Pith. sign in

REVIEW 4 major objections 7 minor 87 references

AtlasVLA: Persistent World-Ego State Modeling for Vision-Language-Action Models

T0 review · 4 major / 7 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 A persistent world-ego state lets a wrist-only VLA beat multi-view baselines on manipulation benchmarks.

desk verdict A coherent dual-memory VLA with strong reported results, but the paper never measures the fidelity of the 3D map that carries the method's weight. read the letter →

arxiv 2608.06729 v1 pith:NI7QBSOD submitted 2026-08-07 cs.RO cs.CV

Guiyu Zhao , Longteng Guo , Yanghong Mei , Zilin Zhu , Yu Zhang , Bin Cao , Mingming Yu , Xingjian He
show 2 more authors
Jie Jiang Jing Liu
This is my paper · ORCID
classification cs.ROcs.CV
keywords vision-language-actionmodelspersistentworldstatememoryego-workingwrist-mountedcamerapartialobservabilitylong-horizonmanipulationvoxelhashingdiffusiontransformer
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 the reactive observe-then-act loop of current vision-language-action (VLA) models is the root cause of two failures in wrist-only robot manipulation: objects vanish from the narrow camera view, and the model loses track of which sub-steps it has already completed. To fix this, AtlasVLA augments the policy with a persistent world-ego state built from a dual-memory architecture. The paper claims that with only a wrist camera, AtlasVLA reaches 97.6% average success on LIBERO and 70.8% on RLBench, and beats representative multi-view baselines by 9.4% absolute points on LIBERO-Long and 17.5% on real-world long-horizon tasks. If the result holds, it means a VLA can overcome partial observability without any external cameras by continuously reconstructing the scene from transient egocentric views.

What carries the argument

The load-bearing object is the 4D Persistent World State Memory: a voxel-hashed spatial map that receives 2D visual tokens lifted into 3D via monocular depth (DepthAnything v3) and camera extrinsics derived from robot kinematics and hand-eye calibration. Incoming tokens are fused into voxels by confidence-weighted averaging (TSDF-style), decayed by a sliding temporal window, and permanently anchored to the first frame. A second mechanism, the Ego-Working Memory, condenses historical context into a small bank of learnable intent tokens that track task progress. These two memories condition a diffusion transformer (DiT) action decoder through step-wise cross-attention, replacing global conditioning with sequential ego-working and world-state attention.

What would settle it

Measure the alignment of the persistent world state against ground-truth geometry during a wrist-only trajectory, for example by comparing the voxel-hashed reconstruction with an external depth sensor's point cloud; if the mean alignment error is comparable to or larger than the 0.025 m voxel size even on successful runs, the claimed spatial-fusion mechanism is not what drives the reported performance.

Watch

Extended reading notes

Core claim

AtlasVLA's central claim is that maintaining a persistent, globally updated 4D world state plus an ego-working memory of task progress lets a VLA operate reliably from a single wrist camera, where standard reactive VLAs fail. The paper demonstrates this by showing state-of-the-art success rates on LIBERO, RLBench, and real-world tasks using exclusively wrist observations, and by showing that removing either memory module causes large drops in real-world long-horizon success (world memory removal: 69.5% to 54.0%; ego memory removal: 69.5% to 56.5%). It further claims that explicit world state conditioning is indispensable, since removing world-state attention drops real-world long-horizon success by 8.0%.

Load-bearing premise

The entire advantage depends on monocular depth estimates and hand-eye and kinematic calibration being accurate and temporally consistent enough that back-projected tokens from different views land in the same voxels; the paper reports no direct metric of the resulting map's geometric accuracy.

Editorial extensions

If this is right

  • If the central claim is correct, wrist-only VLAs can match or beat multi-view models, removing the need for third-person cameras in many tabletop manipulation setups.
  • Long-horizon tasks should no longer degrade from accumulation of forgetting errors; the ego-working memory provides a bounded, scalable history that tracks sub-goal completion.
  • The persistent world state could serve as a shared substrate for other embodied capabilities beyond action decoding, such as spatial reasoning, collision avoidance, and closed-loop replanning.
  • The dual-memory design suggests a general recipe for VLA architectures: make the world model explicit in latent space rather than relying on context windows or implicit recurrent state.
  • The ablation results imply that memory update strategy matters as much as memory existence: naive accumulation of observations is substantially worse than TSDF-style voxel fusion with a sliding window.

Reading between the lines

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

  • A testable extension would be to evaluate AtlasVLA's voxel map against ground-truth geometry; the paper reports no direct metric of map accuracy, so a reader cannot tell whether residual failures come from depth error, calibration drift, or the fusion strategy itself.
  • The permanent first-frame anchor makes sense for static tabletop scenes, but in tasks where objects are moved in and out or removed, an immutable first-frame anchor could actively mislead the policy; the paper does not address dynamic scene changes.
  • The memory architecture is agnostic to the source of depth: if the wrist camera were replaced by an RGB-D sensor or a stereo pair, the same voxel-hash fusion could apply with more reliable geometry, which would likely strengthen the observed gains.
  • The same world-ego memory could be applied beyond manipulation to tasks that require object permanence and hidden-state inference, such as embodied search or navigation, but the paper only evaluates manipulation benchmarks.
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 / 7 minor

Summary. AtlasVLA proposes a dual-memory architecture for wrist-only vision-language-action models: a 4D Persistent World State Memory that lifts monocular 2D observations into a voxel-hashed 3D latent map via Depth Anything v3 depth and hand-eye extrinsics, and an Ego-Working State Memory that tracks task progress with learnable intent queries. The policy is a step-wise conditioned diffusion transformer that attends to both memories. The paper evaluates on LIBERO, RLBench, and real-world Franka tasks, reporting state-of-the-art success rates under a strict wrist-only setting, and claims to outperform multi-view baselines and other memory-augmented VLAs, especially in long-horizon tasks.

Significance. If the reported results hold, the paper addresses a real deployment bottleneck: wrist-only VLAs suffer from partial observability, and a persistent spatial memory could reduce the need for multi-camera setups. The ablation design is a genuine strength: each core component (world memory, ego memory, spatial/temporal PE, world-state conditioning) is removed in turn and consistently degrades performance, providing internal support for the proposed mechanism. The empirical claim is conditional, however, because the central 3D reconstruction step is never directly validated and the reported differences often lack statistical evidence. The work would be a useful step for the robotic-learning community if the map-fidelity and statistical issues are resolved.

major comments (4)
  1. [§3.2, Eqs. (4)–(5)] The update rule is internally inconsistent. Eq. (4) normalizes the new memory value by W_{t-1}(v)+w_t(v), but Eq. (5) stores W_t(v)=λ W_{t-1}(v)+w_t(v). For any λ<1, the weight used in the normalization differs from the stored cumulative weight, so M_t(v) is not the claimed weighted aggregation and the sliding-window forgetting is not actually implemented in the memory value. The paper never specifies λ or states that λ=1, and λ is absent from Table 6. Please reconcile the equations (e.g., replace W_{t-1} by λW_{t-1} in Eq. (4) or explicitly set λ=1) and give the value or schedule used in practice.
  2. [§3.2 and §4.2–4.4] The central mechanism of AtlasVLA is the geometric fidelity of the persistent world state, yet no experiment directly evaluates it. The map is built from monocular depth estimates and hand-eye extrinsics, but the paper reports no depth-error, map-alignment, voxel-consistency, or drift metric. LIBERO and RLBench provide ground-truth depth and known camera poses, so such an evaluation is straightforwardly feasible. Without it, the ablation 'w/o World State Memory' in Table 5 only demonstrates that some memory mechanism improves performance; it does not show that the 3D back-projection and voxel fusion are geometrically faithful. A noisy feature bank could produce the same qualitative benefit even if the geometry is wrong, which would undermine the specific claim that the 4D world state resolves visual blind spots.
  3. [Tables 1–5 and 8–9] All success rates are reported without error bars, confidence intervals, or significance tests, while per-condition trial counts range from 20 (RLBench) to 50 (LIBERO and real-world). With binomial sampling error of roughly 3–7 percentage points, several reported differences are likely noise: for example, Table 5 rows 6/7 show 96.4 vs 96.8 on LIBERO and 67.5 vs 65.0 in real-world long-horizon; Table 9 shows 97.6 vs 97.2 for voxel sizes 0.025 and 0.05. The manuscript should report per-task trial counts, binomial confidence intervals (or repeated-seed statistics), and preferably paired tests for ablation comparisons, and should avoid drawing conclusions from gaps smaller than the measurement resolution.
  4. [Table 1 and §4.2] The headline average comparison is unfair. For methods without LIBERO-90 results (π0, 4D-VLA, OpenVLA-OFT, GE-ACT), the reported average is over only the first four suites, whereas AtlasVLA's average includes all five suites. If LIBERO-90 has different difficulty, this mixes different aggregates; the text's claim that AtlasVLA is '3.4% higher than π0' is therefore not apples-to-apples. Please report the average over the common subset of suites for all methods as the primary comparison, and give per-suite numbers so readers can verify the ranking.
minor comments (7)
  1. [Abstract] The abstract contains a stray curly brace in 'temporal task-progress forgetting}'; remove it.
  2. [§4.2] The sentence 'π0 and MemoryVLA dropping by 3.5% and 2.5% respectively' is inconsistent with Table 1: π0's third-person-to-wrist change is an increase, while 3rd+wrist-to-wrist is the -3.5% drop. Clarify which baselines are being compared.
  3. [§3.2] The term 'TSDF-inspired' is imprecise, because TSDF integration maintains signed distances while the method aggregates latent feature vectors. Consider 'voxel-based weighted averaging with a sliding window' or another descriptor.
  4. [§3.3, Eq. (7)] The consolidation operator Cons(·) is not defined; specify how temporal adjacency and semantic similarity are measured and how merged tokens are formed.
  5. [Table 6] The decay factor λ used in Eq. (5) does not appear in the hyperparameter table; add it.
  6. [References and body text] The notation for the depth model is inconsistent: the body text says 'DepthAnything v3' while the reference is listed as 'DepthAnything3' (Lin et al. 2025). Align the naming.
  7. [Figure 2] Figure 2 is visually overloaded and difficult to read at the reproduction scale; a simplified schematic highlighting the data flow of the two memory update loops would improve accessibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: AtlasVLA's central claims are empirical benchmark results, and its memory components are evaluated by external benchmarks and ablations rather than derived from the quantities they are said to predict.

full rationale

AtlasVLA is an empirical architecture paper. Its main claim is that a wrist-only VLA augmented with a persistent world-ego memory achieves high success rates on LIBERO, RLBench, and real-world tasks. These numbers are evaluated against external benchmarks and internal ablations, not derived from a fitted parameter that is later renamed as a prediction. The world-state memory in Sec. 3.2 is constructed from DepthAnything v3 depth estimates, robot state, and hand-eye calibration; the policy is trained on task demonstrations and evaluated on held-out rollouts, so the quoted success rates are not encoded by construction. The ablations in Table 5 show that removing the world-state memory, ego-working memory, or world-state conditioning degrades performance, which is an independent check on the mechanism. The internal inconsistency between Eq. 4 and Eq. 5 concerning the decayed weight update is a technical correctness concern, not a circularity. The lack of a direct map-accuracy metric is an evidence gap, but it does not mean the argument reduces to its inputs. Cited prior works such as pi0, MemoryVLA, and DepthAnything v3 are used as baselines, background, or off-the-shelf components, not as load-bearing justifications for the paper's own conclusions. No self-definitional, fitted-input, or uniqueness-imported-from-authors pattern is present under the standard for flagging circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 2 invented entities

The paper is an empirical architecture paper rather than a derivation of a physical law or constant. The central claim rests on assumptions about depth quality, hand-eye calibration, semantic preservation during voxel averaging, and the reliability of the first-frame anchor. In addition, several hyperparameters, especially voxel size and ego memory length, are tuned on the same benchmarks that produce the final headline numbers, so the evaluation is not fully independent of model selection.

free parameters (7)
  • world voxel size = 0.025 m
    Selected via ablation on LIBERO and real-world long-horizon tasks (Appendix B.4, Table 9); central to the fusion granularity of the world memory.
  • ego-working memory length = 16
    Selected via ablation comparing 8, 16, and 32 (Appendix B.4, Table 8); directly controls how much task progress history is retained.
  • world state memory capacity = 2048 tokens
    Chosen by design (Table 6); bounds the persistent world state and affects how many observations can be fused.
  • number of ego-working tokens = 4
    Design hyperparameter (Table 6); determines the capacity of the intent query set.
  • CFG scale = 1.5
    Inference hyperparameter (Table 6) chosen for classifier-free guidance strength during action denoising.
  • inference denoising steps = 10
    DDIM sampling steps (Table 6); affects action quality and latency.
  • action chunk size = 16
    Training and inference hyperparameter (Table 6); determines the horizon of predicted actions.
assumptions (5)
  • domain assumption Monocular depth from DepthAnything v3 is accurate and temporally consistent enough for back-projection of wrist-camera features.
    The entire world state memory construction in Sec. 3.2 and Eq. 2 relies on depth estimates being correct; inaccurate depth would corrupt the fused voxel state.
  • domain assumption Camera extrinsics computed from robot state and hand-eye calibration are correct at every timestep.
    Sec. 3.2 derives T_ex from psi(S_t) and the hand-eye matrix T_h2e; any calibration or kinematic error misaligns back-projected tokens.
  • ad hoc to paper Averaging latent visual features across views within a voxel preserves task-relevant object semantics.
    Eq. 4 performs weighted averaging of high-dimensional latent tokens in voxels; the paper assumes this retains the information needed for later action generation, with no direct evaluation of map quality.
  • ad hoc to paper The first frame provides a reliable permanent anchor for the global workspace.
    The 'permanent initialization' rule in Sec. 3.2 assumes the initial wrist view contains the most useful global context; if the first frame is occluded or has poor depth, the anchored state may be wrong and never corrected.
  • domain assumption Success rates in the LIBERO and RLBench simulators transfer to useful real-world manipulation behavior.
    The paper uses simulation results as evidence for the real-world value of the memory architecture, while real-world evaluation is limited to a single robot platform and setup in Sec. 4.4.
invented entities (2)
  • 4D Persistent World State Memory
    purpose: Stores a fused, voxel-hashed latent representation of the workspace that persists as objects leave the wrist camera field of view.
    The quality of this internal map is never measured directly against ground truth; its only observable consequence is end-task success, so there is no falsifiable handle outside the paper's own pipeline.
  • Ego-Working State Memory
    purpose: Tracks historical ego state and subtask progress using learned intent queries consolidated into a memory bank.
    No probe or diagnostic shows what the memory actually encodes; the only evidence is aggregate success-rate improvement in ablations, not independent verification of memory content.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AtlasVLA: Persistent World-Ego State Modeling for Vision-Language-Action Models." pith.science (2026). https://pith.science/paper/NI7QBSOD

@misc{pith2026260806729,
  author       = {Pith},
  title        = {Pith review of: AtlasVLA: Persistent World-Ego State Modeling for Vision-Language-Action Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NI7QBSOD}},
  note         = {Machine review of arXiv:2608.06729}
}
read the original abstract

While Vision-Language-Action (VLA) models have advanced embodied AI, their fundamentally reactive paradigm severely limits performance in partially observable and long-horizon tasks. When restricted to a single wrist-mounted camera, they inevitably suffer from perception forgetting as objects exit the field of view, and temporal task-progress forgetting} during multi-step execution. To overcome these bottlenecks, we propose AtlasVLA, a novel framework that transitions from direct reactive manipulation to proactive reasoning through a persistent world-ego state. AtlasVLA features a dual-memory architecture: a 4D Persistent World State Memory that lifts transient 2D observations into a globally updated, voxel-hashed spatial state to resolve visual blind spots, and an Ego-Working State Memory that tracks historical ego state and task progress. By conditioning a diffusion transformer (DiT) on this joint World-Ego state, AtlasVLA enables robust spatial reasoning. Extensive evaluations across LIBERO, RLBench, and real-world benchmarks demonstrate that AtlasVLA achieves state-of-the-art performance using solely a wrist camera. Remarkably, it decisively outperforms multi-view baselines, yielding absolute success rate improvements of 9.4% on LIBERO-Long and 17.5% in real-world long-horizon tasks.

Figures

Figures reproduced from arXiv: 2608.06729 by the authors.

Figure 1
Figure 1. The dual bottlenecks of current reactive VLAs versus the advantage of AtlasVLA. (A) Partial observation: Wrist-only VLAs suffer from limited FoV and partial observation during manipulation. (B) Task-progress forgetting: Reactive baselines lack temporal context, forgetting completed task progress and state. (C) Our solution: Our AtlasVLA integrates a 4D Persistent World State for spatial awareness and an Ego-Working … view at source ↗
Figure 2
Figure 2. Overall Architecture of AtlasVLA. Relying solely on wrist-camera images I w t , AtlasVLA extracts visual tokens Xw t and lifts them into 3D via depth Dw t and extrinsics Tex t to form an instantaneous world state memory mt. This memory dynamically updates the persistent world state memory Mt−1 via spatial neighborhood fusion and a temporal sliding window. Concurrently, an ego-working memory condenses context into eg… view at source ↗
Figure 3
Figure 3. Qualitative results of AtlasVLA on real-world long-horizon tasks. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Our real-world robotic platform. B.2 Additional Real-world Settings Hardware Setup. All real-world evaluations are con￾ducted using a 7-DoF Franka robotic manipulator. Depart￾ing from conventional multi-camera configurations, our sys￾tem strictly enforces a wrist-only …
Figure 5
Figure 5. Figure 5: Qualitative results of AtlasVLA on real-world long-horizon tasks. [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]
Figure 6
Figure 6. Figure 6: Qualitative results of AtlasVLA on real-world general tasks. [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results of AtlasVLA on LIBERO Benchmark. [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: More Qualitative results of AtlasVLA on long LIBERO-10 Benchmark. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

87 extracted references · 8 canonical work pages

  1. [1]

    arXiv preprint arXiv:2407.07726 , year=

    Paligemma: A versatile 3b vlm for transfer , author=. arXiv preprint arXiv:2407.07726 , year=

    arXiv

  2. [2]

    arXiv preprint arXiv:2502.19645 , year=

    Fine-tuning vision-language-action models: Optimizing speed and success , author=. arXiv preprint arXiv:2502.19645 , year=

    arXiv

  3. [3]

    IEEE Robotics and Automation Letters , year=

    Tinyvla: Towards fast, data-efficient vision-language-action models for robotic manipulation , author=. IEEE Robotics and Automation Letters , year=

  4. [4]

    arXiv preprint arXiv:2412.14058 , year=

    Towards generalist robot policies: What matters in building vision-language-action models , author=. arXiv preprint arXiv:2412.14058 , year=

    arXiv

  5. [5]

    arXiv preprint arXiv:2503.10631 , year=

    Hybridvla: Collaborative diffusion and autoregression in a unified vision-language-action model , author=. arXiv preprint arXiv:2503.10631 , year=

    arXiv

  6. [6]

    FLOWER: Democratizing Generalist Robot Policies with Efficient Vision-Language-Action Flow Policies , author=

  7. [7]

    arXiv preprint arXiv:2410.07864 , year=

    Rdt-1b: a diffusion foundation model for bimanual manipulation , author=. arXiv preprint arXiv:2410.07864 , year=

    arXiv

  8. [8]

    arXiv preprint arXiv:2406.09246 , year=

    Openvla: An open-source vision-language-action model , author=. arXiv preprint arXiv:2406.09246 , year=

    arXiv

Show all 87 references
  1. [9]

    Conference on Robot Learning , pages=

    Rt-2: Vision-language-action models transfer web knowledge to robotic control , author=. Conference on Robot Learning , pages=. 2023 , organization=

    2023

  2. [10]

    arXiv preprint arXiv:2212.06817 , year=

    Rt-1: Robotics transformer for real-world control at scale , author=. arXiv preprint arXiv:2212.06817 , year=

    arXiv

  3. [11]

    arXiv preprint arXiv:2403.01823 , year=

    Rt-h: Action hierarchies using language , author=. arXiv preprint arXiv:2403.01823 , year=

    arXiv

  4. [12]

    arXiv 2023 , author=

    Llama: Open and efficient foundation language models. arXiv 2023 , author=. arXiv preprint arXiv:2302.13971 , volume=

    2023 arXiv

  5. [13]

    Language models are few-shot learners , author=

  6. [14]

    Advances in neural information processing systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in neural information processing systems , volume=

  7. [15]

    2023 , journal =

    OpenAI , title =. 2023 , journal =

    2023

  8. [16]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  9. [17]

    Runpei Dong and Chunrui Han and Yuang Peng and Zekun Qi and Zheng Ge and Jinrong Yang and Liang Zhao and Jianjian Sun and Hongyu Zhou and Haoran Wei and Xiangwen Kong and Xiangyu Zhang and Kaisheng Ma and Li Yi , title =

  10. [18]

    Computer Vision -

    Zekun Qi and Runpei Dong and Shaochen Zhang and Haoran Geng and Chunrui Han and Zheng Ge and Li Yi and Kaisheng Ma , title =. Computer Vision -

  11. [19]

    arXiv preprint arXiv:2501.15830 , year=

    Spatialvla: Exploring spatial representations for visual-language-action model , author=. arXiv preprint arXiv:2501.15830 , year=

    arXiv

  12. [20]

    arXiv preprint arXiv:2508.09071 , year=

    Geovla: Empowering 3d representations in vision-language-action models , author=. arXiv preprint arXiv:2508.09071 , year=

    arXiv

  13. [21]

    IEEE Robotics and Automation Letters , volume=

    Pointvla: Injecting the 3d world into vision-language-action models , author=. IEEE Robotics and Automation Letters , volume=. 2026 , publisher=

    2026

  14. [22]

    arXiv preprint arXiv:2506.07961 , year=

    Bridgevla: Input-output alignment for efficient 3d manipulation learning with vision-language models , author=. arXiv preprint arXiv:2506.07961 , year=

  15. [23]

    arXiv preprint arXiv:2506.22242 , year=

    4d-vla: Spatiotemporal vision-language-action pretraining with cross-scene calibration , author=. arXiv preprint arXiv:2506.22242 , year=

  16. [24]

    arXiv preprint arXiv:2510.17439 , year=

    From spatial to actions: Grounding vision-language-action model in spatial foundation priors , author=. arXiv preprint arXiv:2510.17439 , year=

  17. [25]

    arXiv preprint arXiv:2403.09631 , year=

    3d-vla: A 3d vision-language-action generative world model , author=. arXiv preprint arXiv:2403.09631 , year=

    arXiv

  18. [26]

    arXiv preprint arXiv:2507.00416 , year=

    Evo-0: Vision-language-action model with implicit spatial understanding , author=. arXiv preprint arXiv:2507.00416 , year=

  19. [27]

    URL https://arxiv

    Alphamath almost zero: Process supervision without process, 2024 , author=. URL https://arxiv. org/abs/2405.03553 , year=

    2024 arXiv

  20. [28]

    Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    Confidence improves self-consistency in llms , author=. Findings of the Association for Computational Linguistics: ACL 2025 , pages=

    2025

  21. [29]

    arXiv preprint arXiv:2203.11171 , year=

    Self-consistency improves chain of thought reasoning in language models , author=. arXiv preprint arXiv:2203.11171 , year=

    arXiv

  22. [30]

    Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , pages=

    Improving minimum Bayes risk decoding with multi-prompt , author=. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing , pages=

    2024

  23. [31]

    arXiv preprint arXiv:2407.21787 , year=

    Large language monkeys: Scaling inference compute with repeated sampling , author=. arXiv preprint arXiv:2407.21787 , year=

    arXiv

  24. [32]

    arXiv preprint arXiv:2410.13816 , year=

    Steering your generalists: Improving robotic foundation models via value guidance , author=. arXiv preprint arXiv:2410.13816 , year=

    arXiv

  25. [33]

    arXiv preprint arXiv:2506.17811 , year=

    Robomonkey: Scaling test-time sampling and verification for vision-language-action models , author=. arXiv preprint arXiv:2506.17811 , year=

    arXiv

  26. [34]

    arXiv preprint arXiv:2510.05681 , year=

    Verifier-free Test-Time Sampling for Vision Language Action Models , author=. arXiv preprint arXiv:2510.05681 , year=

    arXiv

  27. [35]

    arXiv preprint arXiv:2510.10975 , year=

    RoVer: Robot Reward Model as Test-Time Verifier for Vision-Language-Action Model , author=. arXiv preprint arXiv:2510.10975 , year=

  28. [36]

    arXiv preprint arXiv:2601.00675 , year=

    RoboReward: General-Purpose Vision-Language Reward Models for Robotics , author=. arXiv preprint arXiv:2601.00675 , year=

  29. [37]

    arXiv preprint arXiv:2403.12945 , year=

    Droid: A large-scale in-the-wild robot manipulation dataset , author=. arXiv preprint arXiv:2403.12945 , year=

    arXiv

  30. [38]

    Conference on Robot Learning , pages=

    Bridgedata v2: A dataset for robot learning at scale , author=. Conference on Robot Learning , pages=. 2023 , organization=

    2023

  31. [39]

    arXiv preprint arXiv:2412.13877 , year=

    Robomind: Benchmark on multi-embodiment intelligence normative data for robot manipulation , author=. arXiv preprint arXiv:2412.13877 , year=

    arXiv

  32. [40]

    Advances in Neural Information Processing Systems , volume=

    Libero: Benchmarking knowledge transfer for lifelong robot learning , author=. Advances in Neural Information Processing Systems , volume=

  33. [41]

    arXiv preprint arXiv:2410.24164 , year=

    pi0: A Vision-Language-Action Flow Model for General Robot Control , author=. arXiv preprint arXiv:2410.24164 , year=

    arXiv

  34. [42]

    arXiv preprint arXiv:2504.16054 , year=

    pi0.5: a Vision-Language-Action Model with Open-World Generalization , author=. arXiv preprint arXiv:2504.16054 , year=

    arXiv

  35. [43]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Vggt: Visual geometry grounded transformer , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  36. [44]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Dust3r: Geometric 3d vision made easy , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  37. [45]

    Advances in Neural Information Processing Systems , volume=

    Depth anything v2 , author=. Advances in Neural Information Processing Systems , volume=

  38. [46]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Depth anything: Unleashing the power of large-scale unlabeled data , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  39. [47]

    arXiv preprint arXiv:2302.12288 , year=

    Zoedepth: Zero-shot transfer by combining relative and metric depth , author=. arXiv preprint arXiv:2302.12288 , year=

    arXiv

  40. [48]

    arXiv preprint arXiv:2503.03480 , year=

    Safevla: Towards safety alignment of vision-language-action model via constrained learning , author=. arXiv preprint arXiv:2503.03480 , year=

    arXiv

  41. [49]

    arXiv preprint arXiv:2405.05941 , year=

    Evaluating Real-World Robot Manipulation Policies in Simulation , author=. arXiv preprint arXiv:2405.05941 , year=

    arXiv

  42. [50]

    arXiv preprint arXiv:2507.04447 , year=

    Dreamvla: a vision-language-action model dreamed with comprehensive world knowledge , author=. arXiv preprint arXiv:2507.04447 , year=

    arXiv

  43. [51]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  44. [52]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Kpconv: Flexible and deformable convolution for point clouds , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  45. [53]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. Iclr , volume=

  46. [54]

    arXiv preprint arXiv:2405.12213 , year=

    Octo: An open-source generalist robot policy , author=. arXiv preprint arXiv:2405.12213 , year=

    arXiv

  47. [55]

    arXiv preprint arXiv:2508.19236 , year=

    Memoryvla: Perceptual-cognitive memory in vision-language-action models for robotic manipulation , author=. arXiv preprint arXiv:2508.19236 , year=

    arXiv

  48. [56]

    2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration 0 , author=. 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2024 , organization=

    2024

  49. [57]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Llava-3d: A simple yet effective pathway to empowering lmms with 3d capabilities , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  50. [58]

    arXiv preprint arXiv:2307.09288 , year=

    Llama 2: Open foundation and fine-tuned chat models , author=. arXiv preprint arXiv:2307.09288 , year=

    arXiv

  51. [59]

    Advances in neural information processing systems , volume=

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space , author=. Advances in neural information processing systems , volume=

  52. [60]

    European conference on computer vision , pages=

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  53. [61]

    Advances in Neural Information Processing Systems , volume=

    Spatialrgpt: Grounded spatial reasoning in vision-language models , author=. Advances in Neural Information Processing Systems , volume=

  54. [62]

    9th Annual Conference on Robot Learning , year=

    3ds-vla: A 3d spatial-aware vision language action model for robust multi-task manipulation , author=. 9th Annual Conference on Robot Learning , year=

  55. [63]

    arXiv preprint arXiv:2511.10647 , year=

    Depth anything 3: Recovering the visual space from any views , author=. arXiv preprint arXiv:2511.10647 , year=

    arXiv

  56. [64]

    arXiv preprint arXiv:2010.11929 , year=

    An image is worth 16x16 words: Transformers for image recognition at scale , author=. arXiv preprint arXiv:2010.11929 , year=

    2010 arXiv

  57. [65]

    IEEE Robotics and Automation Letters , volume=

    Look closer: Bridging egocentric and third-person views with transformers for robotic manipulation , author=. IEEE Robotics and Automation Letters , volume=. 2022 , publisher=

    2022

  58. [66]

    arXiv preprint arXiv:2603.12942 , year=

    Remem-vla: Empowering vision-language-action model with memory via dual-level recurrent queries , author=. arXiv preprint arXiv:2603.12942 , year=

  59. [67]

    arXiv preprint arXiv:2511.09516 , year=

    Map-vla: Memory-augmented prompting for vision-language-action model in robotic manipulation , author=. arXiv preprint arXiv:2511.09516 , year=

  60. [68]

    arXiv preprint arXiv:2605.22283 , year=

    Spatial Memory for Out-of-Vision Manipulation in Vision-Language-Action , author=. arXiv preprint arXiv:2605.22283 , year=

    arXiv

  61. [69]

    arXiv preprint arXiv:2503.06669 , year=

    Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems , author=. arXiv preprint arXiv:2503.06669 , year=

    arXiv

  62. [70]

    arXiv preprint arXiv:2411.19650 , year=

    Cogact: A foundational vision-language-action model for synergizing cognition and action in robotic manipulation , author=. arXiv preprint arXiv:2411.19650 , year=

    arXiv

  63. [71]

    arXiv preprint arXiv:2502.05855 , year=

    Dexvla: Vision-language model with plug-in diffusion expert for general robot control , author=. arXiv preprint arXiv:2502.05855 , year=

    arXiv

  64. [72]

    The International Journal of Robotics Research , volume=

    Diffusion policy: Visuomotor policy learning via action diffusion , author=. The International Journal of Robotics Research , volume=. 2025 , publisher=

    2025

  65. [73]

    International Conference on Learning Representations , volume=

    Rdt-1b: a diffusion foundation model for bimanual manipulation , author=. International Conference on Learning Representations , volume=

  66. [74]

    arXiv preprint arXiv:2601.17885 , year=

    PEAfowl: Perception-Enhanced Multi-View Vision-Language-Action for Bimanual Manipulation , author=. arXiv preprint arXiv:2601.17885 , year=

  67. [75]

    arXiv preprint arXiv:2603.03596 , year=

    Mem: Multi-scale embodied memory for vision language action models , author=. arXiv preprint arXiv:2603.03596 , year=

  68. [76]

    International Conference on Machine Learning , pages=

    VIMA: Robot Manipulation with Multimodal Prompts , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  69. [77]

    2011 10th IEEE international symposium on mixed and augmented reality , pages=

    Kinectfusion: Real-time dense surface mapping and tracking , author=. 2011 10th IEEE international symposium on mixed and augmented reality , pages=. 2011 , organization=

    2011

  70. [78]

    arXiv preprint arXiv:2508.05635 , year=

    Genie envisioner: A unified world foundation platform for robotic manipulation , author=. arXiv preprint arXiv:2508.05635 , year=

    arXiv

  71. [79]

    arXiv preprint arXiv:2511.17502 , year=

    Rynnvla-002: A unified vision-language-action and world model , author=. arXiv preprint arXiv:2511.17502 , year=

    arXiv

  72. [80]

    Advances in Neural Information Processing Systems , volume=

    Fast-in-slow: A dual-system vla model unifying fast manipulation within slow reasoning , author=. Advances in Neural Information Processing Systems , volume=

  73. [81]

    International Conference on Learning Representations , year=

    Denoising Diffusion Implicit Models , author=. International Conference on Learning Representations , year=

  74. [82]

    arXiv preprint arXiv:2207.12598 , year=

    Classifier-Free Diffusion Guidance , author=. arXiv preprint arXiv:2207.12598 , year=

    arXiv

  75. [83]

    IEEE Robotics and Automation Letters , volume=

    Rlbench: The robot learning benchmark & learning environment , author=. IEEE Robotics and Automation Letters , volume=. 2020 , publisher=

    2020

  76. [84]

    arXiv preprint arXiv:2604.19683 , year=

    Mask World Model: Predicting What Matters for Robust Robot Policy Learning , author=. arXiv preprint arXiv:2604.19683 , year=

    arXiv

  77. [85]

    Transactions on Machine Learning Research Journal , year=

    Dinov2: Learning robust visual features without supervision , author=. Transactions on Machine Learning Research Journal , year=

  78. [86]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Sigmoid loss for language image pre-training , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  79. [87]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Scalable diffusion models with transformers , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

Pith tools

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