Pith. sign in

REVIEW 4 major objections 4 minor 61 references

XPolicyLab: A Unified Standard and Open Ecosystem for Robot Policy Evaluation and Deployment

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A single adapter contract can turn pairwise robot-policy integration into reusable per-policy adapters and per-environment clients, the paper claims, with one measured drop from five-plus hours to thirty minutes.

desk verdict A solid, useful infrastructure paper whose central O(N+M) scaling claim is architecturally plausible but empirically under-supported by a thin single-policy effort study; deserves peer review with concrete requests for scaling evidence. read the letter →

arxiv 2608.09892 v2 pith:DTYCNYRU submitted 2026-08-10 cs.RO

XPolicyLab Community , Tianxing Chen , Yue Chen , Tian Nian , Zijian Cai , Guangyu Chen , Wenwei Lin , Qiwei Liang
show 60 more authors
Zanxin Chen Peicheng Xiang Kailun Su Zixuan Li Junyuan Tang Yan Qin Qiangyu Chen Shaolong Zhu Xiang Li Jiahao Zhang Weijie Wan Baijun Chen Honghao Su Kehe Ye Shujia Liu Kaixuan Wang Haotian Liang Yunze Liu Mingleyang Li Yuran Wang Boyu Chen Hongzhe Bi Shuhe Huang Hengkai Tan Jisong Cai Yao Mu Jun Guo Xiaofeng Wang Zheng Zhu Weijie Ke Hengtao Li Yuhang Tang Xiaofan Li Ganlin Yang Zhangzheng Tu Shuai Yang Wenxuan Song Pengxiang Ding Kaidong Zhang Yu Sun Junliang Guo Tong Zhang Yixing Chen Rongxu Cui Zongzheng Zhang Haoxiang Ma Junhao Cai Haoyu Zhang Senqiao Yang Jinhui Ye Pengguang Chen Shu Liu Xiu Su Wenhan Fang Wenhao Li Yichao Cao Chengyao Wang Qiang Chen Ping Luo Wenbo Ding
This is my paper · ORCID
classification cs.RO
keywords robotpolicyevaluationdeploymentadaptercontractstandardizedobservationandactionschemasdependency-isolatedservingvision-language-actionmodelsbenchmarkintegrationreproducibility
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

XPolicyLab aims to fix the boundary between a trained robot policy and the systems that run it: simulators, benchmark suites, and physical robots. Today each policy ships its own inference stack, data conventions, and runtime interfaces, so evaluating N policies across M environments costs $O(NM)$ separate integrations, and silent discrepancies such as camera order or gripper scaling can corrupt comparisons. The paper proposes a minimal adapter contract, common observation and action schemas, and a dependency-isolated server/client bridge, so that policy-specific code is confined to one side of the boundary. It reports that the standard has integrated 42 policies, that the same adapters serve RoboTwin simulation, RoboDojo simulation, and RoboDojo physical evaluation, and that a controlled study measured integration effort for one representative policy dropping from over five hours from scratch to about two hours manually, and to thirty minutes when a coding agent executes the conformance procedure. If the claim holds, reproducible policy comparison stops being bottlenecked by reinventing how to feed each checkpoint to each environment.

What carries the argument

The load-bearing object is the policy adapter contract: four operations, update_obs, get_action, reset, and their batched forms, plus standardized observation, action, and trajectory schemas, with image decoding and transport-level work fixed on the serving side. It carries the argument by making the environment-facing loop identical across policies, which is what converts pairwise integrations into one adapter per policy and one client per environment.

What would settle it

Run the same integration protocol with several policies and several engineering teams unfamiliar with the project; if median time-to-first-rollout through the standard is not far below the from-scratch baseline, or if the environment-facing loop grows with each new policy family, the claim that the contract confines heterogeneity to the policy side would be falsified. A simpler check is to measure hand-written lines of environment-facing glue across adapters, which the paper predicts stays within a few lines of a fixed reference.

Watch

Extended reading notes

Core claim

The central discovery, on the paper's own terms, is that policy integration can be treated as conformance to a specification rather than as bespoke engineering. The contract fixes only the external boundary: update_obs receives a standardized observation, get_action returns a standardized action chunk, reset clears episode state, and batched variants support parallel evaluation, while everything inside, including network architecture, dependencies, checkpoints, preprocessing, and action decoding, stays policy-native. A dependency-isolated serving bridge lets the policy run in its own software environment while the environment client runs in the simulator or robot stack, locally or remotely, with retry-safe request identifiers so reconnects do not re-run inference. The evidence in the paper is that across 42 adapters the model-specific code varies by an order of magnitude but the environment-facing loop stays within a few lines of a fixed reference, and that one adapter is reused across RoboTwin simulation, RoboDojo simulation, and RoboDojo physical evaluation, with the measured effect being the effort drop from over five hours to about two hours to thirty minutes for one VLA policy.

Load-bearing premise

The headline time savings rest on a within-subject study with six engineers, all working on one policy, with no variance or significance testing and a fixed agent stack; if those numbers do not carry over to other policies and integrators, the central quantitative claim is unsupported.

Editorial extensions

If this is right

  • Every new policy needs one adapter and every new environment one client, so no pairing-specific glue code is required and the $O(NM)$ integration cost collapses to $O(N+M)$.
  • The same policy adapter runs in RoboTwin simulation, RoboDojo simulation, and RoboDojo physical evaluation without modifying the model-specific implementation.
  • Standardizing image decoding and action conventions removes a class of silent faults that currently make reported results hard to compare across policies.
  • Conformance can be encoded as machine-readable audit gates, so checking a submission against the standard becomes a deterministic procedure rather than a documentation-reading exercise.
  • If the effort study generalizes, a new policy can be up and running in about thirty minutes with agent assistance instead of a multi-hour integration session.

Reading between the lines

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

  • Editorial inference: if the $O(N+M)$ claim holds broadly, benchmark maintainers can stop shipping model-specific serving stacks entirely and spend scarce physical-robot time on trials rather than integration.
  • Editorial inference: the same contract could plausibly extend to data collection and training loops, since the paper already separates data conversion and training entry points from serving; that is a natural next boundary the paper leaves implicit.
  • Editorial inference: a wider replication with more policies and more integrators would turn the thirty-minute figure from a single-case measurement into a reliable planning estimate, an extension the paper's own framing invites.
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 / 4 minor

Summary. The paper presents XPolicyLab, a standard and open-source infrastructure for connecting robot policies to evaluation environments. It specifies common observation, action, and trajectory schemas, a minimal policy-adapter contract (update_obs, get_action, reset, and batched variants), and a dependency-isolated client/server serving architecture. The central claim is that this reduces integration cost from O(NM) pairwise integrations to O(N+M). The paper reports 42 integrated policy adapters, deployments on RoboTwin and RoboDojo simulation and real-robot leaderboards, and a within-subject study (N=6) in which connecting π0.5 to RoboDojo simulation took over five hours from scratch, about two hours with the manual standard, and about thirty minutes with agent-assisted skills. The paper also claims, without presenting the underlying data, that model-specific code varies by an order of magnitude across adapters while the environment-facing loop stays close to a fixed reference.

Significance. If the claims are substantiated, XPolicyLab would fill a real and widely felt gap: the boundary between trained policies and the simulators, benchmarks, and robots that execute them is indeed fragmented, and a standard schema-plus-adapter architecture with isolated serving is a sensible and potentially reusable solution. The paper deserves credit for releasing the infrastructure, standardizing conformance gates, packaging agent skills, and integrating a large number of policy families. However, the strongest quantitative claims are currently supported more by design reasoning and a single-case study than by evidence. The controlled experiment measures exactly one policy and one environment, with no variance reporting; the code-locality claim is asserted without measurements; and the portability evidence comes from benchmark ecosystems co-developed with overlapping authors. These are load-bearing gaps for the headline O(N+M) claim, although they are addressable within the manuscript's scope by adding data or by carefully qualifying the claims.

major comments (4)
  1. [VI-A, Table V] The controlled study measures one policy (π0.5) integrated into one environment (RoboDojo simulation), so it cannot support the claimed O(N+M) scaling. That claim requires that the marginal cost of adding a new policy be roughly independent of the number of existing environments, and vice versa; demonstrating it requires varying at least one of these dimensions across several cells. The paper itself states in Section IV that 'whether this breadth is genuinely absorbed by the contract, rather than merely tolerated, is an empirical question' and in Section VI-A that the study is 'an indicative case study rather than a controlled measurement.' As written, the abstract and conclusion present the five-hour-to-thirty-minute reduction as an unqualified result. Please either add measurements for additional policies and environments with per-participant variance, or explicitly limit the stated claims to the measured policy-environment pair.
  2. [IV, VII] The statement that 'model-specific code varies by an order of magnitude while the environment-facing loop stays within a few lines of a fixed reference' is asserted in the abstract and conclusion but no supporting measurements are reported. Section IV identifies the question as empirical and refers to Section VI-A for the answer, yet Section VI-A measures integration effort and lines of code for one policy, not the code-locality distribution across the 42 adapters. Please include the underlying measurements, such as a histogram or table of adapter code sizes and environment-client code sizes, or remove and qualify the claim.
  3. [V, VIII] The evidence for portability rests on RoboTwin and RoboDojo, whose benchmark papers (refs. [15] and [16]) share first-author lineage with this paper, and whose leaderboards are administered through XPolicyLab. Deployment in these co-developed settings does not independently demonstrate that the adapters port to third-party environment stacks. Please add an integration with at least one environment outside the project (for example, RLBench, LIBERO, ManiSkill2, or RoboArena) or clearly disclose the extent of the author overlap and explain why the co-developed deployment is still evidence of portability.
  4. [III-A, Abstract] The asymptotic claim 'reduces this cost to O(N+M)' conflates the number of software components with engineering effort. The architecture plausibly yields O(N+M) adapters and clients, but the effort of adding policy i or environment j still depends on how well the common schema absorbs that entity's idiosyncrasies. No formal cost model is given, and the empirical evidence covers one cell, so the claim should either be defined explicitly as a statement about the number of integration components or be replaced by the measured reduction for the evaluated pair.
minor comments (4)
  1. [Table I, Section VIII] The contributor list mentions policies (for example, RLDX-1, Wall-WM, Wall-OSS, and CSU-AI-0) that do not appear in Table I; please clarify whether these are integrated policies that were omitted from the table or are separately maintained integrations.
  2. [Table V] Please report the variance or per-participant values for the time measurements, not only medians, and state precisely how 'time-to-first-successful-rollout' and 'hand-written LoC' were measured, especially in the agent condition where code is generated and then reviewed.
  3. [Tables III and IV] The relationship between 'Score' and 'SR' should be defined explicitly; the text says the score credits partial task progress, but the units and normalization are not stated, which makes the leaderboard tables hard to interpret.
  4. [Abstract] The abstract says the integration effort drops from 'over five hours to two hours,' while Table V reports approximately two hours; the approximate nature of the manual-condition result should be preserved in the abstract.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the O(N+M) claim follows from the defined adapter/client architecture, and the effort study is an external measurement rather than a fitted or self-referential prediction.

full rationale

The paper's main complexity claim, that integrating N policies with M environments reduces to O(N+M), follows directly from the proposed construction: one adapter per policy plus one client per environment (Section III-E). It is therefore a design consequence rather than a prediction derived from fitted data. The controlled study (Section VI-A) measures time-to-first-successful-rollout under three conditions and is an empirical comparison, not a quantity defined in terms of the outcome it claims to support. The paper itself labels the result an 'indicative case study' and acknowledges residual familiarity with the tested policy, which limits external validity but does not make the measurement circular. The deployments on RoboTwin and RoboDojo are cited as external benchmarks; although those benchmarks share authors with XPolicyLab, the paper does not invoke any uniqueness theorem or hidden premise from those citations to force the standard's design. The unsupported assertions, such as model-specific code varying by an order of magnitude across the 42 adapters and the absence of a multi-policy/multi-environment scaling experiment, are evidentiary weaknesses, not instances of a derivation reducing to its own inputs. No fitted parameter is renamed as a prediction, and no self-citation carries a load-bearing argument, so the appropriate circularity score is 0.

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

No numerical parameters are fitted in this systems paper. The three axioms are domain assumptions about expressiveness of the adapter contract, uniformity of integration cost, and adequacy of the transport layer. No new physical or theoretical entities are introduced.

assumptions (3)
  • domain assumption Robot policies can be faithfully expressed through the four-operation adapter contract (update_obs, get_action, reset, batched variants) without loss of generality.
    The whole standard rests on this expressiveness premise; Sections III-B and III-D specify the operations and schema, and no proof is offered that every policy family fits.
  • domain assumption The O(N+M) cost reduction assumes uniform per-policy adapter cost and per-environment client cost, with no hidden couplings between specific policy-environment pairs.
    Section I frames the problem as O(NM) and Section III-A as O(N+M); this is the standard adapter-bus argument and ignores exotic dependencies or per-pair quirks.
  • domain assumption WebSocket/MessagePack transport with retries, keepalive, and instance-identifier checks is reliable and fast enough for robot policy serving, including real-time control.
    Section III-C defines the protocol but provides no latency or reliability benchmarks to validate this premise.

how reviews work

0 comments
Cite this review

Pith. "Pith review of XPolicyLab: A Unified Standard and Open Ecosystem for Robot Policy Evaluation and Deployment." pith.science (2026). https://pith.science/paper/DTYCNYRU

@misc{pith2026260809892,
  author       = {Pith},
  title        = {Pith review of: XPolicyLab: A Unified Standard and Open Ecosystem for Robot Policy Evaluation and Deployment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DTYCNYRU}},
  note         = {Machine review of arXiv:2608.09892}
}
read the original abstract

Robot policy evaluation and deployment remain fragmented by model-specific software dependencies, data representations, and runtime interfaces, so that connecting N policies to M evaluation environments requires O(NM) separate integrations. We present XPolicyLab, a unified standard and open ecosystem that reduces this cost to O(N+M). XPolicyLab specifies common observation, action, and trajectory schemas together with a minimal adapter interface for observation updates, action prediction, batched execution, and episode reset, while a dependency-isolated client/server architecture separates policy inference from environment execution, so that each side retains its native software stack and may run locally or remotely. The ecosystem integrates 42 robot policies and standardizes their installation, debugging, serving, and evaluation workflows. Across these adapters, model-specific code varies by an order of magnitude while the environment-facing loop stays within a few lines of a fixed reference, confirming that the contract confines heterogeneity to the policy side. In a controlled study, conforming to the standard reduces the integration effort of a representative policy from over five hours to two hours, and packaged agent skills reduce it further to thirty minutes. The same adapters serve RoboTwin, RoboDojo simulation, and standardized real-robot evaluation through one interface. XPolicyLab is released as shared infrastructure for reproducible policy comparison and standardized deployment across simulation and physical platforms. Project website: https://xpolicylab.github.io/.

Figures

Figures reproduced from arXiv: 2608.09892 by the authors.

Figure 1
Figure 1. Overview of XPolicyLab. Heterogeneous policy runtimes (left) keep their native dependencies, checkpoints, and processing pipelines, while evaluation environments (right) keep their simulator and robot stacks. XPolicyLab (center) joins the two through a unified lifecycle from installation to evaluation, a minimal adapter contract, standardized observation and action schemas, and a dependency-isolated serving bridge. … view at source ↗
Figure 1
Figure 1. Overview of XPolicyLab. A standard and infrastructure for robot policy evaluation and deployment. Abstract— Robot policy evaluation and deployment remain fragmented by model-specific software dependencies, data representations, and runtime interfaces, so that connecting N policies to M evaluation environments requires O(NM) separate integrations. We present XPolicyLab, a unified standard and open ecosystem that redu… view at source ↗
Figure 2
Figure 2. Cross-platform policy evaluation through XPolicyLab. A shared policy codebase and a standardized serving interface allow the same policy integration to be evaluated in RoboTwin 2.0, RoboDojo simulation, and RoboDojo real-world settings with minimal policy-side adaptation. embodiments, and evaluation protocols, yet share the same policy abstraction and serving architecture. The leaderboards therefore characterize the… view at source ↗
Figures from the paper (2 more)
Figure 2
Figure 2. Figure 2: Infrastructure of XPolicyLab. Heterogeneous policy runtimes (left) keep their native dependencies, checkpoints, and processing pipelines, while evaluation environments (right) keep their simulator and robot stacks. XPolicyLab (center) joins the two through a unified li…
Figure 3
Figure 3. Figure 3: Cross-platform policy evaluation through XPolicy￾Lab. A shared policy codebase and a standardized serving interface allow the same policy integration to be evaluated in RoboTwin 2.0, RoboDojo simulation, and RoboDojo real￾world settings with minimal policy-side adaptat…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 38 linked inside Pith

  1. [15]

    Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation,

    T. Chen, Z. Chen, B. Chen, Z. Cai, Y . Liu, Z. Li, Q. Liang, X. Lin, Y . Ge, Z. Guet al., “Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation,”arXiv preprint arXiv:2506.18088, 2025

    arXiv 2025

  2. [16]

    RoboDojo: A unified sim-and-real benchmark for comprehensive evaluation of generalist robot manipulation policies,

    T. Chen, Y . Chen, Z. Li, J. Tang, K. Su, W. Wan, B. Chen, H. Lu, H. Yan, H. Suet al., “RoboDojo: A unified sim-and-real benchmark for comprehensive evaluation of generalist robot manipulation policies,” arXiv preprint arXiv:2607.04434, 2026

    arXiv 2026

  3. [1]

    Openvla: An open- source vision-language-action model,

    M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. Foster, G. Lam, P. Sanketiet al., “Openvla: An open- source vision-language-action model,”arXiv preprint arXiv:2406.09246, 2024

    arXiv 2024

  4. [2]

    pi0.5: A vision-language-action model with open-world generalization,

    P. Intelligence, K. Black, N. Brown, J. Darpinian, K. Dhabalia, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusaiet al., “ pi0.5: A vision-language-action model with open-world generalization,” arXiv preprint arXiv:2504.16054, 2025

    arXiv 2025

  5. [3]

    Pi 0.7: a steerable generalist robotic foundation model with emergent capabilities,

    P. Intelligence, B. Ai, A. Amin, R. Aniceto, A. Balakrishna, G. Balke, K. Black, G. Bokinsky, S. Cao, T. Charbonnieret al., “Pi 0.7: a steerable generalist robotic foundation model with emergent capabilities,”arXiv preprint arXiv:2604.15483, 2026

    arXiv 2026

  6. [4]

    Octo: An open-source generalist robot policy,

    O. M. Team, D. Ghosh, H. Walke, K. Pertsch, K. Black, O. Mees, S. Dasari, J. Hejna, T. Kreiman, C. Xuet al., “Octo: An open-source generalist robot policy,”arXiv preprint arXiv:2405.12213, 2024

    arXiv 2024

  7. [5]

    World action models are zero-shot policies,

    S. Ye, Y . Ge, K. Zheng, S. Gao, S. Yu, G. Kurian, S. Indupuru, Y . L. Tan, C. Zhu, J. Xianget al., “World action models are zero-shot policies,”arXiv preprint arXiv:2602.15922, 2026

    arXiv 2026

  8. [6]

    Xiaomi-robotics-0: An open-sourced vision-language-action model with real-time execution,

    R. Cai, J. Guo, X. He, P. Jin, J. Li, B. Lin, F. Liu, W. Liu, F. Ma, K. Ma et al., “Xiaomi-robotics-0: An open-sourced vision-language-action model with real-time execution,”arXiv preprint arXiv:2602.12684, 2026

    arXiv 2026

Show all 61 references
  1. [7]

    Hy-embodied-0.5-vla: From vision- language-action models to a real-world robot learning stack,

    H. Zhang, L. Xiang, H. Lin, Z. Huang, M. Wang, D. Zhong, Y . Dong, Y . Wu, Y . Rao, D. Zhanget al., “Hy-embodied-0.5-vla: From vision- language-action models to a real-world robot learning stack,”arXiv preprint arXiv:2606.14409, 2026

    2026 arXiv

  2. [8]

    G3flow: Generative 3d semantic flow for pose-aware and generalizable object manipulation,

    T. Chen, Y . Mu, Z. Liang, Z. Chen, S. Peng, Q. Chen, M. Xu, R. Hu, H. Zhang, X. Liet al., “G3flow: Generative 3d semantic flow for pose-aware and generalizable object manipulation,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 1735–1744

    2025

  3. [9]

    Towards trustworthy embodied intelligence: A systems framework and graded trustworthiness levels,

    X. Yang, T. Chen, H. Su, M. Wang, C. Yu, Z. Tu, Y . Chen, Y . Huo, L. Zhang, Y . Huanget al., “Towards trustworthy embodied intelligence: A systems framework and graded trustworthiness levels,”arXiv preprint arXiv:2607.26121, 2026

    2026 arXiv

  4. [10]

    Lerobot: An open-source library for end-to-end robot learning,

    R. Cadene, S. Alibert, F. Capuano, M. Aractingi, A. Zouitine, P. Kooijmans, J. Choghari, M. Russi, C. Pascal, S. Palmaet al., “Lerobot: An open-source library for end-to-end robot learning,” in International Conference on Learning Representations, vol. 2026, 2026, pp. 122 398–122 417

    2026

  5. [11]

    Fastumi: A scalable and hardware- independent universal manipulation interface with dataset,

    K. Liu, C. Guan, Z. Jia, Z. Wu, X. Liu, T. Wang, S. Liang, P. Chen, P. Zhang, H. Songet al., “Fastumi: A scalable and hardware- independent universal manipulation interface with dataset,”arXiv preprint arXiv:2409.19499, 2024

    2024 arXiv

  6. [12]

    Droid: A large-scale in-the-wild robot manipulation dataset,

    A. Khazatsky, K. Pertsch, S. Nair, A. Balakrishna, S. Dasari, S. Karam- cheti, S. Nasiriany, M. K. Srirama, L. Y . Chen, K. Elliset al., “Droid: A large-scale in-the-wild robot manipulation dataset,”arXiv preprint arXiv:2403.12945, 2024

    2024 arXiv

  7. [13]

    Starvla: A lego-like codebase for vision-language- action model developing,

    S. Community, “Starvla: A lego-like codebase for vision-language- action model developing,”arXiv preprint arXiv:2604.05014, 2026

    2026 arXiv

  8. [14]

    {RLinf}: Flexible and efficient {Large-Scale} reinforcement learning via {Macro-to-Micro} flow transformation,

    C. Yu, Y . Wang, Z. Guo, H. Lin, S. Xu, H. Zang, Q. Zhang, Y . Wu, C. Zhu, J. Huet al., “ {RLinf}: Flexible and efficient {Large-Scale} reinforcement learning via {Macro-to-Micro} flow transformation,” in20th USENIX Symposium on Operating Systems Design and Implementation (OSD...

    2026

  9. [17]

    Bridge data: Boosting gener- alization of robotic skills with cross-domain datasets,

    F. Ebert, Y . Yang, K. Schmeckpeper, B. Bucher, G. Georgakis, K. Daniilidis, C. Finn, and S. Levine, “Bridge data: Boosting gener- alization of robotic skills with cross-domain datasets,”arXiv preprint arXiv:2109.13396, 2021

    2021 arXiv

  10. [18]

    Rlbench: The robot learning benchmark & learning environment,

    S. James, Z. Ma, D. R. Arrojo, and A. J. Davison, “Rlbench: The robot learning benchmark & learning environment,” 2019. [Online]. Available: https://arxiv.org/abs/1909.12271

    2019 arXiv

  11. [19]

    Maniskill2: A unified benchmark for generalizable manipulation skills,

    J. Gu, F. Xiang, X. Li, Z. Ling, X. Liu, T. Mu, Y . Tang, S. Tao, X. Wei, Y . Yaoet al., “Maniskill2: A unified benchmark for generalizable manipulation skills,” inThe Eleventh International Conference on Learning Representations, 2023

    2023

  12. [20]

    Libero: Benchmarking knowledge transfer for lifelong robot learning,

    B. Liu, Y . Zhu, C. Gao, Y . Feng, Q. Liu, Y . Zhu, and P. Stone, “Libero: Benchmarking knowledge transfer for lifelong robot learning,” Advances in Neural Information Processing Systems, vol. 36, pp. 44 776– 44 791, 2023

    2023

  13. [21]

    Rmbench: Memory-dependent robotic manipulation benchmark with insights into policy design,

    T. Chen, Y . Wang, M. Li, Y . Qin, H. Shi, Z. Li, Y . Hu, Y . Zhang, K. Wang, Y . Chenet al., “Rmbench: Memory-dependent robotic manipulation benchmark with insights into policy design,”arXiv preprint arXiv:2603.01229, 2026

    2026 arXiv

  14. [22]

    Univtac: A unified simulation platform for visuo- tactile manipulation data generation, learning, and benchmarking,

    B. Chen, W. Wan, T. Chen, X. Guo, C. Xu, Y . Qi, H. Zhang, L. Wu, T. Xu, Z. Liet al., “Univtac: A unified simulation platform for visuo- tactile manipulation data generation, learning, and benchmarking,” arXiv preprint arXiv:2602.10093, 2026

    2026

  15. [23]

    Robotwin: Dual-arm robot benchmark with generative digital twins,

    Y . Mu, T. Chen, Z. Chen, S. Peng, Z. Lan, Z. Gao, Z. Liang, Q. Yu, Y . Zou, M. Xuet al., “Robotwin: Dual-arm robot benchmark with generative digital twins,” inProceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 27 649–27 660

    2025

  16. [24]

    Benchmarking generalizable bimanual manipulation: Robotwin dual-arm collaboration challenge at cvpr 2025 meis workshop,

    T. Chen, K. Wang, Z. Yang, Y . Zhang, Z. Chen, B. Chen, W. Dong, Z. Liu, D. Chen, T. Yanget al., “Benchmarking generalizable bimanual manipulation: Robotwin dual-arm collaboration challenge at cvpr 2025 meis workshop,”arXiv preprint arXiv:2506.23351, 2025

    2025 arXiv

  17. [25]

    Roboarena: Distributed real-world evaluation of generalist robot policies,

    P. Atreya, K. Pertsch, T. Lee, M. J. Kim, A. Jain, A. Kuramshin, C. Eppner, C. Neary, E. Hu, F. Ramoset al., “Roboarena: Distributed real-world evaluation of generalist robot policies,”arXiv preprint arXiv:2506.18123, 2025

    2025

  18. [26]

    Robochallenge: Large-scale real-robot evaluation of embodied policies,

    A. Yakefu, B. Xie, C. Xu, E. Zhang, E. Zhou, F. Jia, H. Yang, H. Fan, H. Zhang, H. Penget al., “Robochallenge: Large-scale real-robot evaluation of embodied policies,”arXiv preprint arXiv:2510.17950, 2025

    2025

  19. [27]

    pi0: A vision-language-action flow model for general robot control,

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichteret al., “ pi0: A vision-language-action flow model for general robot control,”arXiv preprint arXiv:2410.24164, 2024

    2024 arXiv

  20. [28]

    Gr00t n1: An open foundation model for generalist humanoid robots,

    J. Bjorck, F. Casta˜neda, N. Cherniadev, X. Da, R. Ding, L. Fan, Y . Fang, D. Fox, F. Hu, S. Huanget al., “Gr00t n1: An open foundation model for generalist humanoid robots,”arXiv preprint arXiv:2503.14734, 2025

    2025 arXiv

  21. [29]

    A1: A fully transparent open-source, adaptive and efficient truncated vision-language-action model,

    K. Zhang, J. Zhang, R. Xu, Y . Sun, S. Xue, Y . Wen, X. Guo, M. Guo, W. Liufu, L. Zihouet al., “A1: A fully transparent open-source, adaptive and efficient truncated vision-language-action model,”arXiv preprint arXiv:2604.05672, 2026

    2026 arXiv

  22. [30]

    Aha-wam: Asynchronous horizon-adaptive world-action modeling with observation-guided context routing,

    J. Cai, L. Ling, S. Chu, Z. Liu, J. Kang, Z. Liang, W. Xu, Y . Mao, W. Zhang, X. Yanget al., “Aha-wam: Asynchronous horizon-adaptive world-action modeling with observation-guided context routing,”arXiv preprint arXiv:2606.09811, 2026

    2026 arXiv

  23. [31]

    Abot-m0: Vla foundation model for robotic manipulation with action manifold learning,

    Y . Yang, S. Zeng, T. Lin, X. Chang, D. Qi, J. Xiao, H. Liu, R. Chen, Y . Chen, D. Huoet al., “Abot-m0: Vla foundation model for robotic manipulation with action manifold learning,”arXiv preprint arXiv:2602.11236, 2026

    2026 arXiv

  24. [32]

    Learning fine- grained bimanual manipulation with low-cost hardware,

    T. Z. Zhao, V . Kumar, S. Levine, and C. Finn, “Learning fine- grained bimanual manipulation with low-cost hardware,”arXiv preprint arXiv:2304.13705, 2023

    2023 arXiv

  25. [33]

    Being-h0. 5: Scaling human-centric robot learning for cross-embodiment generalization,

    H. Luo, Y . Wang, W. Zhang, S. Zheng, Z. Xi, C. Xu, H. Xu, H. Yuan, C. Zhang, Y . Wanget al., “Being-h0. 5: Scaling human-centric robot learning for cross-embodiment generalization,”arXiv preprint arXiv:2601.12993, 2026

    2026

  26. [34]

    Dm0: An embodied-native vision-language- action model towards physical ai,

    E. Yu, H. Lv, J. Sun, K. Lin, R. Zhang, Y . Shi, Y . Chen, Z. Chen, Z. Zhang, F. Jiaet al., “Dm0: An embodied-native vision-language- action model towards physical ai,”arXiv preprint arXiv:2602.14974, 2026

    2026

  27. [35]

    Dexora: Open-source vla for high-dof bimanual dexterity,

    Z. Zhang, J. Pang, Z. Yang, K. Li, M. Liao, S. Zhang, G. Chi, J. Guo, H.- a. Gao, M. Shiet al., “Dexora: Open-source vla for high-dof bimanual dexterity,”arXiv preprint arXiv:2605.18722, 2026

    2026 arXiv

  28. [36]

    Diffusion policy: Visuomotor policy learning via action diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via action diffusion,”The International Journal of Robotics Research, vol. 44, no. 10-11, pp. 1684–1704, 2025

    2025

  29. [37]

    Eventvla: Event-driven visual evidence memory for long-horizon vision-language-action policies,

    G. Yang, Z. Tu, Y . Yang, S. Mao, J. Dong, T. Chen, J. Peng, J. Xiong, J. Cao, J. Daiet al., “Eventvla: Event-driven visual evidence memory for long-horizon vision-language-action policies,” arXiv preprint arXiv:2606.20092, 2026

    2026 arXiv

  30. [38]

    Fast-wam: Do world action models need test-time future imagination?

    T. Yuan, Z. Dong, Y . Liu, and H. Zhao, “Fast-wam: Do world action models need test-time future imagination?”arXiv preprint arXiv:2603.16666, 2026

    2026 arXiv

  31. [39]

    Galaxea g0.5 technical report,

    Galaxea Team, “Galaxea g0.5 technical report,” 2026. [Online]. Available: https://opengalaxea.github.io/G05/

    2026

  32. [40]

    Galaxea open-world dataset and g0 dual-system vla model,

    T. Jiang, T. Yuan, Y . Liu, C. Lu, J. Cui, X. Liu, S. Cheng, J. Gao, H. Xu, and H. Zhao, “Galaxea open-world dataset and g0 dual-system vla model,”arXiv preprint arXiv:2509.00576, 2025

    2025 arXiv

  33. [41]

    Gigaworld-policy: An efficient action-centered world–action model,

    A. Ye, B. Wang, C. Ni, G. Huang, G. Zhao, H. Li, H. Li, J. Li, J. Lv, J. Liuet al., “Gigaworld-policy: An efficient action-centered world–action model,”arXiv preprint arXiv:2603.17240, 2026

    2026

  34. [42]

    Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems,

    Q. Bu, J. Cai, L. Chen, X. Cui, Y . Ding, S. Feng, S. Gao, X. He, X. Hu, X. Huanget al., “Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems,”arXiv preprint arXiv:2503.06669, 2025

    2025 arXiv

  35. [43]

    H-rdt: Human manipulation enhanced bimanual robotic manipulation,

    H. Bi, L. Wu, T. Lin, H. Tan, Z. Su, H. Su, and J. Zhu, “H-rdt: Human manipulation enhanced bimanual robotic manipulation,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 40, no. 22, 2026, pp. 18 135–18 143

    2026

  36. [44]

    Internvla-a1: Unifying understanding, generation and action for robotic manipulation,

    J. Cai, Z. Cai, J. Cao, Y . Chen, Z. He, L. Jiang, H. Li, H. Li, Y . Li, Y . Liuet al., “Internvla-a1: Unifying understanding, generation and action for robotic manipulation,”arXiv preprint arXiv:2601.02456, 2026

    2026

  37. [45]

    Internvla-a1. 5: Unifying understanding, latent foresight, and action for compositional generalization,

    H. Ma, J. Cai, X. Xu, H. Li, Y . Yang, Y . Tian, J. Cao, H. Zhu, Z. Qiu, Y . Yanget al., “Internvla-a1. 5: Unifying understanding, latent foresight, and action for compositional generalization,”arXiv preprint arXiv:2607.04988, 2026

    2026 arXiv

  38. [46]

    Lda-1b: Scaling latent dynamics action model via universal embodied data ingestion,

    J. Lyu, K. Liu, X. Zhang, H. Liao, Y . Feng, W. Zhu, T. Shen, J. Chen, J. Zhang, Y . Donget al., “Lda-1b: Scaling latent dynamics action model via universal embodied data ingestion,”arXiv preprint arXiv:2602.12215, 2026

    2026 arXiv

  39. [47]

    Causal world modeling for robot control,

    L. Li, Q. Zhang, Y . Luo, S. Yang, R. Wang, F. Han, M. Yu, Z. Gao, N. Xue, X. Zhu, Y . Shen, and Y . Xu, “Causal world modeling for robot control,” 2026. [Online]. Available: https://arxiv.org/abs/2601.21998

    2026 arXiv

  40. [48]

    A pragmatic vla foundation model,

    W. Wu, F. Lu, Y . Wang, S. Yang, S. Liu, F. Wang, Q. Zhu, H. Sun, Y . Wang, S. Maet al., “A pragmatic vla foundation model,”arXiv preprint arXiv:2601.18692, 2026

    2026 arXiv

  41. [49]

    Molmoact2: Action reasoning models for real-world deployment,

    H. Fang, J. Duan, D. Clay, S. Wang, S. Liu, W. Huang, X. Fan, W.-C. Tsai, S. Chen, Y . R. Wanget al., “Molmoact2: Action reasoning models for real-world deployment,”arXiv preprint arXiv:2605.02881, 2026

    2026 arXiv

  42. [50]

    Starvla- α: Reducing complexity in vision-language- action systems,

    J. Ye, N. Gao, S. Yang, J. Zheng, Z. Wang, Y . Chen, P. Chen, Y . Chen, S. Liu, and J. Jia, “Starvla- α: Reducing complexity in vision-language- action systems,”arXiv preprint arXiv:2604.11757, 2026

    2026 arXiv

  43. [51]

    Rdt-1b: a diffusion foundation model for bimanual manipula- tion,

    S. Liu, L. Wu, B. Li, H. Tan, H. Chen, Z. Wang, K. Xu, H. Su, and J. Zhu, “Rdt-1b: a diffusion foundation model for bimanual manipula- tion,” inInternational Conference on Learning Representations, vol. 2025, 2025, pp. 29 982–30 009

    2025

  44. [52]

    Rise: Self-improving robot policy with compositional world model,

    J. Yang, K. Lin, J. Li, W. Zhang, T. Lin, L. Wu, Z. Su, H. Zhao, Y .-Q. Zhang, L. Chenet al., “Rise: Self-improving robot policy with compositional world model,”arXiv preprint arXiv:2602.11075, 2026

    2026 arXiv

  45. [53]

    Smolvla: A vision-language-action model for affordable and efficient robotics,

    M. Shukor, D. Aubakirova, F. Capuano, P. Kooijmans, S. Palma, A. Zouitine, M. Aractingi, C. Pascal, M. Russi, A. Marafiotiet al., “Smolvla: A vision-language-action model for affordable and efficient robotics,”arXiv preprint arXiv:2506.01844, 2025

    2025 arXiv

  46. [54]

    Spatial forcing: Implicit spatial representation alignment for vision-language-action model,

    F. Li, W. Song, H. Zhao, J. Wang, P. Ding, D. Wang, L. Zeng, and H. Li, “Spatial forcing: Implicit spatial representation alignment for vision-language-action model,” inInternational Conference on Learning Representations, vol. 2026, 2026, pp. 132 324–132 345

    2026

  47. [55]

    Spirit-v1.5: Clean data is the enemy of great robot foundation models,

    S. A. Team, “Spirit-v1.5: Clean data is the enemy of great robot foundation models,”Spirit AI Blog, 2026, https://www.spirit- ai.com/en/blog/spirit-v1-5

    2026

  48. [56]

    Tinyvla: Towards fast, data-efficient vision- language-action models for robotic manipulation,

    J. Wen, Y . Zhu, J. Li, M. Zhu, Z. Tang, K. Wu, Z. Xu, N. Liu, R. Cheng, C. Shenet al., “Tinyvla: Towards fast, data-efficient vision- language-action models for robotic manipulation,”IEEE Robotics and Automation Letters, 2025

    2025

  49. [57]

    X-vla: Soft-prompted transformer as scalable cross-embodiment vision-language-action model,

    J. Zheng, J. Li, Z. Wang, D. Liu, X. Kang, Y . Feng, Y . Zheng, J. Zou, Y . Chen, J. Zenget al., “X-vla: Soft-prompted transformer as scalable cross-embodiment vision-language-action model,” inInternational Conference on Learning Representations, vol. 2026, 2026, pp. 60 580– 60 606

    2026

  50. [58]

    Unified 4d world action modeling from video priors with asynchronous denoising,

    J. Guo, Q. Li, P. Li, Z. Chen, N. Sun, Y . Su, H. Wang, Y . Zhang, X. Li, and H. Liu, “Unified 4d world action modeling from video priors with asynchronous denoising,”arXiv preprint arXiv:2604.26694, 2026

    2026 arXiv

  51. [59]

    Xiaomi-robotics-1: Scaling vision-language- action models with over 100k hours of real-world trajectories,

    X. R. Team, J. Guo, P. Jin, J. Li, P. Li, Y . Li, F. Liu, W. Peng, O. Qin, Y . Suet al., “Xiaomi-robotics-1: Scaling vision-language- action models with over 100k hours of real-world trajectories,”arXiv preprint arXiv:2607.15330, 2026

    2026 arXiv

  52. [60]

    Unit: Toward a unified physical language for human-to-humanoid policy learning and world modeling,

    T. U. o. H. K. XPENG Robotics, Tsinghua University, “Unit: Toward a unified physical language for human-to-humanoid policy learning and world modeling,”https://xpeng-robotics.github.io/unit/, 2026

    2026

  53. [61]

    Rxbrain: Embodied cognition foundation model with joint language-visual reasoning and imagination,

    H. Liang, M. Chen, Y . Huang, Y . Guo, X. Zhu, X. Shi, K. Wang, Y . Mao, W. Zhou, L. Chenet al., “Rxbrain: Embodied cognition foundation model with joint language-visual reasoning and imagination,”arXiv preprint arXiv:2607.14187, 2026

    2026 arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.