Pith. sign in

REVIEW 4 major objections 5 minor 28 references

ProtoAct: Turning Wet-Lab Protocols into Embodied Robotic Actions

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read ProtoAct converts free-form biological protocols into state-aware, embodiment-ready action sequences, and the parsed subtasks support demonstration collection and VLA model training for both simulated and physical robots.

desk verdict A solid systems paper with a genuinely useful benchmark; the RefineChecker input needs disclosure and the real-robot claim needs numbers, but the core parsing contribution deserves peer review. read the letter →

arxiv 2608.01690 v1 pith:ZXXD534X submitted 2026-08-03 cs.RO cs.AI

classification cs.ROcs.AI
keywords wet-labprotocolsprotocolgroundingembodiedroboticexecutionvision-language-actionmodelsretrieval-augmentedgenerationactionschemaBioP2Edatasetlaboratoryautomation
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

ProtoAct claims that free-form biological wet-lab protocols, which routinely omit routine operations and leave parameters context-dependent, can be converted into state-aware, embodiment-ready action sequences without hand-coding every step. The system parses a protocol into natural-language monitoring conditions and subtasks, then grounds them into constrained JSON action functions; manually annotated examples are retrieved for context, a posterior checker revises missing or inconsistent steps, and a schema fixes the action vocabulary. The paper introduces BioP2E, a manually annotated dataset of 22 cell-culture protocols yielding 258 monitoring conditions, 910 subtasks, and 962 grounded action calls, and shows the pipeline works across seven large language models. The parsed subtasks are then used to collect demonstrations and train vision-language-action models, which execute the tasks in simulation and on a physical robot arm. If the claim holds, biologists could hand written protocols to a system that returns robot-executable programs and training data for laboratory automation.

What carries the argument

The load-bearing machinery is the ProtoAct pipeline's three-module design around a common intermediate representation: ProtoRAG retrieves manually annotated protocol fragments to seed context-sensitive parsing; RefineChecker performs posterior checking of the initial monitoring-condition/subtask parse against user-supplied experiment notes and revises four defect types (missing actions, missing parameters, wrong order, wrong granularity); ActSchema maps the refined representation into a constrained JSON action-function sequence using a predefined space of 30 action primitives and 4 monitoring condition types. The key object is the intermediate representation itself—monitoring conditions plus subtask sequences—which preserves procedural dependencies in a form humans can inspect and robots can be trained on, and which makes automatic evaluation (F1, BLEU, SciBERTScore, normalized Levenshtein distance) possible.

What would settle it

Run ProtoAct on held-out protocols whose RefineChecker notes are written by an independent biologist blind to BioP2E's ground-truth, and compare the resulting F1 and Ldn against the paper's reported numbers; if the posterior-checking gain vanishes, the claimed autonomous repair is not robust.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a protocol-grounding pipeline can bridge biological protocol understanding and embodied robotic execution by making explicit what wet-lab instructions leave implicit: the monitoring conditions that gate each step, the physical action primitives, their parameters, and their order. ProtoAct instantiates this as a two-stage representation, first producing a human-inspectable natural-language parse and then constraining it through ActSchema into JSON function sequences, where each function is a named action with fixed parameter fields. The evaluation supports this by reporting strong F1, parameter-similarity, and order-consistency scores across seven LLMs on the 20 held-out BioP2E protocols, and by showing that subtasks parsed from protocols yield successful policy execution in five simulated and three real-robot manipulation tasks.

Load-bearing premise

The load-bearing premise is that the experiment-specific notes fed to RefineChecker can be obtained without already knowing the ground-truth annotations used for scoring; if those notes came from the answer key, the reported posterior-checking gains would partly measure leakage rather than autonomous parsing.

Editorial extensions

If this is right

  • A written cell-culture protocol can produce a robot-executable JSON action program directly, without a human translating each step into machine code.
  • The same subtask representation can be used as language-conditioned instructions for collecting demonstrations, so non-specialists can gather training data by teleoperating the robot through parsed steps.
  • Vision-language-action policies trained with an open-source robot-learning library can execute the parsed subtasks in simulation and on a physical arm, supporting the use of protocol parsing as a data-generation pipeline for embodied models.
  • Because the pipeline is instantiated with seven different LLM backbones with close scores, the protocol grounding is not locked to one model, and future backbone improvements should transfer directly.
  • Ablations attribute distinct roles to the three modules: removing ProtoRAG hurts monitoring-condition recall, removing ActSchema hurts action primitive and parameter regularity, and removing RefineChecker increases omissions and order errors.

Reading between the lines

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

  • The paper's deployment story requires a user to supply experiment-specific notes to RefineChecker; a natural extension is an interactive loop in which the biologist's corrections are the interface, but the paper does not evaluate that interaction directly.
  • Because BioP2E covers only cell-culture protocols from one repository, the 30-primitive action schema may not transfer to other wet-lab domains; testing on PCR or protein-purification protocols would show whether the schema needs extension.
  • The robot validation is at the level of isolated manipulation subtasks, so 'embodiment-ready' should currently be read as subtask-ready; a full multi-hour closed-loop protocol execution remains untested.
  • A decisive test of the posterior-checking benefit would be to source RefineChecker notes from independent biologists blind to the ground-truth annotations; if gains disappear, the measured gain reflects leakage of the answer key rather than autonomous repair.
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 / 5 minor

Summary. The paper presents ProtoAct, a pipeline that converts free-form biological wet-lab protocols into monitoring conditions, subtask sequences, and JSON action-function sequences, using retrieval-augmented parsing (ProtoRAG), posterior checking and revision (RefineChecker), and schema-constrained grounding (ActSchema). The authors introduce BioP2E, a manually annotated dataset of 22 cell-culture protocols with 258 monitoring conditions, 910 subtasks, and 962 action calls, and evaluate the pipeline with seven large language models across recognition, parameter, and order metrics. They also report ablations and use the parsed subtasks to collect demonstrations and train VLA models in simulation and on a physical robot. The central claim is that ProtoAct forms a practical interface between protocol understanding and embodied robotic execution.

Significance. If the claims are substantiated, ProtoAct would be a useful contribution: it offers a new intermediate representation for protocol grounding, a manually annotated benchmark (BioP2E) that the community can reuse, and a systematic comparison of seven LLM backbones with means and standard deviations. The decision to release code and to report simulation results with episode counts are concrete strengths. The main substantive risk is that the RefineChecker improvement reported in Tables 6 and 7 may partly reflect leakage from experiment-specific notes U rather than autonomous parsing ability, and the real-robot validation currently lacks quantitative evidence. These are fixable with additional detail and experiments, but they are load-bearing for the paper's headline claims.

major comments (4)
  1. [Section 3.3, Eq. (3), Appendix B] The input U to RefineChecker is never specified for the 20 evaluation protocols. Appendix B says U includes "key operation flows, experimental constraints, and implicit actions," but the paper does not state who wrote U or whether U was derived from the same manual annotations that define the ground truth (258 monitoring conditions, 910 subtasks). If U was created from those annotations, then the RefineChecker gains in Tables 6 and 7 measure adherence to supplied corrections rather than autonomous recovery of implicit content, and the comparison against the no-RefineChecker ablation is no longer a fair measure of the pipeline's standalone capability. The authors should either specify the exact procedure for obtaining U for each evaluation protocol, show that U is independent of the ground-truth annotations, or rerun the benchmark without U for the claimed autonomous setting.
  2. [Section 4.6, Real-Robot Execution Validation] The real-robot experiment is described only qualitatively: "Both models complete the tasks, with SmolVLA showing more stable execution, consistent with the simulation results." No success rates, number of rollouts, task-wise breakdown, or failure analysis are provided. Since the abstract and conclusion explicitly claim "successful execution in both simulation and real-robot settings," the absence of any quantitative real-robot result makes that part of the central claim unsupported as written. The authors need to report task-level success rates, trial counts, and evaluation protocols for the physical robot experiments, or temper the claim to what the evidence supports.
  3. [Sections 4.2 and 4.5, GPT-5.3 evaluator] GPT-5.3 is both one of the evaluated models in Table 2 and the LLM used as the evaluator in Section 4.5 and Appendix D. Using a model to judge its own outputs creates a self-reference bias, and the paper reports no human-agreement study or calibration for the GPT evaluator scores shown in Appendix E. The authors should either exclude GPT-5.3 from judging outputs that include GPT-5.3's own generations, or provide human correlation results and inter-annotator agreement for the GPT scoring. Without this, the GPT-based conclusions about RefineChecker and ProtoRAG are difficult to interpret.
  4. [Section 4.6 and Appendix F] The embodied validation covers five simulation tasks and three real-robot tasks that are isolated subtasks (tube pickup, lid opening, dish placement, etc.), not complete multi-step protocol executions derived automatically from a full parsed protocol. The claim that "the subtasks generated by ProtoAct can be converted into executable robotic actions" is therefore demonstrated for selected primitive actions, not for the end-to-end pipeline from a raw protocol to a completed experimental procedure. The authors should clarify this scope limitation in the abstract and conclusion, or add an end-to-end execution experiment starting from a parsed protocol and executing a sequence of dependent subtasks.
minor comments (5)
  1. [Section 4.6 vs. Appendix F] Section 4.6 states that the five simulation tasks are constructed in the Pipette platform, while Appendix F says they are constructed in Isaac Sim; the relationship between Pipette and Isaac Sim should be stated explicitly to avoid confusion.
  2. [Tables 6 and 7] The column headers in Tables 6 and 7 are inconsistent: Table 6 repeats "Precision Recall F1" twice for monitor type and monitor parameter, while Table 7 uses the same triple for action primitive and action parameter; please align the headers with the metric names used in the main text.
  3. [Equation (7)] The SciBERTScore formula uses N in the denominator of a sum over i from 1 to N, but N is also used in Equation (6) for the n-gram order; renaming the number of evaluated parameters, for example to M, would remove the notation clash.
  4. [Table 2 caption] The caption says the best and second-best results are marked in bold and underlined, but several cells appear bold and underlined inconsistently; please verify the formatting matches the caption.
  5. [Conclusion, future work] The last sentence of the conclusion lists limitations that are relevant to the main claims; consider moving the RefineChecker dependence on user-provided information into the experimental design section so that readers can evaluate the benchmark validity directly in Section 4.1.

Circularity Check

1 steps flagged · score 4.0 of 10

RefineChecker's U input is defined to include the implicit actions and operation flows the module is credited with recovering; with U's provenance for the 20 test protocols undisclosed, its reported ablation gain is not demonstrated to be autonomous parsing.

  1. self definitional [Section 3.3, Eq. (3); Appendix B; Section 4.4 and Tables 6-7; Section 5]
    "RefineChecker comparesY (0) withU. If the relevant notes have already been covered, the result is kept unchanged. If omissions or errors are identified, the corresponding fragments are reanalyzed... This information includes key operation flows in the protocol, experimental constraints, and implicit actions."

    Eq. (3) defines the refined parse Y(1) as a function of U, and Appendix B defines U as containing 'key operation flows' and 'implicit actions'—exactly the target content RefineChecker is credited with adding in Section 4.4 ('supplement missing content through posterior checking'). For the 20 BioP2E evaluation protocols, the paper never states how U was obtained. If U came from the same manual annotations used as ground truth (258 monitoring conditions, 910 subtasks), the F1/Ldn improvements in Tables 6-7 reflect adherence to provided corrections, not autonomous recovery; the measured RefineChecker gain reduces by construction to copying the supplied answer key. The conclusion admits the dependence but does not resolve benchmark validity.

full rationale

Most of ProtoAct is not circular: ProtoRAG retrieves from a held-out pair of reference protocols with the 20 test protocols excluded, ActSchema's constrained JSON space is a representational choice rather than a predicted result, and the Pipette simulation plus real-robot VLA experiments give independent (if narrow) execution evidence. The compromising step is RefineChecker. Equation (3) makes the revised parse a function of user-supplied U, while Appendix B defines U as including the very implicit actions and operation flows the module is credited with recovering; with U's provenance for the test protocols undisclosed, the ablation gains in Tables 6-7 are not demonstrated to be autonomous parsing and could be answer-key leakage. The conclusion explicitly concedes 'RefineChecker depends on experiment-specific information provided by users,' confirming the dependence without removing the benchmark-validity problem. Because the rest of the pipeline retains independent content and the robot execution is an external check, this is partial rather than total circularity, hence score 4.

Assumptions & free parameters 3 free parameters · 5 assumptions · 2 invented entities

The framework is entirely evaluated against a benchmark that the authors designed and annotated, so the central claim rests on the validity of that annotation standard and on unspecified experiment-specific notes for RefineChecker. No physical constants or fitted scientific parameters appear; the free parameters listed are manual design choices in the pipeline. The invented entities are author-created datasets and schemas, not natural postulates, and they currently lack external verification because the dataset is not separately released.

free parameters (3)
  • Retrieval top-k for ProtoRAG
    The number of reference fragments retrieved from the knowledge base is a hand-chosen design parameter not reported in the paper; retrieval quality directly affects parsing output.
  • Action primitive vocabulary size and monitoring condition types in ActSchema = 30 action primitives and 4 monitoring condition types
    The action schema is a manually designed closed set; the evaluation and all downstream claims depend on this schema being sufficient for cell-culture protocols.
  • Augmentation hyperparameters for simulation trajectories = not reported
    Scene illumination and execution speed are varied with unspecified ranges, and the augmentation filter uses a task-specific success criterion that is not fully specified.
assumptions (5)
  • domain assumption Human expert annotation of BioP2E is a valid and complete ground truth for protocol parsing.
    Section 4.1 states annotations are manually reviewed by experts for validity, completeness, and executability; all automatic metrics compare against this standard.
  • domain assumption User-provided checking information U exists and is accurate in deployment and evaluation.
    Section 3.3 defines RefineChecker as a function of U, and Section 5 lists dependence on user information as a limitation; the source of U in the 20-protocol evaluation is not specified.
  • domain assumption GPT-5.3 can reliably judge semantic validity and executability of parsing results.
    Section 4.2 uses a GPT-based evaluator following G-Eval, but the same model family is among the evaluated models, so judge reliability and impartiality are assumed.
  • domain assumption The 30-primitive ActSchema is sufficient to represent all cell-culture operations in the selected protocols.
    Appendix C defines the closed action space; if real protocols require actions outside this space, the evaluation would overstate coverage.
  • domain assumption Isaac Sim and Pipette simulation with teleoperation demonstrations are faithful proxies for real wet-lab manipulation.
    Section 4.6 and Appendix F use simulation for VLA training and evaluation; transfer to physical robots is asserted but not quantitatively reported.
invented entities (2)
  • BioP2E dataset
    purpose: Manually annotated resource mapping 22 cell-culture protocols to monitoring conditions, subtasks, and grounded action calls; used for evaluation and demonstration collection.
    The dataset is introduced by the authors; the manuscript provides statistics but no separate dataset URL, license, or commit hash, so external verification and reuse are not yet possible.
  • ActSchema action space
    purpose: A closed set of 4 monitoring condition types and 30 action primitives with parameter fields that constrains LLM output into JSON functions.
    The schema is defined by the authors and is the standard against which outputs are scored; without an external benchmark, its completeness is untested.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ProtoAct: Turning Wet-Lab Protocols into Embodied Robotic Actions." pith.science (2026). https://pith.science/paper/ZXXD534X

@misc{pith2026260801690,
  author       = {Pith},
  title        = {Pith review of: ProtoAct: Turning Wet-Lab Protocols into Embodied Robotic Actions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZXXD534X}},
  note         = {Machine review of arXiv:2608.01690}
}
read the original abstract

Biological wet-lab protocols are written for trained researchers and often leave routine operations, state-dependent conditions, and contextual parameters implicit, making them difficult to translate into robot-executable actions. We present ProtoAct, a structured protocol-grounding framework that converts free-form biological procedures into state-aware, embodiment-ready action sequences. ProtoAct uses ProtoRAG to retrieve manually annotated examples for context-sensitive parsing, employs RefineChecker to detect and revise missing or inconsistent steps, and applies ActSchema to map the refined procedure into constrained JSON function sequences. We further introduce BioP2E, for which we manually annotate 22 cell-culture protocols into 258 monitoring conditions, 910 executable subtasks, and 962 grounded action calls. Evaluation across seven large language models demonstrates that ProtoAct can be effectively instantiated with different backbones. Ablations confirm that retrieval, posterior checking, and schema constraints make complementary contributions. The parsed subtasks further support demonstration collection and VLA model training, enabling successful execution in both simulation and real-robot settings. ProtoAct thus provides a practical interface between biological protocol understanding and embodied robotic execution.

Figures

Figures reproduced from arXiv: 2608.01690 by the authors.

Figure 1
Figure 1. Architecture of the ProtoAct system. and revise the preliminary result according to experiment￾specific notes provided by the user. This process preserves the correctly parsed content while incorporating domain knowl￾edge to address the large language model’s incomplete under￾standing of experimental logic. Finally, under the constraints of ActSchema, the monitoring conditions and subtask se￾quences are mapped into … view at source ↗
Figure 2
Figure 2. Architecture of the ActSchema module. 3.3 RefineChecker: Post-parsing Verification and Targeted Revision RefineChecker is used to verify and revise the initial nat￾ural language parsing result. An architecture diagram for RefineChecker is included in the supplementary material. The input of RefineChecker consists of two parts: the initial parsing result Y (0), and the checking information U provided by the user for … view at source ↗
Figure 3
Figure 3. Cases of four types of parsing defects. This mechanism is lightweight and interactive, making it suitable for forming a feedback loop between human review and model based parsing. 3.4 ActSchema: Constrained Function Selection and Parameter Filling ActSchema is used to convert the revised intermediate rep￾resentation Y (1) into a structured action function sequence. Its structure is shown in [PITH_FULL_IMAGE:figures… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Example embodied execution tasks in simulation and real-robot settings. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Workflow of the ProtoRAG module [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Architecture of the RefineChecker module. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Annotation procedure of the BioP2E dataset. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Action primitive frequency heatmap for selected BioP2E cell culture protocols. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Action type distribution across BioP2E cell culture protocols. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Distribution of manually annotated subtask counts [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Output example of GPT-based parsing quality [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: GPT score comparison for model-module combinations across six cell culture protocol parsing tasks. PR and RC [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: Online evaluation workflow of VLA models in the Isaac Sim simulation platform. [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages

  1. [1]

    , title =

    O'Donoghue, Odhran and Shtedritski, Aleksandar and Ginger, John and Abboud, Ralph and Ghareeb, Ali and Rodriques, Samuel G. , title =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

  2. [2]

    2025 , eprint =

    Yi, Seungjun and Lim, Jaeyoung and Yoon, Juyong , title =. 2025 , eprint =

  3. [3]

    Qiu, Yizhou and Huang, Ziheng and Wang, Zeyu and Liu, Haotian and Qiao, Yuxuan and Hu, Yuchen and Sun, Shuyang and Peng, Hao and Xu, R. X. and Sun, Maosong , title =. 2025 , eprint =

  4. [4]

    and Yachie, Akihiro , title =

    Jiang, Shuo and Evans-Yamamoto, Daniel and Bersenev, Dennis and Palaniappan, Sucheendra K. and Yachie, Akihiro , title =. SLAS Technology , volume =

  5. [5]

    RoboCulture: A Robotics Platform for Automated Biological Experimentation , year =

    Angers, Kevin and Darvish, Kourosh and Yoshikawa, Natsuki and Okhovatian, Shadi and Bannerman, Duncan and Yakavets, Ihar and Shkurti, Florian and Aspuru-Guzik, Al. RoboCulture: A Robotics Platform for Automated Biological Experimentation , year =. 2505.14941 , archivePrefix =

  6. [6]

    2026 , eprint =

    Du, Zhen and Wang, Zeyu and Fei, Haoran and Cao, Xiaoyu and Xiao, Tian and Wang, Qi and Jin, Hao and Gu, Jun and Lu, Qing and Liu, Zhi , title =. 2026 , eprint =

  7. [7]

    Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks , booktitle =

    Lewis, Patrick and Perez, Ethan and Piktus, Aleksandra and Petroni, Fabio and Karpukhin, Vladimir and Goyal, Naman and K. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks , booktitle =

  8. [8]

    Proceedings of the 37th International Conference on Machine Learning , year =

    Guu, Kelvin and Lee, Kenton and Tung, Zora and Pasupat, Panupong and Chang, Ming-Wei , title =. Proceedings of the 37th International Conference on Machine Learning , year =

Show all 28 references
  1. [9]

    and Lespiau, Jean-Baptiste and Damoc, Bogdan and Clark, Aidan , title =

    Borgeaud, Sebastian and Mensch, Arthur and Hoffmann, Jordan and Cai, Trevor and Rutherford, Eliza and Millican, Katie and van den Driessche, George B. and Lespiau, Jean-Baptiste and Damoc, Bogdan and Clark, Aidan , title =. Proceedings of the 39th International Conference on M...

  2. [10]

    Journal of Machine Learning Research , volume =

    Izacard, Gautier and Lewis, Patrick and Lomeli, Maria and Hosseini, Lucas and Petroni, Fabio and Schick, Timo and Dwivedi-Yu, Jane and Joulin, Armand and Riedel, Sebastian and Grave, Edouard , title =. Journal of Machine Learning Research , volume =

  3. [11]

    and Cox, Sam and Schilter, Oliver and Baldassari, Carlo and White, Andrew D

    Bran, Andres M. and Cox, Sam and Schilter, Oliver and Baldassari, Carlo and White, Andrew D. and Schwaller, Philippe , title =. Nature Machine Intelligence , volume =

  4. [12]

    Ichter, Brian and Brohan, Anthony and Chebotar, Yevgen and Finn, Chelsea and Hausman, Karol and Herzog, Alexander and Ho, Daniel and Ibarz, Julian and Irpan, Alex and Jang, Eric and Julian, Ryan and Kalashnikov, Dmitry and Levine, Sergey and Lu, Yao and Parada, Carolina and Ra...

  5. [13]

    Proceedings of the IEEE International Conference on Robotics and Automation , pages =

    Liang, Jacky and Huang, Wenlong and Xia, Fei and Xu, Peng and Hausman, Karol and Ichter, Brian and Florence, Pete and Zeng, Andy , title =. Proceedings of the IEEE International Conference on Robotics and Automation , pages =

  6. [14]

    Proceedings of the IEEE International Conference on Robotics and Automation , pages =

    Singh, Ishika and Blukis, Valts and Mousavian, Arsalan and Goyal, Ankit and Xu, Danfei and Tremblay, Jonathan and Fox, Dieter and Thomason, Jesse and Garg, Animesh , title =. Proceedings of the IEEE International Conference on Robotics and Automation , pages =

  7. [15]

    Proceedings of the 7th Annual Conference on Robot Learning , year =

    Huang, Wenlong and Wang, Chen and Zhang, Ruohan and Li, Yunzhu and Wu, Jiajun and Fei-Fei, Li , title =. Proceedings of the 7th Annual Conference on Robot Learning , year =

  8. [16]

    and Salazar, Grecia and Ryoo, Michael S

    Zitkovich, Brianna and Yu, Tianhe and Xu, Sichun and Xu, Peng and Xiao, Ted and Xia, Fei and Wu, Jialin and Wohlhart, Paul and Welker, Stefan and Wahid, Ayzaan and Vuong, Quan and Vanhoucke, Vincent and Tran, Huong and Soricut, Radu and Singh, Anikait and Singh, Jaspiar and Se...

  9. [17]

    and Sanketi, Pannag R

    Kim, Moo Jin and Pertsch, Karl and Karamcheti, Siddharth and Xiao, Ted and Balakrishna, Ashwin and Nair, Suraj and Rafailov, Rafael and Foster, Ethan P. and Sanketi, Pannag R. and Vuong, Quan and Kollar, Thomas and Burchfiel, Benjamin and Tedrake, Russ and Sadigh, Dorsa and Le...

  10. [18]

    Proceedings of Robotics: Science and Systems , year =

    Black, Kevin and Brown, Noah and Driess, Danny and Esmail, Adnan and Equi, Michael Robert and Finn, Chelsea and Fusai, Niccolo and Groom, Lachy and Hausman, Karol and Ichter, Brian and Jakubczak, Szymon and Jones, Tim and Ke, Liyiming and Levine, Sergey and Li-Bell, Adrian and...

  11. [19]

    2025 , eprint =

    Shukor, Mustafa and Aubakirova, Dana and Capuano, Francesco and Kooijmans, Pepijn and Palma, Steven and Zouitine, Adil and Aractingi, Michel and Pascal, Caroline and Russi, Martino and Marafioti, Andres and Alibert, Simon and Cord, Matthieu and Wolf, Thomas and Cadene, Remi , ...

  12. [20]

    Advances in Neural Information Processing Systems , volume =

    Liu, Bo and Zhu, Yifeng and Gao, Chongkai and Feng, Yihao and Liu, Qiang and Zhu, Yuke and Stone, Peter , title =. Advances in Neural Information Processing Systems , volume =

  13. [21]

    The Fourteenth International Conference on Learning Representations , year =

    Lan, Zhiqian and Jiang, Yuxuan and Wang, Ruiqi and Xie, Xuanbing and Zhang, Rongkui and Zhu, Yicheng and Li, Peihang and Yang, Tianshuo and Chen, Tianxing and Gao, Haoyu and Yang, Xiaokang and Li, Xuelong and Zhang, Hongyuan and Mu, Yao and Luo, Ping , title =. The Fourteenth ...

  14. [22]

    , title =

    Teytelman, Leonid and Stoliartchouk, Anna and Kindler, Lenny and Hurwitz, Bonnie L. , title =. PLOS Biology , volume =

  15. [23]

    Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics , pages =

    Papineni, Kishore and Roukos, Salim and Ward, Todd and Zhu, Wei-Jing , title =. Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics , pages =

  16. [24]

    Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing , pages =

    Beltagy, Iz and Lo, Kyle and Cohan, Arman , title =. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing , pages =

  17. [25]

    , title =

    Levenshtein, Vladimir I. , title =. Soviet Physics Doklady , volume =

  18. [26]

    Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

    Liu, Yang and Iter, Dan and Xu, Yichong and Wang, Shuohang and Xu, Ruochen and Zhu, Chenguang , title =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =

  19. [27]

    The Fourteenth International Conference on Learning Representations , year =

    Cadene, Remi and Alibert, Simon and Capuano, Francesco and Aractingi, Michel and Zouitine, Adil and Kooijmans, Pepijn and Choghari, Jade and Russi, Martino and Pascal, Caroline and Palma, Steven and Aubakirova, Dana and Shukor, Mustafa and Moss, Jess and Soare, Alexander and L...

  20. [28]

    2026 , eprint =

    Liu, Zhe and Jin, Huanbo and Du, Zhaohui and Wang, Zhe and Zhou, Dongzhan and Pan, Minting and Xu, He and Li, Peijia and Gu, Jiaming and Lu, Quan and Wang, Qi and Ji, Bin and Xiao, Ting , title =. 2026 , eprint =

Pith tools

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