pith. machine review for the scientific record. sign in

arxiv: 2604.15671 · v1 · submitted 2026-04-17 · 💻 cs.RO

Recognition: unknown

Long-Term Memory for VLA-based Agents in Open-World Task Execution

Hua Chen, Jiabao Zhao, Weixin Mao, Xu Huang, Yinhao Li

Pith reviewed 2026-05-10 09:20 UTC · model grok-4.3

classification 💻 cs.RO
keywords VLAlong-term memorychemical automationdual-layer memoryembodied agentsroboticstask planning
0
0 comments X

The pith

ChemBot adds dual-layer memory to VLA agents for persistent learning in long chemical experiments.

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

The paper proposes ChemBot as a solution to the short-term memory problem in Vision-Language-Action models when applied to complex chemical laboratory tasks. Standard VLA approaches treat each planning and execution step independently, leading to repeated mistakes in multi-stage protocols. ChemBot introduces a dual-layer memory system that stores and retrieves successful trajectories, paired with a Model Context Protocol server to manage tools and sub-agents. It also uses future-state asynchronous inference to keep execution smooth. Experiments show this leads to safer and more successful robot performance in real lab settings.

Core claim

ChemBot is a dual-layer, closed-loop framework integrating an autonomous AI agent with a progress-aware VLA model (Skill-VLA) for hierarchical task decomposition and execution. It uses a dual-layer memory architecture to consolidate successful trajectories into retrievable assets and a Model Context Protocol (MCP) server for sub-agent and tool orchestration. A future-state-based asynchronous inference mechanism mitigates trajectory discontinuities, resulting in superior operational safety, precision, and task success rates over VLA baselines in long-horizon chemical experimentation.

What carries the argument

The dual-layer memory architecture, which consolidates successful trajectories into retrievable assets for reuse by the VLA agent.

If this is right

  • Reduces reliance on inefficient trial-and-error by reusing proven strategies in multi-stage protocols.
  • Achieves higher task success rates through accumulated experience in long-horizon tasks.
  • Enhances operational safety and precision in chemical lab automation on collaborative robots.
  • Enables hierarchical decomposition and smooth execution via Skill-VLA and asynchronous inference.

Where Pith is reading between the lines

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

  • Similar memory consolidation techniques could apply to other embodied domains like household robotics or manufacturing where long sequences of actions are needed.
  • The approach highlights the value of closing the loop between execution outcomes and future planning in VLA systems.
  • Further work might explore how memory retrieval scales with the number of stored trajectories without increasing latency.

Load-bearing premise

The dual-layer memory architecture and Model Context Protocol server can reliably consolidate, retrieve, and apply successful trajectories without introducing new failure modes or excessive overhead in real chemical lab settings.

What would settle it

A repeated trial of the same long-horizon chemical experiment protocol using both ChemBot and baseline VLA agents, checking if ChemBot's success rate stays consistently higher while its memory retrieval does not cause new errors or slowdowns.

Figures

Figures reproduced from arXiv: 2604.15671 by Hua Chen, Jiabao Zhao, Weixin Mao, Xu Huang, Yinhao Li.

Figure 1
Figure 1. Figure 1: Paradigm Shift from Traditional Manual to Agent-Driven Experi [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of ChemBot, a closed-loop framework for chemical task decomposition and action execution, centered on an AI agent integrating chain-backtracking subtask generation and dual-layer memory mechanisms, with a progress-aware Vision-Language-Action Model for low-level execution. • Skill-VLA Module: An augmented VLA architecture featuring a progress-prediction head and a continuous inference pipeline… view at source ↗
Figure 3
Figure 3. Figure 3: Sequence diagram of iterative task decomposition via multi￾agent collaboration, illustrating a closed-loop workflow where an agent incrementally generates, validates, and refines subtasks through tool invo￾cation and dual-layer memory retrieval until sequence completion. Overall Task: To purify crude salt by removing insoluble impurities via dissolution, filtration, and evaporation, and to test for residua… view at source ↗
Figure 4
Figure 4. Figure 4: Structured scene representation generated by the Scene Describer. The module leverages Set-of-Mark prompting to parse raw lab images into a structured memory dashboard, detailing task-related items, their interactable states, affordances, and spatial relationships to provide visual grounding for subsequent task planning. Subtask Generator. The Subtask Reasoner and Reflector constitute an incremental planni… view at source ↗
Figure 5
Figure 5. Figure 5: Detailed network architecture of Skill-VLA. (a) The main VLA model for continuous action generation and subtask progress prediction. The state of the robotic arm and the hidden state of the VLM are provided as inputs to the Action Head and the Progress Head, respectively. (b) Internal structure of the Progress Head, an auxiliary module for the estimation of task progress. iteratively denoising pure noise v… view at source ↗
Figure 6
Figure 6. Figure 6: Heating to Dissolve Solid Dissolution Water Bath Heating Acid-base Neutralization Centrifuge Tube Transfer Reagent Combination Phenolphthalein Solid Weighing Water Bath Cooling Test Tube Heating CuSO + Na CO Test Tube Shaking 0 1000 2000 3000 4000 5000 6000 Number of Sub-task Segments Place Grasp Transfer / Pour Other Stir Shake Ignite / Heat Push [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: A visualization of the predicted task progress for opening a wide [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Bar chart of success rates for three chemistry tasks (error bars show 68% Wilson score intervals). Skill-VLA with subtask-level training outperforms the full-trajectory baselines (π0.5, GR00T) across all tasks, validating its advantage in long-horizon tasks. Results. The results in [PITH_FULL_IMAGE:figures/full_fig_p006_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of motion trajectories for Joint 4 of the UR robot between synchronous and asynchronous inference. Asynchronous inference results in smoother trajectories and reduced execution time. fully autonomous scientific assistants capable of end-to-end experiment orchestration and analysis. REFERENCES [1] R. Jiang, W. Lin, S. Wen, F. Zhu, T. Luan, and G. Ouyang, “Devel￾opment of a full automation solid p… view at source ↗
read the original abstract

Vision-Language-Action (VLA) models have demonstrated significant potential for embodied decision-making; however, their application in complex chemical laboratory automation remains restricted by limited long-horizon reasoning and the absence of persistent experience accumulation. Existing frameworks typically treat planning and execution as decoupled processes, often failing to consolidate successful strategies, which results in inefficient trial-and-error in multi-stage protocols. In this paper, we propose ChemBot, a dual-layer, closed-loop framework that integrates an autonomous AI agent with a progress-aware VLA model (Skill-VLA) for hierarchical task decomposition and execution. ChemBot utilizes a dual-layer memory architecture to consolidate successful trajectories into retrievable assets, while a Model Context Protocol (MCP) server facilitates efficient sub-agent and tool orchestration. To address the inherent limitations of VLA models, we further implement a future-state-based asynchronous inference mechanism to mitigate trajectory discontinuities. Extensive experiments on collaborative robots demonstrate that ChemBot achieves superior operational safety, precision, and task success rates compared to existing VLA baselines in complex, long-horizon chemical experimentation.

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 / 1 minor

Summary. The paper introduces ChemBot, a dual-layer closed-loop framework that integrates an autonomous AI agent with a progress-aware VLA model (Skill-VLA) for hierarchical task decomposition and execution in chemical laboratory automation. It employs a dual-layer memory architecture to consolidate and retrieve successful trajectories, a Model Context Protocol (MCP) server for sub-agent and tool orchestration, and a future-state-based asynchronous inference mechanism to mitigate trajectory discontinuities. The central claim is that extensive experiments on collaborative robots demonstrate superior operational safety, precision, and task success rates relative to existing VLA baselines in complex, long-horizon chemical tasks.

Significance. If the experimental results can be substantiated with quantitative metrics, ablations, and controls, the dual-layer memory and closed-loop design could offer a practical mechanism for persistent experience accumulation in embodied VLA systems, addressing a recognized limitation in long-horizon reasoning for specialized real-world domains.

major comments (1)
  1. [Abstract] Abstract: the superiority claim in safety, precision, and success rates is presented without any supporting metrics, baseline definitions, experiment counts, error bars, ablation results isolating the memory layer or MCP server, or description of trajectory consolidation/retrieval procedures. This absence prevents verification that the proposed architecture, rather than Skill-VLA or hardware factors, produced the reported gains.
minor comments (1)
  1. The manuscript title emphasizes open-world task execution while the abstract and framework description focus exclusively on chemical laboratory protocols; clarify the intended scope and any generalization beyond the chemical domain.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment below and will revise the abstract to improve verifiability while preserving its concise nature.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the superiority claim in safety, precision, and success rates is presented without any supporting metrics, baseline definitions, experiment counts, error bars, ablation results isolating the memory layer or MCP server, or description of trajectory consolidation/retrieval procedures. This absence prevents verification that the proposed architecture, rather than Skill-VLA or hardware factors, produced the reported gains.

    Authors: We agree that the abstract would benefit from greater specificity to allow immediate assessment of the claims. The full manuscript already contains the requested elements in Section 4 (Experiments), including quantitative success rates with error bars across 50+ trials per task, explicit baseline definitions (e.g., standard VLA models without memory), ablation studies isolating the dual-layer memory and MCP server contributions, and detailed descriptions of trajectory consolidation/retrieval in Section 3.2. To directly address the concern, we will revise the abstract to incorporate key summary metrics (e.g., relative success rate improvements and experiment scale) and a brief note on the ablations, ensuring readers can attribute gains to the proposed components rather than Skill-VLA or hardware alone. revision: yes

Circularity Check

0 steps flagged

No circularity: framework description and experimental claims are independent of self-referential definitions or fitted inputs

full rationale

The paper introduces ChemBot as a dual-layer memory architecture plus MCP server and asynchronous inference for VLA agents, then asserts empirical superiority from experiments on collaborative robots. No equations, parameter fits, predictions derived from inputs, self-citations as load-bearing premises, or ansatzes appear in the provided text. The claimed gains in safety/precision/success rates are presented as outcomes of external experiments rather than quantities defined by the architecture itself, satisfying the condition for a self-contained proposal against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 4 invented entities

Only the abstract is available, so the ledger is limited to components explicitly named; no free parameters or external benchmarks are described.

axioms (1)
  • domain assumption VLA models can be extended with external memory layers and orchestration servers to overcome long-horizon limitations.
    Invoked by the proposal of ChemBot and Skill-VLA without further justification in the abstract.
invented entities (4)
  • ChemBot dual-layer closed-loop framework no independent evidence
    purpose: Integrate autonomous AI agent with progress-aware VLA for hierarchical decomposition and execution
    New system proposed in the paper; no independent evidence supplied.
  • Skill-VLA no independent evidence
    purpose: Progress-aware VLA model for task execution
    New model variant introduced; no independent evidence supplied.
  • dual-layer memory architecture no independent evidence
    purpose: Consolidate successful trajectories into retrievable assets
    New architecture proposed; no independent evidence supplied.
  • Model Context Protocol (MCP) server no independent evidence
    purpose: Facilitate sub-agent and tool orchestration
    New server component proposed; no independent evidence supplied.

pith-pipeline@v0.9.0 · 5489 in / 1537 out tokens · 46991 ms · 2026-05-10T09:20:08.586842+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

44 extracted references · 19 canonical work pages · 10 internal anchors

  1. [1]

    Devel- opment of a full automation solid phase microextraction method for investigating the partition coefficient of organic pollutant in complex sample,

    R. Jiang, W. Lin, S. Wen, F. Zhu, T. Luan, and G. Ouyang, “Devel- opment of a full automation solid phase microextraction method for investigating the partition coefficient of organic pollutant in complex sample,”Journal of Chromatography A, vol. 1406, pp. 27–33, 2015

    2015

  2. [2]

    Organic synthesis in a modular robotic system driven by a chemical programming language,

    S. Steiner, J. Wolf, S. Glatzel, A. Andreou, J. M. Granda, G. Keenan, T. Hinkley, G. Aragon-Camarasa, P. J. Kitson, D. Angeloneet al., “Organic synthesis in a modular robotic system driven by a chemical programming language,”Science, vol. 363, no. 6423, p. eaav2211, 2019

    2019

  3. [3]

    A mobile robotic chemist,

    B. Burger, P. M. Maffettone, V . V . Gusev, C. M. Aitchison, Y . Bai, X. Wang, X. Li, B. M. Alston, B. Li, R. Cloweset al., “A mobile robotic chemist,”Nature, vol. 583, no. 7815, pp. 237–241, 2020

    2020

  4. [4]

    An all-round ai-chemist with a scientific mind,

    Q. Zhu, F. Zhang, Y . Huang, H. Xiao, L. Zhao, X. Zhang, T. Song, X. Tang, X. Li, G. Heet al., “An all-round ai-chemist with a scientific mind,”National Science Review, vol. 9, no. 10, p. nwac190, 2022

    2022

  5. [5]

    Gemini Robotics: Bringing AI into the Physical World

    G. R. Team, S. Abeyruwan, J. Ainslie, J.-B. Alayrac, M. G. Arenas, T. Armstrong, A. Balakrishna, R. Baruch, M. Bauza, M. Blokzijlet al., “Gemini robotics: Bringing ai into the physical world,”arXiv preprint arXiv:2503.20020, 2025

    work page internal anchor Pith review arXiv 2025

  6. [6]

    GPT-4 Technical Report

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkatet al., “Gpt-4 technical report,”arXiv preprint arXiv:2303.08774, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  7. [7]

    π 0: 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. 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,” 2024

    2024

  8. [8]

    GR00T N1: An open foundation model for generalist humanoid robots,

    NVIDIA, J. Bjorck, N. C. Fernando Casta ˜neda, 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. Yu, ...

    2025

  9. [9]

    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

    work page internal anchor Pith review arXiv 2025

  10. [10]

    Voyager: An Open-Ended Embodied Agent with Large Language Models

    G. Wang, Y . Xie, Y . Jiang, A. Mandlekar, C. Xiao, Y . Zhu, L. Fan, and A. Anandkumar, “V oyager: An open-ended embodied agent with large language models,”arXiv preprint arXiv:2305.16291, 2023

    work page internal anchor Pith review arXiv 2023

  11. [11]

    Organa: A robotic assistant for automated chemistry experimentation and characterization,

    K. Darvish, M. Skreta, Y . Zhao, N. Yoshikawa, S. Som, M. Bog- danovic, Y . Cao, H. Hao, H. Xu, A. Aspuru-Guziket al., “Organa: A robotic assistant for automated chemistry experimentation and characterization,”Matter, vol. 8, no. 2, 2025

    2025

  12. [12]

    A multiagent-driven robotic ai chemist enabling autonomous chemical research on demand,

    T. Song, M. Luo, X. Zhang, L. Chen, Y . Huang, J. Cao, Q. Zhu, D. Liu, B. Zhang, G. Zouet al., “A multiagent-driven robotic ai chemist enabling autonomous chemical research on demand,”Journal of the American Chemical Society, vol. 147, no. 15, pp. 12 534–12 545, 2025

    2025

  13. [13]

    DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

    D. Guo, D. Yang, H. Zhang, J. Song, P. Wang, Q. Zhu, R. Xu, R. Zhang, S. Ma, X. Biet al., “Deepseek-r1: Incentivizing reason- ing capability in llms via reinforcement learning,”arXiv preprint arXiv:2501.12948, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  14. [14]

    Robomatrix: A skill-centric hierarchical framework for scalable robot task planning and execution in open- world,

    W. Mao, W. Zhong, Z. Jiang, D. Fang, Z. Zhang, Z. Lan, H. Li, F. Jia, T. Wang, H. Fanet al., “Robomatrix: A skill-centric hierarchical framework for scalable robot task planning and execution in open- world,”arXiv preprint arXiv:2412.00171, 2024

    work page arXiv 2024

  15. [15]

    Robochemist: Long-horizon and safety- compliant robotic chemical experimentation

    Z. Zhang, C. Yue, H. Xu, M. Liao, X. Qi, H. Gao, Z. Wang, and H. Zhao, “Robochemist: Long-horizon and safety-compliant robotic chemical experimentation,”CoRR, vol. abs/2509.08820, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2509.08820

    work page doi:10.48550/arxiv.2509.08820 2025

  16. [16]

    Openclaw: Your own personal AI assistant,

    P. Steinberger and OpenClaw Contributors, “Openclaw: Your own personal AI assistant,” GitHub repository, 2026, version 2026.3.1. Accessed: 2026-03-02. [Online]. Available: https://github.com/openclaw/openclaw

    2026

  17. [17]

    Towards general computer control: A multimodal agent for red dead redemption ii as a case study

    W. Tan, W. Zhang, X. Xu, H. Xia, Z. Ding, B. Li, B. Zhou, J. Yue, J. Jiang, Y . Liet al., “Cradle: Empowering foundation agents towards general computer control,”arXiv preprint arXiv:2403.03186, 2024

    work page arXiv 2024

  18. [18]

    Zhang, H

    C. Zhang, H. Huang, C. Ni, J. Mu, S. Qin, S. He, L. Wang, F. Yang, P. Zhao, C. Duet al., “Ufo2: The desktop agentos,”arXiv preprint arXiv:2504.14603, 2025

    work page arXiv 2025

  19. [19]

    Anygrasp: Robust and efficient grasp perception in spatial and temporal domains,

    H.-S. Fang, C. Wang, H. Fang, M. Gou, J. Liu, H. Yan, W. Liu, Y . Xie, and C. Lu, “Anygrasp: Robust and efficient grasp perception in spatial and temporal domains,”IEEE Transactions on Robotics, vol. 39, no. 5, pp. 3929–3945, 2023

    2023

  20. [20]

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

    A. Brohan, Y . Chebotar, C. Finn, K. Hausman, A. Herzog, D. Ho, J. Ibarz, A. Irpan, E. Jang, R. Julianet al., “Do as i can, not as i say: Grounding language in robotic affordances,” inConference on robot learning. PMLR, 2023, pp. 287–318

    2023

  21. [21]

    Language agent tree search unifies reasoning acting and planning in language models,

    A. Zhou, K. Yan, M. Shlapentokh-Rothman, H. Wang, and Y .-X. Wang, “Language agent tree search unifies reasoning acting and planning in language models,” 2023

    2023

  22. [22]

    Ghost in the minecraft: Generally capable agents for open-world enviroments via large language mod- els with text-based knowledge and memory

    X. Zhu, Y . Chen, H. Tian, C. Tao, W. Su, C. Yang, G. Huang, B. Li, L. Lu, X. Wanget al., “Ghost in the minecraft: Generally capable agents for open-world environments via large language models with text-based knowledge and memory,”arXiv preprint arXiv:2305.17144, 2023

    work page arXiv 2023

  23. [23]

    Reasoning with language model is planning with world model,

    S. Hao, Y . Gu, H. Ma, J. Hong, Z. Wang, D. Wang, and Z. Hu, “Reasoning with language model is planning with world model,” in Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 8154–8173

    2023

  24. [24]

    Creative robot tool use with large language models,

    M. Xu, P. Huang, W. Yu, S. Liu, X. Zhang, Y . Niu, T. Zhang, F. Xia, J. Tan, and D. Zhao, “Creative robot tool use with large language models,”arXiv preprint arXiv:2310.13065, 2023

    work page arXiv 2023

  25. [25]

    Take a step back: Evoking reasoning via abstraction in large language models,

    H. S. Zheng, S. Mishra, X. Chen, H.-T. Cheng, E. H. Chi, Q. V . Le, and D. Zhou, “Take a step back: Evoking reasoning via abstraction in large language models,” inInternational Conference on Learning Representations, B. Kim, Y . Yue, S. Chaudhuri, K. Fragkiadaki, M. Khan, and Y . Sun, Eds., vol. 2024, 2024, pp. 20 279–20 316

    2024

  26. [26]

    Wonderful team: Zero-shot physical task planning with visual llms,

    Z. Wang, R. Shen, and B. Stadie, “Wonderful team: Zero-shot physical task planning with visual llms,”arXiv preprint arXiv:2407.19094, 2024

    work page arXiv 2024

  27. [27]

    Being-0: A humanoid robotic agent with vision-language models and modular skills,

    H. Yuan, Y . Bai, Y . Fu, B. Zhou, Y . Feng, X. Xu, Y . Zhan, B. F. Karlsson, and Z. Lu, “Being-0: A humanoid robotic agent with vision-language models and modular skills,”arXiv preprint arXiv:2503.12533, 2025

    work page arXiv 2025

  28. [28]

    Rt-2: Vision-language-action models transfer web knowledge to robotic control,

    B. Zitkovich, T. Yu, S. Xu, P. Xu, T. Xiao, F. Xia, J. Wu, P. Wohlhart, S. Welker, A. Wahidet al., “Rt-2: Vision-language-action models transfer web knowledge to robotic control,” inConference on Robot Learning. PMLR, 2023, pp. 2165–2183

    2023

  29. [29]

    $\pi_{0.5}$: 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., “π 0.5: a vision-language-action model with open-world generalization,”arXiv preprint arXiv:2504.16054, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  30. [30]

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

    M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. P. Foster, P. R. Sanketi, Q. Vuong, T. Kollar, B. Burchfiel, R. Tedrake, D. Sadigh, S. Levine, P. Liang, and C. Finn, “Openvla: An open-source vision- language-action model,” inProceedings of The 8th Conference on Robot Learning, ser. Proceedings of Machine Learning Re...

    2025

  31. [31]

    Clap: A closed-loop diffusion transformer action foundation model for robotic manipulation,

    M. Li, Y . Dong, Y . Zhou, and C. Yang, “Clap: A closed-loop diffusion transformer action foundation model for robotic manipulation,” in 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2025, pp. 9808–9815

    2025

  32. [32]

    A review of large language models and autonomous agents in chemistry,

    M. C. Ramos, C. J. Collison, and A. D. White, “A review of large language models and autonomous agents in chemistry,”Chemical science, 2025

    2025

  33. [33]

    Autonomous chemical research with large language models,

    D. A. Boiko, R. MacKnight, B. Kline, and G. Gomes, “Autonomous chemical research with large language models,”Nature, vol. 624, no. 7992, pp. 570–578, 2023

    2023

  34. [34]

    Large language models for chemistry robotics,

    N. Yoshikawa, M. Skreta, K. Darvish, S. Arellano-Rubach, Z. Ji, L. Bjørn Kristensen, A. Z. Li, Y . Zhao, H. Xu, A. Kuramshinet al., “Large language models for chemistry robotics,”Autonomous Robots, vol. 47, no. 8, pp. 1057–1086, 2023

    2023

  35. [35]

    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

    work page internal anchor Pith review arXiv 2023

  36. [36]

    Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V

    J. Yang, H. Zhang, F. Li, X. Zou, C. Li, and J. Gao, “Set-of-mark prompting unleashes extraordinary visual grounding in gpt-4v,”arXiv preprint arXiv:2310.11441, 2023

    work page internal anchor Pith review arXiv 2023

  37. [37]

    A universal system for digitization and automatic execution of the chemical synthesis literature,

    S. H. M. Mehr, M. Craven, A. I. Leonov, G. Keenan, and L. Cronin, “A universal system for digitization and automatic execution of the chemical synthesis literature,”Science, vol. 370, no. 6512, pp. 101– 108, 2020

    2020

  38. [38]

    Task bench: A pa- rameterized benchmark for evaluating parallel runtime performance,

    E. Slaughter, W. Wu, Y . Fu, L. Brandenburg, N. Garcia, W. Kautz, E. Marx, K. S. Morris, Q. Cao, G. Bosilcaet al., “Task bench: A pa- rameterized benchmark for evaluating parallel runtime performance,” inSC20: International Conference for High Performance Computing, Networking, Storage and Analysis. IEEE, 2020, pp. 1–15

    2020

  39. [39]

    Rouge: A package for automatic evaluation of summaries,

    C.-Y . Lin, “Rouge: A package for automatic evaluation of summaries,” inText summarization branches out, 2004, pp. 74–81

    2004

  40. [40]

    BERTScore: Evaluating Text Generation with BERT

    T. Zhang, V . Kishore, F. Wu, K. Q. Weinberger, and Y . Artzi, “Bertscore: Evaluating text generation with bert,”arXiv preprint arXiv:1904.09675, 2019

    work page internal anchor Pith review arXiv 1904

  41. [41]

    A normalized levenshtein distance metric,

    L. Yujian and L. Bo, “A normalized levenshtein distance metric,”IEEE transactions on pattern analysis and machine intelligence, vol. 29, no. 6, pp. 1091–1095, 2007

    2007

  42. [42]

    Qwen3 Technical Report

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lvet al., “Qwen3 technical report,”arXiv preprint arXiv:2505.09388, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  43. [43]

    Doubao-1.6 large language model,

    ByteDance, “Doubao-1.6 large language model,” 2025, accessed: 2026-03-04. [Online]. Available: https://www.volcengine.com/product/doubao

    2025

  44. [44]

    Training-time action conditioning for efficient real-time chunking.arXiv preprint arXiv:2512.05964, 2025

    K. Black, A. Z. Ren, M. Equi, and S. Levine, “Training-time ac- tion conditioning for efficient real-time chunking,”arXiv preprint arXiv:2512.05964, 2025

    work page arXiv 2025