpreter estimates objects, initial state, and goal from the instruction and image, writing them as a structured PDDL problem. Second, an integrated task-and-motion-planning module first finds a symbolic action sequence and then verifies and samples collision-free motions for it, treating learned contact-rich skills such as slicing as fixed start-to-end stages with enforced pose constraints. Third, a corrective-planning module takes the motion planner's detailed failure reports—collision partners, failed grasp, reachability—and re-prompts the interpreter with these constraints to produce a revised problem, repeating until success or a limit. The paper's key modification is making the motion planner emit structured, semantically meaningful failure feedback that a vision-language model can reason over, since raw planner scenes are not directly usable by such models.

What would settle it

Run the learned slicing skill in isolation over the same food types and poses used in the experiments and record how frequently it lands on the predefined ending pose; if that rate is low, the optimistic planning assumption would fail in practice and the real-robot success figures would not reflect typical deployments.

Watch

Extended reading notes

Core claim

The paper's central claim is that grounding a vision-language interpreter in an integrated task-and-motion-planning loop yields verifiable, interpretable, and more successful bimanual manipulation than generating action plans directly from a vision-language model. In the proposed design, the interpreter converts a linguistic instruction and a scene image into a complete problem specification in the Planning Domain Definition Language (PDDL), a standardized symbolic planning format; a high-level symbolic planner then searches for an action sequence, and a multi-stage motion planner verifies that the sequence can be realized as collision-free trajectories, incorporating learned skills such as slicing as constrained planning stages. When verification fails, a corrective-planning module converts the motion failure into natural-language feedback, re-prompts the interpreter to revise the specification, and replans. Evaluated over five cooking tasks, the full system reaches higher success rates than the direct-planning baseline, especially as task complexity grows, and the corrective loop is shown to be a major source of the gain. The approach is also demonstrated on a real dual-arm robot.

Load-bearing premise

The plan's real-world success depends on the learned slicing skill reliably returning the robot to the predefined ending pose after every execution, and the paper does not measure how often that actually happens.

Editorial extensions

If this is right

  • A robot can accept a new instruction and scene image and produce a plan that is checked for logical and physical feasibility before any motion is executed.
  • Two-arm tasks with inter-arm collision constraints no longer require a hand-tuned planner per task; the same domain and loop handles pick-and-place, obstacle avoidance, slicing, and serving.
  • When motion planning fails, the system does not simply stop: it converts the failure into a revised symbolic problem and attempts a new plan, which is what produces most of the reported success-rate gain.
  • Contact-rich skills can be plugged into the same framework as staged skills with enforced start and end poses, so adding new skills extends the planner without rewriting the symbolic domain.
  • Non-expert users could direct a dual-arm cooking robot by language while keeping the plan inspectable in a structured symbolic form before execution.

Reading between the lines

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

  • Inference: the size of the corrective-planning gain suggests that a large share of VLM planning errors are motion-level geometric mistakes rather than pure language-understanding errors; if perception were perfect, the gap between direct planning and the hybrid loop might shrink but would not vanish, because the loop also corrects contradictory goal specifications.
  • Inference: the same failure-feedback loop could be ported to other contact-rich skills by training each skill with an explicit terminating pose and wrapping it as a constrained stage, extending the framework beyond slicing to pouring, kneading, or assembly.
  • Inference: a simple controlled experiment—feeding the same five problems to the framework with and without the corrective loop while holding the foundation model and perception fixed—would isolate how much of the gain comes from feedback quality versus model revision ability.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes ViLaIn-TAMP, a hybrid planning framework that uses a vision-language interpreter to convert natural-language instructions and scene images into PDDL problem specifications, then grounds these specifications in an integrated Task and Motion Planning system built on Fast Downward and MoveIt Task Constructor for dual-arm cooking tasks. A corrective-planning module feeds structured motion-failure feedback back to the VLM to revise the PDDL problem and replan. The framework is evaluated on five simulated bimanual manipulation tasks and validated on a physical dual-arm UR5e system, with the paper reporting that ViLaIn-TAMP outperforms a direct VLM-as-a-planner baseline and that corrective planning substantially improves success rate.

Significance. The architectural direction is valuable: grounding a VLM in a symbolic planner and a verified motion-planning layer, with iterative correction from structured failure feedback, is a credible way to make VLM-generated plans verifiable and executable for bimanual tasks. The paper also contributes an open dataset, a modified MTC that emits structured failure messages, and a real-robot demonstration, and it evaluates the framework with both proprietary and open-source foundation models. However, the central quantitative claim is not currently supported as reported: the headline success-rate margin changes between the abstract, introduction, and results section, trial counts and confidence intervals are absent for the main comparison, and one load-bearing assumption about the learned slicing skill is unmeasured. These issues are fixable, but they must be addressed before the claimed effect size can be accepted.

major comments (4)
  1. [Abstract, §I, §V-B] The headline success-rate advantage is reported inconsistently: the abstract states 17.5% (with a 32.9% CP boost), the introduction states 18% (with a 32% boost), and §V-B states an average margin of 10% over the baseline. No confidence intervals, significance tests, or per-task trial counts accompany Fig. 5 or Table I for the main GPT-4o evaluation; only the open-source-model ablation in §V-C specifies 50 trials per task. A 7.5-point discrepancy in the central effect size, with no measure of sampling variability, means the claimed superiority over the baseline is not verifiable from the manuscript. Please report exact trial counts, repeated-seed information, confidence intervals, and reconcile the three stated margins.
  2. [§V-A, Baseline-CP] The baseline is described as taking O, L, D as input and directly generating action sequences, while ViLaIn-TAMP takes the raw scene observation S (image) in addition to domain knowledge D and the linguistic instruction L. If O contains only detected object labels, as suggested by the Object Estimator description in §IV-A, then the comparison gives the proposed method access to strictly more perceptual information than the baseline, and part of the reported margin could be attributable to that input asymmetry rather than to grounding in TAMP. Please specify exactly what O contains and either give the baseline the same visual input as ViLaIn-TAMP or explicitly control for the input modality.
  3. [§IV-B, §V-D] The integrated TAMP system treats the learned slicing skill as a mock MTC stage and states: "During execution, we assume the skill reliably returns the robot to the predefined ending pose." The real-robot validation in §V-D reports no quantitative measure of skill-level reliability or success, and the text says execution halts if the ending pose is not reached. Because slicing is one of the core long-horizon behaviors evaluated in the paper, the end-to-end success claim is conditional on an unmeasured component; please report skill-level success rates and failure counts from the real-robot experiments.
  4. [§V-C, Table I] The numbers for the no-CP condition are difficult to reconcile. Section V-C reports a 45% average success rate for ViLaIn-TAMP-No-CP, but Table I's without-ICL column has per-task success rates of 92, 84, 40, 90, and 80, which average 77.2%. If the 45% refers to a different configuration, such as the baseline or an ablation with different inputs, that must be stated explicitly; as written, a reader cannot tell which condition Table I describes or how the CP boost of 32.9%/32% was computed.
minor comments (4)
  1. [Algorithm 1, §V-A] The notation for the maximum number of corrective-planning attempts alternates among N_CP,max, N CP,max, and N CP, max; please use a single consistent notation throughout.
  2. [Fig. 5, Table I] Figure 5 and Table I do not clearly indicate which bars or rows correspond to the with-ICL and without-ICL conditions; the captions should state the color/pattern legend and the exact configuration family (ViLaIn-TAMP-CP vs. ViLaIn-TAMP-No-CP).
  3. [Algorithm 1, §V-B] Algorithm 1 calls EXECUTE-REAL-ROBOT after every successful plan, but the primary results in §V-B are reported as a mixture of planning and execution success in simulation; please clarify which evaluation results involve real-robot execution and which are simulated.
  4. [§V-D] In the Adaptive Execution paragraph, the list items are labeled (A) and (B), but the sentence "grouping only (A) actions for potential replanning" is immediately followed by a description of processing (B) actions; rephrasing would make the grouping and replanning logic clearer.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the central performance claim is an empirically tested comparison against an external baseline, not a consequence of the paper's definitions or self-citations.

full rationale

The paper's central claim is that ViLaIn-TAMP outperforms a direct VLM-as-a-planner baseline and that corrective planning further improves success. This is supported by success-rate measurements over 10/9 problems per task; the margin is not entailed by the construction. ViLaIn and the RL slicing skill are drawn from prior work by overlapping authors ([15], [43]), but they enter as components whose contribution is re-evaluated in the integrated system against an independent baseline, so those self-citations are not load-bearing for the headline result. No fitted parameter is renamed as a prediction, and no equation makes the output equal to the input by definition. The CP effect is measured by ablating the module, not assumed. The inconsistent reported margins (17.5%, 18%, 10%) and missing trial counts/confidence intervals are empirical reproducibility concerns, not circularity. The assumption that the slicing skill reliably reaches its ending pose is an unmeasured external-skill assumption, not a self-referential derivation. The derivation chain is therefore self-contained in the sense relevant to circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the correctness of the hand-crafted PDDL domain, black-box planner and tool implementations, the reliability of the learned slicing skill, and the fidelity of the simulator. No numerical parameters are fitted to data in a statistical sense, but the reported success rates depend on hand-chosen settings (N_CP,max and ICL count).

free parameters (2)
  • Maximum corrective planning attempts (N_CP,max) = 3
    Chosen by hand; ablation shows success rate depends on it (0, 1, 3).
  • Number of in-context examples (ICL) = 1
    The paper compares 0 vs 1 ICL examples; mean success rate increases by 5.2%, so this hand-chosen value affects results.
assumptions (5)
  • domain assumption The hand-crafted PDDL cooking domain correctly captures all necessary predicates and actions for the five tasks.
    Section III and V-A define the domain; if it is incomplete, the planner cannot represent valid plans.
  • standard math Fast Downward is a correct symbolic planner for the generated PDDL problems.
    Section IV-B uses it as a black box; its correctness is assumed.
  • domain assumption MTC stage planning either finds a collision-free trajectory if one exists or returns accurate failure comments.
    Section IV-B and IV-C rely on MTC feedback to drive corrective planning; inaccurate feedback would mislead revision.
  • domain assumption The RL slicing skill reliably returns the robot to the predefined ending pose.
    Section IV-B explicitly states this assumption; it is load-bearing for plan execution, especially on the real robot.
  • domain assumption The simulation environment used for quantitative evaluation faithfully models the robot and physics.
    Section V-A reports simulated success rates; transfer to real hardware is not quantitatively validated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Grounded Vision-Language Interpreter for Long-Horizon Bimanual Task and Motion Planning." pith.science (2026). https://pith.science/paper/3Q4FEPXF

@misc{pith2026250603270,
  author       = {Pith},
  title        = {Pith review of: Grounded Vision-Language Interpreter for Long-Horizon Bimanual Task and Motion Planning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3Q4FEPXF}},
  note         = {Machine review of arXiv:2506.03270}
}
read the original abstract

While recent advances in vision-language models have accelerated language-guided robot planning, their black-box nature lacks the safety guarantees and interpretability crucial for real-world deployment. Conversely, classical symbolic planners offer rigorous safety verification but require significant expert knowledge for setup. Moreover, most existing methods are limited to single-arm pick-and-place tasks, leaving bimanual manipulation largely underexplored, despite its tightly interdependent subtasks and the need to explicitly manage inter-arm collisions. To bridge this gap, this paper proposes ViLaIn-TAMP, a hybrid planning framework for enabling verifiable, interpretable, and autonomous bimanual robot behaviors. ViLaIn-TAMP comprises three main components: (1) a Vision-Language Interpreter (ViLaIn) adapted from a prior work that converts multimodal inputs into structured PDDL problem specifications, (2) an integrated Task and Motion Planning (TAMP) system that grounds these specifications in actionable trajectory sequences through symbolic and geometric constraint reasoning, explicitly verifying feasibility before execution, and (3) a corrective planning (CP) module which receives structured motion failure feedback and feeds it as constraints back to ViLaIn to refine the specification. We design challenging bimanual manipulation tasks in a cooking domain to evaluate our framework, where experimental results show that ViLaIn-TAMP outperforms a VLM-as-a-planner baseline by 17.5% in mean success rate, with the CP module boosting it further by 32.9%. We further validate ViLaIn-TAMP on a physical dual-arm robotic system. Project page: https://omron-sinicx.github.io/ViLaIn-TAMP

Figures

Figures reproduced from arXiv: 2506.03270 by the authors.

Figure 1
Figure 1. We develop ViLaIn-TAMP, a novel end-to-end planning framework for long-horizon manipulation that 1) converts multimodal inputs into PDDL problems, 2) finds feasible motion plans via an integrated TAMP system, and 3) reasons over detailed failure feedback to revise and replan using corrective planning. ViLaIn-TAMP is capable of solving real￾world, long-horizon bimanual cooking tasks. LLM/VLM-as-a-planner approach, wh… view at source ↗
Figure 2
Figure 2. Overview of ViLaIn-TAMP framework. In the ViLaIn part (A), given a linguistic instruction in natural language and an image as a scene observation, the ViLaIn module generates a complete PDDL problem. In the TAMP part (B), the generated PDDL problem is passed to the integrated TAMP module, which solves the problem for a sequence of symbolic actions and collision-free motion trajectories. If successful, the complete p… view at source ↗
Figure 3
Figure 3. Overview of Corrective Planning Module. ViLaIn-TAMP implements a 3-step corrective planning (CP) approach, which involves 1) re-prompting the model with the failure feedback to 2) revise the PDDL problem, and then 3) replanning using the revised PDDL problem. context of our slicing task, certain grasp poses for fixturing an object by an arm may interfere with slicing and cause collisions with the other arm. Conventi… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Example of Motion Planning Failure Feedback and Visualiza￾tion in RViz. In MTC, both successful and failed motion plans are published and can be visualized in RViz before actual execution, allowing human introspection. Our custom MTC implementation extracts these failu…
Figure 5
Figure 5. Figure 5: Comparison of ViLaIn-TAMP and the baseline, evaluating their performance on five cooking tasks with and without corrective planning (CP). The maximum number of CP attempts is set to 3. ViLaIn-TAMP consistently outperforms the baseline in all tasks. CP is effective in b…
Figure 6
Figure 6. Figure 6: Failure mode count per method. 10%, with even more significant improvements observed as task complexity increases. In Tabl. I, we report planning￾related metrics for each of the tasks, including average task planning attempts, average motion planning attempts, and plan…
Figure 8
Figure 8. Figure 8: Evaluation of Open-Source Models. The results of GPT-4o-CP-ICL are plotted for reference. 4 [5] (14B parameters) and Alibaba’s Qwen2.5-Coder-32B￾Instruct [46] (32B parameters). Note that these models are text-only and that Qwen2.5-VL-7B-Instruct is used as the object e…
Figure 9
Figure 9. Figure 9: Overview of our robotic system. Dual arm robotic system for real-world validation on a cooking domain. actions rely on Reinforcement Learning (RL) with compli￾ance control. Prior work [43] showed this approach adapts to unseen objects and can capture an object’s force …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 47 canonical work pages

  1. [1]

    Automated Planning for Robotics,

    E. Karpas and D. Magazzeni, “Automated Planning for Robotics,” Annual Review of Control, Robotics, and Autonomous Systems, vol. 3, no. 1, pp. 417–439, May 2020

  2. [2]

    Integrated task and motion planning,

    C. R. Garrettet al., “Integrated task and motion planning,”Annual review of control, robotics, and autonomous systems, vol. 4, no. 1, pp. 265–293, 2021

  3. [3]

    GPT-4 technical report,

    OpenAI, “GPT-4 technical report,” 2024

  4. [4]

    The Llama 3 herd of models,

    A. Grattafioriet al., “The Llama 3 herd of models,” 2024

  5. [5]

    Phi-4 technical report,

    M. Abdinet al., “Phi-4 technical report,” 2024

  6. [6]

    GPT-4o system card,

    OpenAI, “GPT-4o system card,” 2024

  7. [7]

    Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context,

    G. Team, “Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context,” 2024

  8. [8]

    Visual instruction tuning,

    H. Liuet al., “Visual instruction tuning,” inNeurIPS, vol. 36, 2023, pp. 34 892–34 916

Show all 48 references
  1. [9]

    Inner monologue: Embodied reasoning through planning with language models,

    W. Huanget al., “Inner monologue: Embodied reasoning through planning with language models,” inCoRL, vol. 205, 2023, pp. 1769– 1782

  2. [10]

    ProgPrompt: Generating situated robot task plans using large language models,

    I. Singhet al., “ProgPrompt: Generating situated robot task plans using large language models,” inICRA, 2023, pp. 11 523–11 530

  3. [11]

    Generalized planning in PDDL domains with pre- trained large language models,

    T. Silveret al., “Generalized planning in PDDL domains with pre- trained large language models,” inAAAI, vol. 38, 2024, pp. 20 256– 20 264

  4. [12]

    ReplanVLM: Replanning robotic tasks with visual language models,

    A. Meiet al., “ReplanVLM: Replanning robotic tasks with visual language models,”RA-L, vol. 9, no. 11, pp. 10 201–10 208, 2024

  5. [13]

    LLM3: Large language model-based task and motion planning with motion failure reasoning,

    S. Wanget al., “LLM3: Large language model-based task and motion planning with motion failure reasoning,” inIROS, 2024, pp. 12 086– 12 092

  6. [14]

    Guiding long-horizon task and motion planning with vision language models,

    Z. Yanget al., “Guiding long-horizon task and motion planning with vision language models,” inICRA, 2025, pp. 16 847–16 853

  7. [15]

    Vision-language interpreter for robot task planning,

    K. Shiraiet al., “Vision-language interpreter for robot task planning,” inICRA, 2024, pp. 2051–2058

  8. [16]

    PDDL2.1: An extension to PDDL for express- ing temporal planning domains,

    M. Fox and D. Long, “PDDL2.1: An extension to PDDL for express- ing temporal planning domains,”JAIR, vol. 20, pp. 61–124, 2003

  9. [17]

    Reducing the barrier to entry of complex robotic software: a MoveIt! case study,

    D. Colemanet al., “Reducing the barrier to entry of complex robotic software: a MoveIt! case study,”ArXiV, 2014

  10. [18]

    Extending the motion planning frame- work—MoveIt with advanced manipulation functions for industrial ap- plications,

    P. M. Fresnilloet al., “Extending the motion planning frame- work—MoveIt with advanced manipulation functions for industrial ap- plications,”Robotics and Computer-Integrated Manufacturing, vol. 83, p. 102559, 2023

  11. [19]

    MoveIt! Task Constructor for Task-Level Motion Planning,

    M. Görneret al., “MoveIt! Task Constructor for Task-Level Motion Planning,” inICRA, 2019, pp. 190–196

  12. [20]

    The fast downward planning system,

    M. Helmert, “The fast downward planning system,”JAIR, vol. 26, pp. 191–246, 2006

  13. [21]

    DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning,

    DeepSeek-AI, “DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning,” 2025

  14. [22]

    Qwen2.5-VL technical report,

    S. Baiet al., “Qwen2.5-VL technical report,”ArXiV, 2025

  15. [23]

    Bc-z: Zero-shot task generalization with robotic imitation learning,

    E. Janget al., “Bc-z: Zero-shot task generalization with robotic imitation learning,” inCoRL, vol. 164, 2022, pp. 991–1002

  16. [24]

    Code as policies: Language model programs for embodied control,

    J. Lianget al., “Code as policies: Language model programs for embodied control,” inICRA, 2023, pp. 9493–9500

  17. [25]

    Open X-Embodiment: Robotic learning datasets and RT-X models,

    A. O’Neillet al., “Open X-Embodiment: Robotic learning datasets and RT-X models,” inICRA, 2024, pp. 6892–6903

  18. [26]

    π 0: A vision-language-action flow model for general robot control,

    K. Blacket al., “π 0: A vision-language-action flow model for general robot control,” 2024

  19. [27]

    Real-world robot applications of foundation models: a review,

    K. Kawaharazukaet al., “Real-world robot applications of foundation models: a review,”Advanced Robotics, vol. 38, no. 18, pp. 1232–1254, 2024

  20. [28]

    Gemini robotics: Bringing ai into the physical world,

    G. R. Team, “Gemini robotics: Bringing ai into the physical world,” 2025

  21. [29]

    Planning with large language models via corrective re-prompting,

    S. S. Ramanet al., “Planning with large language models via corrective re-prompting,” inNeurIPS 2022 F oundation Models for Decision Making Workshop, 2022

  22. [30]

    Text2Motion: From natural language instructions to feasible plans,

    K. Linet al., “Text2Motion: From natural language instructions to feasible plans,”Autonomous Robots, vol. 47, no. 8, pp. 1345–1365, 2023

  23. [31]

    LLM+P: Empowering large language models with optimal planning proficiency,

    B. Liuet al., “LLM+P: Empowering large language models with optimal planning proficiency,”ArXiV, 2023

  24. [32]

    Translating natural language to planning goals with large-language models,

    Y . Xieet al., “Translating natural language to planning goals with large-language models,”ArXiv, 2023

  25. [33]

    PDDLStream: Integrating symbolic planners and blackbox samplers via optimistic adaptive planning,

    C. R. Garrettet al., “PDDLStream: Integrating symbolic planners and blackbox samplers via optimistic adaptive planning,” inICAPS, vol. 30, 2020, pp. 440–448

  26. [34]

    Practical task and motion planning for robotic food preparation,

    J. Siburianet al., “Practical task and motion planning for robotic food preparation,” inSII. IEEE, 2025, pp. 1229–1234

  27. [35]

    Long-horizon manipulation of unknown objects via task and motion planning with estimated affordances,

    A. Curtiset al., “Long-horizon manipulation of unknown objects via task and motion planning with estimated affordances,” inICRA. IEEE, 2022, pp. 1940–1946

  28. [36]

    Combined task and motion planning through an extensible planner-independent interface layer,

    S. Srivastavaet al., “Combined task and motion planning through an extensible planner-independent interface layer,” inICRA. IEEE, 2014, pp. 639–646

  29. [37]

    Task and motion planning with large language models for object rearrangement,

    Y . Dinget al., “Task and motion planning with large language models for object rearrangement,” inIROS. IEEE, 2023, pp. 2086–2092

  30. [38]

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

    K. Ranaet al., “Sayplan: Grounding large language models using 3d scene graphs for scalable robot task planning,” inCoRL, vol. 229, 2023, pp. 23–72

  31. [39]

    AutoTAMP: Autoregressive task and motion planning with LLMs as translators and checkers,

    Y . Chenet al., “AutoTAMP: Autoregressive task and motion planning with LLMs as translators and checkers,” inICRA, 2024, pp. 6695– 6702

  32. [40]

    Castl: Constraints as specifications through llm translation for long-horizon task and motion planning,

    W. Guoet al., “Castl: Constraints as specifications through llm translation for long-horizon task and motion planning,” inICRA. IEEE, 2025, pp. 11 957–11 964

  33. [41]

    Cape: Corrective actions from precondition errors using large language models,

    S. S. Ramanet al., “Cape: Corrective actions from precondition errors using large language models,” inICRA. IEEE, 2024, pp. 14 070– 14 077

  34. [42]

    AHA: A vision-language-model for detecting and reasoning over failures in robotic manipulation,

    J. Duanet al., “AHA: A vision-language-model for detecting and reasoning over failures in robotic manipulation,” inICLR, 2025

  35. [43]

    SliceIt! - a dual simulator framework for learning robot food slicing,

    C. C. Beltran-Hernandezet al., “SliceIt! - a dual simulator framework for learning robot food slicing,” inICRA. IEEE, 2024, pp. 4296–4302

  36. [44]

    Generating consistent pddl domains with large language models,

    P. Smirnovet al., “Generating consistent pddl domains with large language models,”ArXiV, 2024

  37. [45]

    Large language models as planning domain gener- ators,

    J. Oswaldet al., “Large language models as planning domain gener- ators,” inICAPS, vol. 34, 2024, pp. 423–431

  38. [46]

    Qwen2.5-coder technical report,

    B. Huiet al., “Qwen2.5-coder technical report,” 2024

  39. [47]

    SAM 2: Segment anything in images and videos,

    N. Raviet al., “SAM 2: Segment anything in images and videos,” ArXiV, 2024

  40. [48]

    InterPreT: Interactive predicate learning from language feedback for generalizable task planning,

    M. Hanet al., “InterPreT: Interactive predicate learning from language feedback for generalizable task planning,” inRSS, 2024

Pith tools

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