pith. sign in

arxiv: 2606.07107 · v1 · pith:Y5NEIOEVnew · submitted 2026-06-05 · 💻 cs.RO

Coarse-to-Control: Action-Token Planning for Vision-Language-Action Models

Jinhao Wu , Shiduo Zhang , Yicheng Liu , Xiaopeng Yu , Sixian Li , Siyin Wang , Hang Zhao , Jing Huo
show 4 more authors
Yang Gao Jingjing Gong Xipeng Qiu Yu-Gang Jiang
This is my paper

Pith reviewed 2026-06-27 21:55 UTC · model grok-4.3

classification 💻 cs.RO
keywords vision-language-actionaction-token planningcoarse-to-controllong-horizon manipulationdiscrete action vocabularyplan-execute VLArobot control
0
0 comments X

The pith

VLA models that first predict a compact sequence of coarse action tokens in a shared discrete vocabulary outperform direct action generation on long-horizon tasks.

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

Vision-language-action models usually map observations straight to motor commands, but early mistakes compound across many steps in extended tasks. Coarse-to-Control adds an explicit planning stage inside the action-token space: the model first outputs a short sequence of coarse tokens that outline the future trajectory, then produces the executable actions conditioned on that outline. Because planning and execution draw from the same discrete vocabulary, the intermediate plan requires no separate translation and remains close to actual control signals. Experiments across simulation benchmarks and real robot setups show consistent gains, with the biggest improvements appearing on multi-stage problems. The approach therefore demonstrates that native action-space planning can stabilize long sequences without external planners or abstract representations.

Core claim

The paper shows that a plan-execute VLA architecture succeeds by letting the policy predict a compact sequence of coarse action tokens first, then generate executable tokens conditioned on this plan, where both stages share one unified discrete action vocabulary so that the plan supplies directly usable guidance rather than an abstract hint.

What carries the argument

The unified discrete action vocabulary that lets coarse planning tokens and fine execution tokens occupy the same space and thereby keep the plan on the control manifold.

If this is right

  • Gains appear consistently across LIBERO, SimplerEnv-WidowX, and real manipulation tasks, with the largest margins on long-horizon multi-stage problems.
  • Planning occurs inside the existing action vocabulary, so no separate language planner or translation module is required.
  • The shared vocabulary ensures the plan remains executable without drifting into abstract representations.
  • The same architecture can be applied to existing VLA backbones by extending their output sequences to include planning tokens before execution tokens.

Where Pith is reading between the lines

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

  • The method could be tested on tasks with changing goals by regenerating the coarse plan periodically rather than once at the start.
  • Optimal length and granularity of the coarse sequence remain open parameters that future work could tune per task class.
  • Exposing the coarse plan tokens during execution might allow human operators to intervene at the trajectory level instead of the low-level command level.
  • The same coarse-to-fine structure might stabilize other long-sequence prediction problems that currently suffer from compounding token errors.

Load-bearing premise

A compact sequence of coarse action tokens predicted in the shared discrete vocabulary will supply directly actionable guidance that stays close to the control manifold without any additional translation.

What would settle it

A controlled experiment on LIBERO or real-world tasks in which inserting the coarse planning stage produces equal or lower success rates than direct action generation would falsify the claimed benefit.

Figures

Figures reproduced from arXiv: 2606.07107 by Hang Zhao, Jing Huo, Jingjing Gong, Jinhao Wu, Shiduo Zhang, Sixian Li, Siyin Wang, Xiaopeng Yu, Xipeng Qiu, Yang Gao, Yicheng Liu, Yu-Gang Jiang.

Figure 1
Figure 1. Figure 1: Comparison of reasoning paradigms for VLA control: No CoT, textual CoT, visual CoT, [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Plan-execute VLA framework. The policy conditions on observations, language, and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Real-world task success rates (%). We report final task success for each task and the [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative analysis of joint-mode action-token planning. To qualitatively inspect this mechanism, we use two diagnostics: attention over image tokens and a decoded coarse trajectory from the predicted planning tokens. Figure 5a compares the first-frame attention map produced with planning tokens against the no-plan baseline. With planning, attention concentrates more strongly on task-relevant regions arou… view at source ↗
Figure 6
Figure 6. Figure 6: Average success on the multi-stage real-world tasks [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Real-world setup and representative keyframes from the four physical manipulation tasks [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Joint plan-execute action tokenizer. Execution mode represents short-horizon executable [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
read the original abstract

Most vision-language-action (VLA) models map observations directly to actions without explicit intermediate planning, which limits performance on long-horizon tasks where early mistakes compound. We propose Coarse-to-Control, a plan-execute VLA that introduces planning natively in the action-token space. The key idea is to let the policy first predict a compact sequence of coarse action tokens that summarize the intended future trajectory, and then generate executable action tokens conditioned on this plan. Because both planning and execution share a unified discrete action vocabulary, the plan stays close to the control manifold and provides directly actionable guidance rather than an abstract hint that must be translated back to motor commands. Experiments on LIBERO, SimplerEnv-WidowX, and real-world manipulation tasks show that action-token planning consistently improves over direct action generation, with the largest gains on long-horizon multi-stage tasks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript proposes Coarse-to-Control, a plan-execute vision-language-action (VLA) model. It first predicts a compact sequence of coarse action tokens summarizing the intended future trajectory in a shared discrete action vocabulary, then generates executable action tokens conditioned on this plan. The shared vocabulary is intended to keep the plan close to the control manifold and directly actionable. Experiments on LIBERO, SimplerEnv-WidowX, and real-world manipulation tasks are claimed to show consistent improvements over direct action generation, with largest gains on long-horizon multi-stage tasks.

Significance. If the experimental results hold with proper controls, the approach could be significant for VLA research by integrating planning natively in the action-token space rather than relying on abstract language plans that require translation. The unified vocabulary is a clean design choice that avoids additional modules. However, the absence of any reported metrics, baselines, ablations, or statistical details in the available text prevents assessment of whether the claimed gains are real, robust, or merely incremental.

major comments (1)
  1. [Abstract] Abstract: The central claim that 'action-token planning consistently improves over direct action generation, with the largest gains on long-horizon multi-stage tasks' is asserted without any baselines, metrics (e.g., success rate, horizon length), statistical details, or ablation results. This renders the experimental contribution impossible to evaluate and is load-bearing for the paper's main thesis.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and for identifying the lack of quantitative support in the abstract. We address the point directly below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that 'action-token planning consistently improves over direct action generation, with the largest gains on long-horizon multi-stage tasks' is asserted without any baselines, metrics (e.g., success rate, horizon length), statistical details, or ablation results. This renders the experimental contribution impossible to evaluate and is load-bearing for the paper's main thesis.

    Authors: We agree that the abstract presents the central claim without accompanying numbers, baselines, or statistical details, which makes it difficult for a reader to assess the strength of the result from the abstract alone. The full manuscript contains the supporting experiments (success rates, direct-action baselines, horizon-stratified results, and ablations) in the results section. To resolve the referee's concern we will revise the abstract to include a concise statement of the key quantitative findings (e.g., average success-rate lift and the differential gain on long-horizon tasks) drawn from those experiments. This change keeps the abstract self-contained while preserving the original claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper proposes an empirical method (Coarse-to-Control) for VLA models that predicts coarse action tokens before executable ones, sharing a discrete vocabulary. The abstract and description contain no equations, derivations, fitted parameters presented as predictions, or self-citation chains. The central claim rests on experimental results across LIBERO, SimplerEnv-WidowX, and real-world tasks rather than any self-referential reduction. No load-bearing step reduces by construction to its inputs, satisfying the criteria for a self-contained empirical contribution with score 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities described in the abstract.

pith-pipeline@v0.9.1-grok · 5714 in / 893 out tokens · 21830 ms · 2026-06-27T21:55:56.490575+00:00 · methodology

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

41 extracted references · 29 canonical work pages · 22 internal anchors

  1. [1]

    RT-1: Robotics Transformer for Real-World Control at Scale

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, et al. Rt-1: Robotics transformer for real-world control at scale.arXiv preprint arXiv:2212.06817, 2022. doi:10.48550/arXiv.2212.06817. URLhttps://arxiv.org/abs/2212.06817

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2212.06817 2022

  2. [2]

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, X. Chen, et al. Rt-2: Vision-language-action models transfer web knowledge to robotic control.arXiv preprint arXiv:2307.15818, 2023. doi:10.48550/arXiv.2307.15818. URLhttps://arxiv.org/abs/2307.15818

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2307.15818 2023

  3. [3]

    Octo: An Open-Source Generalist Robot Policy

    Octo Model Team, D. Ghosh, H. Walke, K. Pertsch, K. Black, et al. Octo: An open-source generalist robot policy.arXiv preprint arXiv:2405.12213, 2024. doi:10.48550/arXiv.2405. 12213. URLhttps://arxiv.org/abs/2405.12213

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2405 2024

  4. [5]

    Black, N

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Haus- man, B. Ichter, S. Jakubczak, T. Jones, L. Ke, S. Levine, A. Li-Bell, M. Mothukuri, S. Nair, K. Pertsch, L. X. Shi, J. Tanner, Q. Vuong, A. Walling, H. Wang, and U. Zhilinsky.π 0: A vision-language-action flow model for general robot control.arXiv preprint arXiv:2410.24164,

    Pith/arXiv arXiv

  5. [6]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    doi:10.48550/arXiv.2410.24164. URLhttps://arxiv.org/abs/2410.24164

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2410.24164

  6. [7]

    D. A. Rosenbaum.Human motor control. Academic press, 2009

    2009

  7. [8]

    N. A. Bernstein.The Co-ordination and Regulation of Movements. Pergamon Press, Oxford, 1967

    1967

  8. [9]

    K. S. Lashley. The problem of serial order in behavior. In L. A. Jeffress, editor,Cerebral Mechanisms in Behavior, pages 112–136. Wiley, New York, 1951

    1951

  9. [10]

    Zawalski, W

    M. Zawalski, W. Chen, K. Pertsch, O. Mees, C. Finn, and S. Levine. Robotic control via embodied chain-of-thought reasoning.arXiv preprint arXiv:2407.08693, 2024. doi:10.48550/ arXiv.2407.08693. URLhttps://arxiv.org/abs/2407.08693

    Pith/arXiv arXiv 2024

  10. [11]

    ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planning

    C.-P. Huang, Y .-H. Wu, M.-H. Chen, Y .-C. F. Wang, and F.-E. Yang. Thinkact: Vision-language-action reasoning via reinforced visual latent planning.arXiv preprint arXiv:2507.16815, 2025. doi:10.48550/arXiv.2507.16815. URLhttps://arxiv.org/abs/ 2507.16815

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2507.16815 2025

  11. [12]

    Q. Zhao, Y . Lu, M. J. Kim, Z. Fu, Z. Zhang, Y . Wu, Z. Li, Q. Ma, S. Han, C. Finn, A. Handa, M.-Y . Liu, D. Xiang, G. Wetzstein, and T.-Y . Lin. Cot-vla: Visual chain-of-thought reasoning for vision-language-action models.arXiv preprint arXiv:2503.22020, 2025. doi:10.48550/ arXiv.2503.22020. URLhttps://arxiv.org/abs/2503.22020

    Pith/arXiv arXiv 2025

  12. [13]

    DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge

    W. Zhang, H. Liu, Z. Qi, Y . Wang, X. Yu, J. Zhang, R. Dong, J. He, F. Lu, H. Wang, Z. Zhang, L. Yi, W. Zeng, and X. Jin. Dreamvla: A vision-language-action model dreamed with compre- hensive world knowledge.arXiv preprint arXiv:2507.04447, 2025. doi:10.48550/arXiv.2507. 04447. URLhttps://arxiv.org/abs/2507.04447

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2507 2025

  13. [14]

    D. Qu, H. Song, Q. Chen, Y . Yao, X. Ye, Y . Ding, Z. Wang, J. Gu, B. Zhao, D. Wang, and X. Li. Spatialvla: Exploring spatial representations for visual-language-action model.arXiv preprint arXiv:2501.15830, 2025. doi:10.48550/arXiv.2501.15830. URLhttps://arxiv. org/abs/2501.15830

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2501.15830 2025

  14. [16]

    Huang, M

    H. Huang, M. Cen, K. Tan, X. Quan, G. Huang, and H. Zhang. Graphcot-vla: A 3d spatial- aware reasoning vision-language-action model for robotic manipulation with ambiguous in- structions.arXiv preprint arXiv:2508.07650, 2025. doi:10.48550/arXiv.2508.07650. URL https://arxiv.org/abs/2508.07650

    work page doi:10.48550/arxiv.2508.07650 2025

  15. [17]

    FAST: Efficient Action Tokenization for Vision-Language-Action Models

    K. Pertsch, K. Stachowicz, B. Ichter, D. Driess, S. Nair, Q. Vuong, O. Mees, C. Finn, and S. Levine. Fast: Efficient action tokenization for vision-language-action models.arXiv preprint arXiv:2501.09747, 2025. doi:10.48550/arXiv.2501.09747. URLhttps://arxiv.org/abs/ 2501.09747

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2501.09747 2025

  16. [18]

    M. J. Kim, C. Finn, and P. Liang. Fine-tuning vision-language-action models: Optimizing speed and success.arXiv preprint arXiv:2502.19645, 2025. doi:10.48550/arXiv.2502.19645. URLhttps://arxiv.org/abs/2502.19645

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2502.19645 2025

  17. [19]

    $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization

    K. Black, N. Brown, J. Darpinian, K. Dhabalia, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fu- sai, M. Y . Galliker, D. Ghosh, L. Groom, K. Hausman, B. Ichter, S. Jakubczak, T. Jones, L. Ke, D. LeBlanc, S. Levine, A. Li-Bell, M. Mothukuri, et al.π 0.5: a vision-language- action model with open-world generalization.arXiv preprint arXiv:2504.16054, 2025. doi...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2504.16054 2025

  18. [20]

    Schumacher, H

    Z. Zhong, J. Li, J. He, H. Yan, X. Gong, G. Zhao, Y . Cai, J. Gao, X. Yan, B. Liu, Y . Chen, L. Yang, and H. Li. Dualcot-vla: Visual-linguistic chain of thought via parallel reasoning for vision-language-action models.arXiv preprint arXiv:2603.22280, 2026. doi:10.48550/arXiv. 2603.22280. URLhttps://arxiv.org/abs/2603.22280

    work page internal anchor Pith review doi:10.48550/arxiv 2026

  19. [21]

    J. Lee, J. Duan, H. Fang, Y . Deng, S. Liu, B. Li, B. Fang, J. Zhang, Y . R. Wang, S. Lee, W. Han, W. Pumacay, A. Wu, R. Hendrix, K. Farley, E. VanderBilt, A. Farhadi, D. Fox, and R. Krishna. Molmoact: Action reasoning models that can reason in space.arXiv preprint arXiv:2508.07917, 2025. doi:10.48550/arXiv.2508.07917. URLhttps://arxiv.org/abs/ 2508.07917

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2508.07917 2025

  20. [22]

    H. Fang, J. Duan, D. Clay, S. Wang, S. Liu, W. Huang, X. Fan, W.-C. Tsai, S. Chen, Y . R. Wang, S. Xing, J. Cho, J. S. Park, A. Eftekhar, P. Sushko, K. Farley, A. Wadhwa, C. Harrison, W. Han, Y .-C. Lee, E. VanderBilt, R. Hendrix, S. Ellawela, L. Ngoo, J. Chai, Z. Ren, A. Farhadi, D. Fox, and R. Krishna. Molmoact2: Action reasoning models for real-world d...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2605.02881 2026

  21. [23]

    Zhong, Y

    L. Zhong, Y . Liu, Y . Wei, Z. Xiong, M. Yao, S. Liu, and G. Ren. Acot-vla: Action chain- of-thought for vision-language-action models.arXiv preprint arXiv:2601.11404, 2026. doi: 10.48550/arXiv.2601.11404. URLhttps://arxiv.org/abs/2601.11404

    work page doi:10.48550/arxiv.2601.11404 2026

  22. [24]

    Y . Liu, S. Zhang, Z. Dong, B. Ye, T. Yuan, X. Yu, L. Yin, C. Lu, J. Shi, L. J.-T. Yu, L. Zheng, T. Jiang, J. Gong, X. Qiu, and H. Zhao. Faster: Toward efficient autoregressive vision language action modeling via neural action tokenization.arXiv preprint arXiv:2512.04952, 2025. doi: 10.48550/arXiv.2512.04952. URLhttps://arxiv.org/abs/2512.04952

    work page doi:10.48550/arxiv.2512.04952 2025

  23. [25]

    Z. Dong, Y . Liu, S. Zhang, B. Ye, Y . Yuan, F. Ni, J. Gong, X. Qiu, H. Zhao, Y . Li, et al. Actioncodec: What makes for good action tokenizers.arXiv preprint arXiv:2602.15397, 2026

    arXiv 2026

  24. [26]

    Y . Wang, H. Zhu, M. Liu, J. Yang, H.-S. Fang, and T. He. Vq-vla: Improving vision-language- action models via scaling vector-quantized action tokenizers. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 11089–11099, 2025

    2025

  25. [27]

    B. Liu, Y . Zhu, C. Gao, Y . Feng, Q. Liu, Y . Zhu, and P. Stone. Libero: Benchmarking knowledge transfer for lifelong robot learning. InAdvances in Neural Information Processing Systems, volume 36, 2023. URLhttps://proceedings.neurips.cc/paper_files/ paper/2023/hash/8c3c666820ea055a77726d66fc7d447f-Abstract-Datasets_and_ Benchmarks.html. 10

    2023

  26. [28]

    X. Li, K. Hsu, J. Gu, K. Pertsch, O. Mees, H. R. Walke, C. Fu, I. Lunawat, I. Sieh, S. Kir- mani, S. Levine, J. Wu, C. Finn, H. Su, Q. Vuong, and T. Xiao. Evaluating real-world robot manipulation policies in simulation.arXiv preprint arXiv:2405.05941, 2024

    Pith/arXiv arXiv 2024

  27. [29]

    SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics

    M. Shukor, D. Aubakirova, F. Capuano, P. Kooijmans, S. Palma, A. Zouitine, M. Ar- actingi, C. Pascal, M. Russi, A. Marafioti, S. Alibert, M. Cord, T. Wolf, and R. Cadene. Smolvla: A vision-language-action model for affordable and efficient robotics.arXiv preprint arXiv:2506.01844, 2025. doi:10.48550/arXiv.2506.01844. URLhttps://arxiv.org/abs/ 2506.01844

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.01844 2025

  28. [30]

    GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

    NVIDIA, J. Bjorck, F. Casta ˜neda, N. Cherniadev, X. Da, R. Ding, L. J. Fan, Y . Fang, D. Fox, F. Hu, S. Huang, J. Jang, Z. Jiang, J. Kautz, K. Kundalia, L. Lao, Z. Li, Z. Lin, K. Lin, G. Liu, E. Llontop, L. Magne, A. Mandlekar, A. Narayan, S. Nasiriany, S. Reed, Y . L. Tan, G. Wang, Z. Wang, J. Wang, Q. Wang, J. Xiang, Y . Xie, Y . Xu, Z. Xu, S. Ye, Z. Y...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.14734 2025

  29. [31]

    J. Cen, C. Yu, H. Yuan, Y . Jiang, S. Huang, J. Guo, X. Li, Y . Song, H. Luo, F. Wang, D. Zhao, and H. Chen. Worldvla: Towards autoregressive action world model.arXiv preprint arXiv:2506.21539, 2025. doi:10.48550/arXiv.2506.21539. URLhttps://arxiv.org/abs/ 2506.21539

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2506.21539 2025

  30. [32]

    Y . Wang, X. Li, W. Wang, J. Zhang, Y . Li, Y . Chen, X. Wang, and Z. Zhang. Unified vision- language-action model.arXiv preprint arXiv:2506.19850, 2025. doi:10.48550/arXiv.2506. 19850. URLhttps://arxiv.org/abs/2506.19850

    work page doi:10.48550/arxiv.2506 2025

  31. [33]

    Q. Lv, W. Kong, H. Li, J. Zeng, Z. Qiu, D. Qu, H. Song, Q. Chen, X. Deng, and J. Pang. F1: A vision-language-action model bridging understanding and generation to actions.arXiv preprint arXiv:2509.06951, 2025. doi:10.48550/arXiv.2509.06951. URLhttps://arxiv. org/abs/2509.06951

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2509.06951 2025

  32. [34]

    J. Chen, W. Song, P. Ding, Z. Zhou, H. Zhao, F. Tang, D. Wang, and H. Li. Unified diffusion vla: Vision-language-action model via joint discrete denoising diffusion process. arXiv preprint arXiv:2511.01718, 2025. doi:10.48550/arXiv.2511.01718. URLhttps: //arxiv.org/abs/2511.01718

    work page doi:10.48550/arxiv.2511.01718 2025

  33. [35]

    Q. Li, Y . Liang, Z. Wang, L. Luo, X. Chen, M. Liao, F. Wei, Y . Deng, S. Xu, Y . Zhang, X. Wang, B. Liu, J. Fu, J. Bao, D. Chen, Y . Shi, J. Yang, and B. Guo. Cogact: A foundational vision-language-action model for synergizing cognition and action in robotic manipulation. arXiv preprint arXiv:2411.19650, 2024. doi:10.48550/arXiv.2411.19650. URLhttps:// a...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2411.19650 2024

  34. [36]

    K. Li, L. Clouatre, R. Pathak, and S. Ermon. Chain of thought empowers transformers to solve inherently serial problems.arXiv preprint arXiv:2402.12875, 2024. doi:10.48550/arXiv.2402. 12875. URLhttps://arxiv.org/abs/2402.12875

    work page doi:10.48550/arxiv.2402 2024

  35. [37]

    M. Nye, A. J. Andreassen, G. Gur-Ari, H. Michalewski, J. Austin, D. Bieber, D. Do- han, A. Lewkowycz, M. Bosma, D. Luan, C. Sutton, L. Soares, Y . Hu, D. Chen, and O. Habryka. Show your work: Scratchpads for intermediate computation with language models.arXiv preprint arXiv:2112.00114, 2021. doi:10.48550/arXiv.2112.00114. URL https://arxiv.org/abs/2112.00114

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2112.00114 2021

  36. [38]

    X. Wang, J. Wei, D. Schuurmans, Q. V . Le, E. H. Chi, and S. Narang. Self-consistency im- proves chain of thought reasoning in language models.arXiv preprint arXiv:2203.11171, 2023. doi:10.48550/arXiv.2203.11171. URLhttps://arxiv.org/abs/2203.11171. 11

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2203.11171 2023

  37. [39]

    K. Tian, Y . Jiang, Z. Yuan, B. Peng, and L. Wang. Visual autoregressive modeling: Scalable image generation via next-scale prediction.arXiv preprint arXiv:2404.02905, 2024. doi:10. 48550/arXiv.2404.02905. URLhttps://arxiv.org/abs/2404.02905

    arXiv 2024

  38. [40]

    BridgeData V2: A dataset for robot learning at scale

    H. Walke, K. Black, A. Lee, M. J. Kim, M. Du, C. Zheng, T. Zhao, P. Hansen-Estruch, Q. Vuong, A. He, V . Myers, K. Fang, C. Finn, and S. Levine. Bridgedata v2: A dataset for robot learning at scale.arXiv preprint arXiv:2308.12952, 2023. doi:10.48550/arXiv.2308.12952. URLhttps://arxiv.org/abs/2308.12952

    work page doi:10.48550/arxiv.2308.12952 2023

  39. [41]

    QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation

    D. Kalashnikov, A. Irpan, P. Pastor, J. Ibarz, A. Herzog, E. Jang, D. Quillen, E. Holly, M. Kalakrishnan, V . Vanhoucke, and S. Levine. Qt-opt: Scalable deep reinforcement learn- ing for vision-based robotic manipulation.arXiv preprint arXiv:1806.10293, 2018. doi: 10.48550/arXiv.1806.10293. URLhttps://arxiv.org/abs/1806.10293

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1806.10293 2018

  40. [42]

    Khazatsky, K

    A. Khazatsky, K. Pertsch, S. Nair, A. Balakrishna, S. Dasari, S. Karamcheti, S. Nasiriany, et al. Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945,

    Pith/arXiv arXiv

  41. [43]

    DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset

    doi:10.48550/arXiv.2403.12945. URLhttps://arxiv.org/abs/2403.12945. 12 Appendix A Additional Ablations The main paper reports the core simulation, real-world, and design-ablation results. Here we provide additional diagnostic breakdowns and implementation details that support the design choices. A.1 LIBERO Design Ablations Planning Horizon.Table 3 reports...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2403.12945 2048