Pith. sign in

REVIEW 3 major objections 4 minor 51 references

Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Mimir claims that separating world memory from task memory, and dynamically grounding each active goal in recalled world evidence before acting, is what drives long-horizon embodied success; across 13 vision-language backbones it reports…

desk verdict Mimir's memory architecture is solid and broadly evaluated, but the main gains may hinge on an undisclosed simulator feedback signal. read the letter →

arxiv 2608.04933 v1 pith:RTXZ5WW2 submitted 2026-08-05 cs.RO

Haoming Xu , Zhenlin He , Hengyi Wang , Jiafeng Xu , Hao Dong This is my paper
classification cs.RO
keywords embodiedagentslong-horizonmanipulationneuro-symbolicmemorydynamicgroundingworldtaskvision-languagemodelsEB-Habitat
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

The paper argues that long-horizon embodied agents fail not because they forget too much, but because their memory does not expose an actionable state for the goal currently being executed. Mimir therefore separates world memory, which tracks where objects are and what state they are in, from task memory, which tracks what remains undone and what has failed. Before each action, a grounding step binds the active goal to the best-supported candidate recalled from world memory, fills missing source locations, and attaches the evidence behind the binding. On EB-ALFRED and EB-Habitat, the paper reports that this design improves success rate for all 13 tested backbones, with gains up to 42.5 percentage points and a mean gain of 23.0 points, and that a 32B open model reaches 86.0% success on the EB-Habitat Long-horizon subset, above every closed-source model tested. If true, the explicit task–world interface, not backbone scale, is what makes long-horizon performance possible.

What carries the argument

The load-bearing mechanism is the closed loop $(T_t,W_t) \to$ grounding $\to$ abstract action $\to$ embodiment adapter $\to$ execution feedback, with the grounding operator $\hat{g}_t = G(g_t,R_t,F_t,A_e,x,o_t)$ as the decisive step. Recall returns a bounded candidate list from world memory with each candidate's parent and an evidence score; grounding filters candidates that violate constraints or are contradicted by failures, selects the best-evidence candidate, fills an omitted source from the remembered parent, and returns $\bot_{\mathrm{grd}}$ rather than an unsupported binding. That fail-closed property forces the agent to explore or block a goal instead of acting on an ungrounded query. The two memory write paths are what make the loop stable: observation updates add or refine hypotheses, action updates only write effects supported by $\Delta_t$, and task memory's explicit goal-status map $s_t$ decides completion by whether $\Delta_t$ establishes the goal postcondition.

What would settle it

Run Mimir on EB-Habitat with the $\Delta_t$ channel disabled so world memory is updated only from raw observations, and compare success rate with the full system. If it collapses toward the 12.5% of the 'without world memory' ablation, the reported gains depend on privileged state feedback rather than on the memory organization itself.

Watch

Extended reading notes

Core claim

Mimir's central claim is that memory for embodied agents should be modeled as two evolving states—a world memory $W_t$ holding an episode-level tree of entities, parent relations, attributes, and evidence, and a task memory $T_t$ holding an ordered goal agenda, per-goal status, hand state, failed bindings, and completion evidence. The two are coupled by dynamic grounding: task memory exposes the next unfinished goal, goal-conditioned recall $R(W_t^+, Q(g_t))$ returns a bounded list of candidate entities with their remembered parents and supporting evidence, and the grounding operator $G$ commits to one admissible candidate, resolves a missing source from that candidate's parent, and attaches evidence before planning and execution. World memory is updated by distinct observation and action operators, and the action operator writes only changes supported by the environment feedback $y_t=(\sigma_t,\Delta_t)$, where $\Delta_t$ is the feedback-supported change in object relation, object state, or hand state. The paper claims that this separation and grounding, evaluated over thirteen multimodal backbones on EB-ALFRED and EB-Habitat, improves success rate and goal-condition success rate in every case, with the largest gains exactly where the bare backbone is weakest, and that the resulting system outperforms prior agent and memory systems under a matched backbone and surpasses closed-source foundation models on long-horizon tasks.

Load-bearing premise

The load-bearing premise is that the environment supplies a structured state-change signal, $\Delta_t$, stating exactly which object relation, object state, or hand state changed after each action, and that world memory is updated only from this signal. A physical robot would have to recover that signal from raw vision, and the large world-memory gains rest on it.

Editorial extensions

If this is right

  • With Mimir, InternVL3-8B reaches EB-ALFRED SR 60.0 and GC 68.5, roughly matching or exceeding its 38B and 78B counterparts, so structured memory can substitute for some backbone scale.
  • Removing world memory drops EB-Habitat SR from 65.0 to 12.5 with Qwen3-VL-8B and from 71.5 to 12.5 with Qwen3-VL-32B, so the scene-belief store is the source of the largest single-component gain in interactive environments.
  • Removing task memory costs 23.0 SR points on Qwen3-VL-8B and 37.5 points on Qwen3-VL-32B on EB-ALFRED, showing that an ordered goal agenda carries much of the instruction-following benefit.
  • Under the same Qwen2.5-VL-72B backbone, Mimir raises EB-Habitat SR by 16.0 points and GC by 13.6 points over the strongest prior memory system, RoboMemory.
  • Instruct mode with Mimir outperforms Thinking mode with Mimir on all four metrics, suggesting that externalized progress and evidence can substitute for additional internal reasoning in these tasks.

Reading between the lines

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

  • Editorial inference: the dependence on the structured feedback signal $\Delta_t$ is the main transfer risk; a direct test would disable or corrupt that signal and measure how much of the world-memory gain survives.
  • Editorial inference: since world memory is reset every episode, extending Mimir with persistent cross-episode semantic facts is a natural next step that the paper does not explore.
  • Editorial inference: the fail-closed grounding behavior suggests a cheap diagnostic for any embodied planner: count decisions that act without an evidence-supported binding; the paper's account predicts those actions are a major source of failure in long-horizon tasks.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes Mimir, a neuro-symbolic memory system for long-horizon embodied manipulation. Mimir maintains two separable memories: a world memory that stores scene entities, relations, states, and evidence, and a task memory that tracks an ordered goal agenda, progress, hand state, failures, and constraints. Before each action, a grounding module binds the active goal to recalled world candidates, resolves missing source locations, and attaches evidence before planning. The system is evaluated on EB-ALFRED and EB-Habitat across 13 multimodal backbones, reporting consistent success-rate and goal-condition gains, with maximum gains of 42.5 and 37.5 percentage points on the two benchmarks, and a 86.0% success rate on the EB-Habitat Long-horizon subset using a 32B open-source backbone. The paper claims that the architecture, rather than backbone capacity, is the primary driver of the improvements, supported by ablations and a matched-backbone comparison against prior agent and memory systems.

Significance. If the empirical claim holds, the contribution is valuable: the explicit separation of world memory from task memory and the dynamic grounding operation provide a transparent interface between scene belief and execution progress, addressing a recognized bottleneck in long-horizon embodied agents. The breadth of the evaluation is a genuine strength: testing 13 backbones of varying scale and reporting uniformly positive deltas is stronger evidence than a single-model comparison. The failure analysis is also honest and useful, identifying cases where memory cannot compensate for ambiguous or missing perception. However, the significance depends on resolving a load-bearing ambiguity about the source of the environment feedback signal used to update world memory, and on clarifying the evaluation protocol with respect to variance and comparison subsets.

major comments (3)
  1. [Method, Eq. (4) and Eq. (8); Ablation, Table 4] The source of y_t = (sigma_t, Delta_t) is never disclosed. In Eq. (8), the action update writes 'only effects supported by Delta_t', and the text describes Delta_t as 'feedback-supported change in object relation, object state, or hand state'. If Delta_t is obtained from simulator ground truth (e.g., the environment's internal object-relation and hand-state fields), then the world-memory component is copying privileged telemetry rather than inferring scene belief from perception. In that case, the 'Without world memory' ablation in Table 4, which drops EB-Habitat SR from 65.0 to 12.5 (Qwen3-VL-8B) and from 71.5 to 12.5 (Qwen3-VL-32B), measures the value of a privileged state-change signal, not the value of the proposed memory architecture per se. The manuscript must specify exactly how Delta_t is computed in EB-ALFRED and EB-Habitat, and should include an additional ablation where Delta_t is either estimated from observations only or removed entirely, so that the contribution of the grounding and memory structure can be separated from the contribution of the privileged signal.
  2. [Experiments, Metrics and Table 1] The evaluation protocol is under-specified with respect to variance. The Metrics section states 'Each task is executed once' and then says the experiments were 'repeated ... and consistently obtained the same results', which is internally ambiguous: either the benchmark is deterministic given temperature 0, or the repetition is not described precisely. Because the central claim is that Mimir 'consistently improves' across all 13 backbones, the paper should state explicitly whether the environment is deterministic, report the number of seeds or runs if any, and provide a per-task breakdown of the four tasks that make up the Table 1 averages (Base, Common-sense, Complex-instruction, Long-horizon). Without this, the reported deltas are point estimates with no measure of variability.
  3. [Experiments, Table 2] Table 2 is not directly comparable with Table 1. The caption says each value is the 'unweighted mean of Base and Long-horizon', following RoboMemory, whereas Table 1 reports four-task averages over Base, Common-sense, Complex-instruction, and Long-horizon. The claim that Mimir improves the overall average SR by 8.5% over prior systems is therefore based on a different subset than the headline numbers in Table 1. The comparison should be repeated on a consistent subset: either report the full four-task averages for all methods, or clearly state that the 8.5% improvement refers only to the Base+Long-horizon subset and confirm that the advantage persists on the other two tasks.
minor comments (4)
  1. [Abstract] There is a typo in the first sentence: 'Long-horizonembodiedtaskrequires' should be 'Long-horizon embodied tasks require'.
  2. [Metrics] The sentence 'And werepeated the experiments multipletimes and consistently obtained the same results' has grammatical errors and should be rewritten; it should also specify whether the repetition refers to multiple episodes of the same task or to re-running with different random seeds.
  3. [Method, Eq. (9)] The candidate list size K_t in Eq. (9) is a free parameter, but no sensitivity analysis is reported. A short discussion of how K_t is chosen and whether results are stable to its value would improve reproducibility.
  4. [Experiments, Table 3] The phrase 'With a open-source backbone' should be 'With an open-source backbone'. In addition, the comparison with closed-source models would be clearer if the inference budget or decoding configuration of the closed-source baselines were specified, since the paper states only the maximum generation budget for all entries.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; Mimir's empirical gains are measured on external benchmarks and are not forced by construction.

full rationale

The paper makes an empirical performance claim, not a derivation. Mimir's components are defined (world memory T_t, task memory W_t, grounding Ĝ_t in Eqs. 1-11) and then evaluated on the external EmbodiedBench (EB-ALFRED, EB-Habitat) tasks, with success measured by benchmark evaluators rather than by Mimir's own internal state or equations. The baseline comparisons use the same backbone with and without Mimir, so the reported gains are not fitted or renamed predictions; they are measured differences in task completion. Ablations remove entire components (world memory or task memory) and report benchmark effects, which is a component analysis rather than a circular derivation. The load-bearing concern about y_t = (σ_t, Δ_t) being privileged simulator feedback (Section Method, Eq. 4, Eq. 8) is a correctness risk about transfer to physical robots, not circularity: the paper never claims to infer Δ_t from perception, and the failure analysis explicitly states memory 'cannot reconstruct evidence that never entered the perceptual stream.' No self-citation chain or imported uniqueness theorem is used to force the central claim. The design choices (e.g., separating world and task memory) are presented as a hypothesis, and the experiments test it against external baselines. Therefore, there is no equation-level reduction, no fitted-parameter-as-prediction, and no load-bearing self-citation; the result is self-contained as an empirical evaluation.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

No physically invented entities are introduced. The free parameters and axioms listed are the main unstated or privileged elements the results rest on: simulator state-change feedback, a pre-existing goal agenda, and reliable VLM grounding. The candidate list size is an unreported design choice that could affect behavior.

free parameters (1)
  • candidate list size K_t
    The retrieval operator returns a bounded list R_t (Eq. 9), but the bound K_t and the evidence weights eta_t,k are not specified. These hand-chosen implementation values affect which candidates grounding can select.
assumptions (3)
  • domain assumption The environment returns a structured feedback signal y_t = (sigma_t, Delta_t) that identifies supported changes in object relations, object states, and hand state.
    Invoked in Eq. (4) and Eq. (8). Without this privileged state-change signal, world memory cannot be updated the way Mimir does, and the ablation shows world memory is essential on EB-Habitat.
  • domain assumption A symbolic goal agenda G with ordering ⪯ is available before execution, decomposing the instruction into ordered subgoals.
    Used throughout Task Memory (Eq. 2-3). The paper does not describe how G is initialized or validated; Mimir's progress tracking presumes this decomposition is correct.
  • domain assumption The VLM grounding module can reliably select the correct candidate from the recalled candidate set given the current observation and memory descriptions.
    Grounding operator G in Eq. (10) delegates disambiguation to a VLM restricted to candidates. The paper's own failure analysis admits that ambiguous language or poor visual evidence can cause incorrect or impossible grounding.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments." pith.science (2026). https://pith.science/paper/RTXZ5WW2

@misc{pith2026260804933,
  author       = {Pith},
  title        = {Pith review of: Mimir: A Neuro-Symbolic Memory System with Dynamic Grounding for Embodied Agents in Interactive Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RTXZ5WW2}},
  note         = {Machine review of arXiv:2608.04933}
}
read the original abstract

Long-horizon embodied task requires agents to act under partial observability while preserving both scene belief and execution progress. Flat histories or implicit policy states may contain past observations, but they do not provide an explicit interface for deciding which world facts support the currently active goal. We introduce Mimir, a neuro-symbolic memory that separates world memory from task memory and dynamically grounds them before each action. World memory maintains object locations, object states, and perceptual evidence, while task memory maintains an ordered goal agenda, progress state, hand state, failures, and execution constraints. A grounding module binds the active goal to recalled world candidates, fills missing source locations, and attaches evidence before planning and embodiment-specific execution. Across tested backbones, Mimir consistently improves on different EB-ALFRED and EB-Habitat tasks, with maximum gains of 42.5% and average gains of 23.0%, respectively. Compared with the best results among prior agent and memory systems evaluated under the same backbone, Mimir improves the overall average success rate by 8.5%. Finally, on the EB-Habitat Long-horizon subset, Mimir achieves 86.0% success rate, substantially outperforming current closed-source models. Our code will be released soon.

Figures

Figures reproduced from arXiv: 2608.04933 by the authors.

Figure 1
Figure 1. Motivation of Mimir. Long-horizon embodied tasks require memory of both physical world state and task progress. Flat-history agents can retain past observations yet still lose object locations or confuse completed and pending subgoals. Mimir maintains explicit world and task memories, and dynamically grounds the active goal against recalled world evidence before action. failures, and translates grounded abstract act… view at source ↗
Figure 2
Figure 2. Overview of Mimir. Task memory preserves execution progress, while world memory maintains scene entities, relations, and evidence. Goal-conditioned recall retrieves candidates for the active goal, and dynamic grounding selects an evidence-supported binding for planning and embodiment-specific execution. Observations and execution feedback update the two memories online. For each non-root entity v ∈ Vt, the relations… view at source ↗
Figure 3
Figure 3. Representative failure cases under ambiguous in [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 48 canonical work pages

  1. [1]

    Baddeley and Graham Hitch , title =

    Alan D. Baddeley and Graham Hitch , title =. Psychology of Learning and Motivation , volume =

    work page

  2. [2]

    Organization of Memory , pages =

    Endel Tulving , title =. Organization of Memory , pages =

    work page

  3. [3]

    2023 , eprint =

    Shunyu Yao and Jeffrey Zhao and Dian Yu and Nan Du and Izhak Shafran and Karthik Narasimhan and Yuan Cao , title =. 2023 , eprint =

    work page 2023

  4. [4]

    2023 , eprint =

    Noah Shinn and Federico Cassano and Edward Berman and Ashwin Gopinath and Karthik Narasimhan and Shunyu Yao , title =. 2023 , eprint =

    work page 2023

  5. [5]

    2023 , eprint =

    Guanzhi Wang and Yuqi Xie and Yunfan Jiang and Ajay Mandlekar and Chaowei Xiao and Yuke Zhu and Linxi Fan and Anima Anandkumar , title =. 2023 , eprint =

    work page 2023

  6. [6]

    Patil and Ion Stoica and Joseph E

    Charles Packer and Sarah Wooders and Kevin Lin and Vivian Fang and Shishir G. Patil and Ion Stoica and Joseph E. Gonzalez , title =. 2023 , eprint =

    work page 2023

  7. [7]

    Michael Ahn and Anthony Brohan and Noah Brown and Yevgen Chebotar and Omar Cortes and Byron David and Chelsea Finn and Chuyuan Fu and Keerthana Gopalakrishnan and Karol Hausman and Alex Herzog and Daniel Ho and Jasmine Hsu and Julian Ibarz and Brian Ichter and Alex Irpan and Eric Jang and Rosario Jauregui Ruano and Kyle Jeffrey and Sally Jesmonth and Nikh...

    work page 2022

  8. [8]

    Sadler and Wei-Lun Chao and Yu Su , title =

    Chan Hee Song and Jiaman Wu and Clayton Washington and Brian M. Sadler and Wei-Lun Chao and Yu Su , title =. 2023 , eprint =

    work page 2023

Show all 51 references
  1. [9]

    Danny Driess and Fei Xia and Mehdi S. M. Sajjadi and Corey Lynch and Aakanksha Chowdhery and Brian Ichter and Ayzaan Wahid and Jonathan Tompson and Quan Vuong and Tianhe Yu and Wenlong Huang and Yevgen Chebotar and Pierre Sermanet and Daniel Duckworth and Sergey Levine and Vin...

    2023

  2. [10]

    2024 , eprint =

    Minghao Chen and Yihang Li and Yanting Yang and Shiyu Yu and Binbin Lin and Xiaofei He , title =. 2024 , eprint =

    2024

  3. [11]

    2024 , eprint =

    Dayuan Fu and Biqing Qi and Yihuai Gao and Che Jiang and Guanting Dong and Bowen Zhou , title =. 2024 , eprint =

    2024

  4. [12]

    2024 , eprint =

    Zixuan Wang and Bo Yu and Junzhe Zhao and Wenhao Sun and Sai Hou and Shuai Liang and Xing Hu and Yinhe Han and Yiming Gan , title =. 2024 , eprint =

    2024

  5. [13]

    2025 , eprint =

    Mingcong Lei and Honghao Cai and Yuyuan Yang and Yimou Wu and Jinke Ren and Zezhou Cui and Liangchen Tan and Junkun Hong and Gehan Hu and Shuangyu Zhu and Shaohan Jiang and Ge Wang and Junyuan Tan and Zhenglin Wan and Zheng Li and Zhen Li and Shuguang Cui and Yiming Zhao and Y...

    2025

  6. [14]

    arXiv preprint arXiv:2311.17406 , year =

    Siwei Chen and Anxing Xiao and David Hsu , title =. arXiv preprint arXiv:2311.17406 , year =

    arXiv

  7. [15]

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

    Byeonghwi Kim and Jinyeon Kim and Yuyeong Kim and Cheolhong Min and Jonghyun Choi , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

  8. [16]

    SayPlan: Grounding Large Language Models Using

    Krishan Rana and Jesse Haviland and Sourav Garg and Jad Abou-Chakra and Ian Reid and Niko S. SayPlan: Grounding Large Language Models Using. Proceedings of the 7th Conference on Robot Learning , pages =

  9. [17]

    Tenenbaum and Antonio Torralba and Florian Shkurti and Liam Paull , title =

    Qiao Gu and Alihusein Kuwajerwala and Sacha Morin and Krishna Murthy Jatavallabhula and Bipasha Sen and Aditya Agarwal and Corban Rivera and William Paul and Kirsty Ellis and Rama Chellappa and Chuang Gan and Celso Miguel de Melo and Joshua B. Tenenbaum and Antonio Torralba an...

    2023

  10. [18]

    2024 , eprint =

    Yuncong Yang and Han Yang and Jiachen Zhou and Peihao Chen and Hongxin Zhang and Yilun Du and Chuang Gan , title =. 2024 , eprint =

    2024

  11. [19]

    2023 , eprint =

    Anthony Brohan and Noah Brown and Justice Carbajal and Yevgen Chebotar and Xi Chen and Krzysztof Choromanski and Tianli Ding and Danny Driess and Avinava Dubey and Chelsea Finn and Pete Florence and Chuyuan Fu and Montse Gonzalez Arenas and Keerthana Gopalakrishnan and Kehang ...

    2023

  12. [20]

    2024 , eprint =

    Moo Jin Kim and Karl Pertsch and Siddharth Karamcheti and Ted Xiao and Ashwin Balakrishna and Suraj Nair and Rafael Rafailov and Ethan Foster and Grace Lam and Pannag Sanketi and Quan Vuong and Thomas Kollar and Benjamin Burchfiel and Russ Tedrake and Dorsa Sadigh and Sergey L...

    2024

  13. [21]

    2025 , eprint =

    Hao Shi and Bin Xie and Yingfei Liu and Lin Sun and Fengrong Liu and Tiancai Wang and Erjin Zhou and Haoqiang Fan and Xiangyu Zhang and Gao Huang , title =. 2025 , eprint =

    2025

  14. [22]

    2025 , eprint =

    Ajay Sridhar and Jennifer Pan and Satvik Sharma and Chelsea Finn , title =. 2025 , eprint =

    2025

  15. [23]

    2025 , eprint =

    Runhao Li and Wenkai Guo and Zhenyu Wu and Changyuan Wang and Haoyuan Deng and Zhenyu Weng and Yap-Peng Tan and Ziwei Wang , title =. 2025 , eprint =

    2025

  16. [24]

    2025 , eprint =

    Rui Yang and Hanyang Chen and Junyu Zhang and Mark Zhao and Cheng Qian and Kangrui Wang and Qineng Wang and Teja Venkat Koripella and Marziyeh Movahedi and Manling Li and Heng Ji and Huan Zhang and Tong Zhang , title =. 2025 , eprint =

    2025

  17. [25]

    Atkinson and Richard M

    Richard C. Atkinson and Richard M. Shiffrin , title =. Psychology of Learning and Motivation , volume =

  18. [26]

    2025 , eprint =

    Prateek Chhikara and Dev Khant and Saket Aryan and Taranjeet Singh and Deshraj Yadav , title =. 2025 , eprint =

    2025

  19. [27]

    Tenenbaum and Tianmin Shu and Chuang Gan , title =

    Hongxin Zhang and Weihua Du and Jiaming Shan and Qinhong Zhou and Yilun Du and Joshua B. Tenenbaum and Tianmin Shu and Chuang Gan , title =. 2023 , eprint =

    2023

  20. [28]

    2025 , eprint =

    Mingcong Lei and Ge Wang and Yiming Zhao and Zhixin Mai and Qing Zhao and Yao Guo and Zhen Li and Shuguang Cui and Yatong Han and Jinke Ren , title =. 2025 , eprint =

    2025

  21. [29]

    2025 , eprint =

    Marc Glocker and Peter H. 2025 , eprint =

    2025

  22. [30]

    2025 , eprint =

    Wonje Choi and Jinwoo Park and Sanghyun Ahn and Daehee Lee and Honguk Woo , title =. 2025 , eprint =

    2025

  23. [31]

    2026 , eprint =

    Xiaoyu Ma and Lianyu Hu and Wenbing Tang and Zixuan Hu and Zeqin Liao and Zhizhen Wu and Yang Liu , title =. 2026 , eprint =

    2026

  24. [32]

    2026 , eprint =

    Xin Ding and Xinrui Wang and Yifan Yang and Hao Wu and Shiqi Jiang and Qianxi Zhang and Liang Mi and Hanxin Zhu and Kun Li and Yunxin Liu and Zhibo Chen and Ting Cao , title =. 2026 , eprint =

    2026

  25. [33]

    2026 , eprint =

    Zebin Yang and Tong Xie and Baotong Lu and Shaoshan Liu and Bo Yu and Meng Li , title =. 2026 , eprint =

    2026

  26. [34]

    2022 , eprint =

    Chenguang Huang and Oier Mees and Andy Zeng and Wolfram Burgard , title =. 2022 , eprint =

    2022

  27. [35]

    Tenenbaum and Celso Miguel de Melo and Madhava Krishna and Liam Paull and Florian Shkurti and Antonio Torralba , title =

    Krishna Murthy Jatavallabhula and Alihusein Kuwajerwala and Qiao Gu and Mohd Omama and Tao Chen and Alaa Maalouf and Shuang Li and Ganesh Iyer and Soroush Saryazdi and Nikhil Keetha and Ayush Tewari and Joshua B. Tenenbaum and Celso Miguel de Melo and Madhava Krishna and Liam ...

    2023

  28. [36]

    2023 , eprint =

    Matthew Chang and Theophile Gervet and Mukul Khanna and Sriram Yenamandra and Dhruv Shah and So Yeon Min and Kavit Shah and Chris Paxton and Saurabh Gupta and Dhruv Batra and Roozbeh Mottaghi and Jitendra Malik and Devendra Singh Chaplot , title =. 2023 , eprint =

    2023

  29. [37]

    2025 , eprint =

    Joel Loo and Zhanxin Wu and David Hsu , title =. 2025 , eprint =

    2025

  30. [38]

    2025 , eprint =

    Wenbo Hu and Yining Hong and Yanjun Wang and Leison Gao and Zibu Wei and Xingcheng Yao and Nanyun Peng and Yonatan Bitton and Idan Szpektor and Kai-Wei Chang , title =. 2025 , eprint =

    2025

  31. [39]

    2025 , eprint =

    Yufan Mao and Hanjing Ye and Wenlong Dong and Chengjie Zhang and Hong Zhang , title =. 2025 , eprint =

    2025

  32. [40]

    2025 , eprint =

    Lingfeng Zhang and Xiaoshuai Hao and Yingbo Tang and Haoxiang Fu and Xinyu Zheng and Pengwei Wang and Zhongyuan Wang and Wenbo Ding and Shanghang Zhang , title =. 2025 , eprint =

    2025

  33. [41]

    2023 , eprint =

    Open. 2023 , eprint =

    2023

  34. [42]

    2024 , eprint =

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

    2024

  35. [43]

    2025 , eprint =

    Min Lin and Xiwen Liang and Bingqian Lin and. 2025 , eprint =

    2025

  36. [44]

    2026 , eprint =

    Xinying Guo and Chenxi Jiang and Hyun Bin Kim and Ying Sun and Yang Xiao and Yuhang Han and Jianfei Yang , title =. 2026 , eprint =

    2026

  37. [45]

    Marcel Torne and Karl Pertsch and Homer Walke and Kyle Vedder and Suraj Nair and Brian Ichter and Allen Z. Ren and Haohuan Wang and Jiaming Tang and Kyle Stachowicz and Karan Dhabalia and Michael Equi and Quan Vuong and Jost Tobias Springenberg and Sergey Levine and Chelsea Fi...

    2026

  38. [46]

    2026 , eprint =

    Hang Li and Fengyi Shen and Dong Chen and Liudi Yang and Xudong Wang and Jinkui Shi and Zhenshan Bing and Ziyuan Liu and Alois Knoll , title =. 2026 , eprint =

    2026

  39. [47]

    2026 , eprint =

    Zijian Zeng and Fei Ding and Huiming Yang and Xianwei Li , title =. 2026 , eprint =

    2026

  40. [48]

    2026 , eprint =

    Yuheng Lei and Zhixuan Liang and Hongyuan Zhang and Ping Luo , title =. 2026 , eprint =

    2026

  41. [49]

    2026 , eprint =

    Haoyang Li and Yang You and Hao Su and Leonidas Guibas , title =. 2026 , eprint =

    2026

  42. [50]

    Cradle: Empowering Foundation Agents Towards General Computer Control , year =

    Weihao Tan and Wentao Zhang and Xinrun Xu and Haochong Xia and Ziluo Ding and Boyu Li and Bohan Zhou and Junpeng Yue and Jiechuan Jiang and Yewen Li and Ruyi An and Molei Qin and Chuqiao Zong and Longtao Zheng and Yujie Wu and Xiaoqiang Chai and Yifei Bi and Tianbao Xie and Pe...

    doi arXiv

  43. [51]

    2025 , eprint =

    Huajie Tan and Xiaoshuai Hao and Cheng Chi and Minglan Lin and Yaoxu Lyu and Mingyu Cao and Dong Liang and Zhuo Chen and Mengsi Lyu and Cheng Peng and Chenrui He and Yulong Ao and Yonghua Lin and Pengwei Wang and Zhongyuan Wang and Shanghang Zhang , title =. 2025 , eprint =. d...

    2025 doi arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.