Pith. sign in

REVIEW 4 major objections 5 minor 29 references

World Action Models predict actions plus imagined futures; this paper claims cross-view depth reprojection consistency from a frozen geometry model ranks those futures, enabling label-free robot rollout selection.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 17:53 UTC pith:B5KFVNSW

load-bearing objection Useful training-free test-time scaling for WAMs with modest but consistent gains; the main open question is whether the frozen geometry model's scores are trustworthy on synthetic frames. the 4 major comments →

arxiv 2607.17454 v1 pith:B5KFVNSW submitted 2026-07-20 cs.RO

Test-Time Scaling for World Action Models via Zero-Shot Geometric Evaluation

Zesen Zhao , Minkyoung Cho , Hui shen , Boyuan Zheng , Kunxiao Gao , Yulong Cao , Z.Morley Mao This is my paper
classification cs.RO
keywords test-time scalingworld action modelsrobot manipulationbest-of-N selectioncross-view depth reprojectiongeometric consistencyaction–future consistency gatetraining-free verification
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Robot policies called World Action Models (WAMs) generate both an action chunk and a predicted future video at each decision point. This paper tries to establish that the imagined future itself is usable as a test-time selection signal: a cheap action–future consistency gate decides when the initial rollout looks internally inconsistent, and when it does, a frozen geometry model ranks additional rollouts by cross-view depth reprojection inconsistency—how poorly the wrist-camera predicted depth map, projected into the primary camera, matches the primary view's predicted depth. The claim is that this training-free, task-label-free check improves fixed-budget Best-of-N success in every tested setting—for example, RoboCasa group average from 66.3% to 68.4% with Cosmos Policy—and that gating recovers about 74.8% of the always-on gain while sampling extra candidates at only 26.2% of decision points. A sympathetic reader would care because it suggests WAM inference can be scaled selectively using only the model's own multi-view predictions, without learned verifiers or environment feedback.

Core claim

The central discovery is that cross-view depth reprojection inconsistency is a strong task-label-free selector for WAM rollouts. The paper defines the score as the average absolute log-ratio between the depth predicted directly for the primary view and the wrist-view depth map reprojected into the primary camera, using a frozen geometry model (VGGT-Ω) for depth and pose; lower scores mean the two predicted views correspond to one coherent 3D scene. Around this score the paper builds GeoBoN, a fixed-budget Best-of-N selector that improves N=8 success in all five benchmark–backbone settings, and Gated GeoBoN, which invokes sampling only when an optical-flow action–future consistency test flags

What carries the argument

The load-bearing object is the cross-view depth reprojection inconsistency score (Eq. 5): for each sampled rollout, feed the predicted primary and wrist frames into the frozen geometry model VGGT-Ω, then average |log(d_proj/d_vggt)| over pixels where the projected wrist points land in the primary view with positive depth and sufficient model confidence. It converts 'do these two imagined views depict the same 3D scene?' into a single number. The action–future gate (Section 3.1) computes cosine agreement between dense optical flow in a capsule around the projected end-effector trajectory and the motion implied by the generated action chunk; a low agreement triggers additional sampling. Both s

Load-bearing premise

The method assumes that a frozen depth-and-pose model, when fed the synthetic frames a WAM imagines, produces errors small and uncorrelated enough across candidates that the lowest reprojection error really marks the most coherent future; if those errors are large or correlated, the selector picks a biased outlier. This premise enters at Eq. 5 (Section 3.2) and is not directly validated.

What would settle it

Take a fixed set of RoboCasa tasks, sample N=8 rollouts per decision, run every candidate in simulation, and check whether the candidate with the lowest Eq. 5 reprojection error succeeds more often than random or value-head picks; if low reprojection error does not track simulation success on the WAM's predicted frames, the central claim fails. This can be tested offline from the paper's candidate dumps.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Fixed-budget GeoBoN improves N=8 task success in every evaluated setting; on RoboCasa the group average rises from 66.3% to 68.4% with Cosmos Policy and from 80.8% to 82.5% with X-WAM.
  • Gated GeoBoN recovers on average 74.8% of the always-on success gain while triggering extra sampling at only 26.2% of decision points, cutting latency substantially (e.g., 3.65s to 1.29s on RoboCasa/Cosmos Policy).
  • Cross-view reprojection is the only selector among the tested training-free signals (confidence-only, future-consensus) that improves closed-loop success in all settings; it also beats the Cosmos Policy value head at identical candidate budgets.
  • The gate is an informative compute-allocation signal: at equal trigger rates it raises the 'GeoBoN helps' rate above a random trigger in all settings, by up to 24 points on RoboCasa/X-WAM.
  • Larger candidate pools create a multiple-comparisons failure mode: false low-score selections rise from 7% at N=2 to 32% at N=16, explaining why Best-of-N success saturates or degrades and motivating moderate budgets with selective invocation.

Where Pith is reading between the lines

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

  • A natural extension the paper leaves implicit: the same reprojection score could serve as a self-supervised preference signal for fine-tuning WAMs, since it predicts rollout quality without success labels; one could generate on-policy candidates and optimize against the score.
  • The gate-plus-evaluator split suggests a compute-adaptive controller that tunes the gate threshold per task or per domain; the paper fixes τgate = −0.2 across all settings, and the diagnostics show gate informativeness varies (e.g., only +4.1 points on RoboTwin/Motus), so adapting the threshold could recover more gain.
  • Because the geometric evaluator is model-agnostic and operates on multi-view predicted frames, the same recipe could transfer to other multi-view world models or video prediction benchmarks outside manipulation, where no action labels or task rewards are available.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Gated GeoBoN, a training-free test-time scaling method for World Action Models (WAMs). A cheap action–future consistency gate decides whether to sample additional rollouts; when it triggers, a frozen VGGT-Ω geometry model ranks candidates by cross-view depth reprojection inconsistency (Eq. 5). Fixed-budget GeoBoN is reported to improve N=8 task success in all five benchmark–backbone settings (e.g., RoboCasa/Cosmos 66.3→68.4; RoboCasa/X-WAM 80.8→82.5), and gating recovers on average 74.8% of the always-on gain while triggering on 26.2% of decision points. Additional offline diagnostics compare selectors and gate against alternatives, and identify false low-score selections as a failure mode.

Significance. If the mechanism is correct, this is a useful contribution: it shows that WAM-exposed future frames can support task-label-free, training-free rollout selection and selective compute allocation, with closed-loop evidence across multiple backbones and benchmarks, and it compares favorably to a learned value head. The strengths are the multi-setting closed-loop evaluation, the comparison against future-consensus and value-head baselines, and the attempt to isolate the gate and selector contributions. However, the central geometric evaluator is not validated on the distribution it is applied to (WAM-generated frames), and the false-selection analysis is defined in a way that may be circular; these issues must be resolved before the results can be fully trusted.

major comments (4)
  1. [§3.2, Eq. (5); §4.5/Fig. 2] The cross-view depth reprojection score is computed entirely under VGGT-Ω's estimated depth and camera pose on predicted future frames. The paper provides no evidence that VGGT-Ω is calibrated on WAM-generated frames; the X-WAM rows use the backbone's native depth head, so they do not validate the VGGT branch. Since the score is used only to compare candidates, correlated VGGT errors across candidates could produce the observed rankings without true geometric consistency. Fig. 2/§4.5 shows false low-score selections increasing from 7% (N=2) to 32% (N=16), confirming that the score is noisy on predicted frames. Given the modest closed-loop gains (RoboCasa/Cosmos +2.1; X-WAM Door/Drawer −4.6), a negative-control experiment is required: e.g., compare VGGT reprojection scores on predicted frames against simulator ground-truth depth/pose, or show that low-score selections correspond to genuin
  2. [§4.5/Fig. 2] The false low-score selection analysis defines 'false' via LPIPS near-duplicates below 0.02 and a duplicate-pair variation threshold. This is circular: a low reprojection score could be a true signal of geometric consistency even if images are perceptually near-identical, and LPIPS < 0.02 is an extremely tight threshold. The analysis also lacks counts of near-duplicate pairs, error bars, and comparison against a ground-truth consistency label. As this analysis is used to explain saturation/degradation at large N, please either define false selections with respect to ground-truth geometry/task outcome, or present it as a heuristic with appropriate caveats and statistical detail.
  3. [§3.1/§4.1] All thresholds are fixed, but no sensitivity analysis is reported. The gate threshold τ_gate = −0.2 is particularly surprising: the gate triggers only when the cosine between optical flow and projected action motion is strongly negative, whereas an inconsistency detector would be expected to trigger at low positive or near-zero values. The trigger rates (14.2–34.5%) and the claimed compute savings depend directly on this choice. Please report a threshold sweep on a held-out set, or justify the choice by a principled argument, to show that Gated GeoBoN's benefits are not tied to a single fragile hyperparameter.
  4. [Table 3/§4.4] The selector comparison lacks confidence intervals for online success and error recovery. In RoboTwin/Motus, GeoBoN (89.9%) is actually worse than Consensus (90.5%), and in RoboCasa/Cosmos, VGGT Conf (65.9%) is below baseline (66.3%). The phrase 'most consistent selector' needs statistical support (e.g., paired confidence intervals over seeds/tasks); otherwise, the offline ER advantage may not transfer to closed-loop settings.
minor comments (5)
  1. [Table 1] Clarify how the group average and its confidence interval are computed (e.g., paired by seed/task), and report the number of tasks and seeds per setting.
  2. [Fig. 2] Add error bars and sample sizes, and define the denominator for the false-selection rate.
  3. [Related Work] Typo: 'WA V' should be 'WAV' (World Action Verifier).
  4. [§4.2] Per-task improvements are often within error bars; emphasize the group-level results to avoid overclaiming category-level gains.
  5. [General] Consider releasing evaluation scripts and a reproducibility statement to support the 'training-free' and 'zero-shot' claims.

Circularity Check

0 steps flagged

No material circularity: the selection score and gate are fixed, label-free functions of rollouts and a frozen geometry model.

full rationale

The derivation is self-contained as an empirical method rather than a first-principles prediction. The selector in Eq. (5) is defined only from the predicted frame pair and frozen VGGT-Omega: the paper states 'we feed the predicted frame pair (I_hat_pri_i, I_hat_wri_i) into frozen VGGT-Omega' and averages log depth ratios; it contains no task success label, oracle term, or fitted parameter. The action-future gate in Eq. (4) is a fixed cosine comparison between optical flow and projected action displacement, with all thresholds 'fixed across experiments'. The oracle appears only in offline diagnostics: 'the oracle is used only for offline analysis', so it cannot leak into the closed-loop selection rule. The X-WAM variant uses the backbone's own depth head, which weakens the independence of the metric but is not circular: it is still a fixed scoring rule applied to candidate rollouts, not a fit to benchmark outcomes. Citations to WAM backbones, VGGT-Omega, and benchmarks are external; no load-bearing claim is justified by a self-citation. The concern that VGGT-Omega may be miscalibrated on WAM-generated frames is a validity/robustness risk, not an equivalence-by-construction, so it does not raise the circularity score.

Axiom & Free-Parameter Ledger

5 free parameters · 3 axioms · 0 invented entities

The central method introduces five hand-set thresholds and relies on two heuristics (action-future consistency, cross-view geometric consistency) plus a frozen external geometry model. It introduces no new entities and fits no parameters to the benchmark outcomes.

free parameters (5)
  • gate threshold tau_gate = -0.2
    Cosine threshold below which the action-future gate triggers additional sampling; fixed across experiments, not tuned per benchmark.
  • VGGT confidence threshold gamma_conf = 0.5
    Threshold for including pixels in the reprojection error; chosen a priori.
  • idle displacement threshold delta_idle = 1 cm
    Used to ignore arms with negligible motion; chosen a priori.
  • capsule radius rho_cap = 30 pixels
    Radius around projected end-effector trajectory for averaging optical flow; chosen a priori.
  • LPIPS near-duplicate threshold = 0.02
    Used in failure analysis to identify near-identical candidate futures; diagnostic threshold, not part of the method.
axioms (3)
  • domain assumption Cross-view geometric consistency of predicted future frames is a task-label-free indicator of rollout quality
    The method assumes that a consistent 3D scene across views correlates with task success, which is not proven but supported by experiments.
  • domain assumption Agreement between optical flow of the predicted future and projected end-effector motion indicates a reliable rollout
    The gate assumes that if the action matches the visual motion, the rollout is likely correct; this is a heuristic.
  • domain assumption VGGT-Ω provides accurate depth and camera poses on synthetic predicted frames
    The method relies on a frozen geometry model that was not trained on WAM-generated images; its reliability is tested only indirectly.

pith-pipeline@v1.3.0-alltime-deepseek · 129 in / 9432 out tokens · 130405 ms · 2026-08-01T17:53:12.472007+00:00 · methodology

0 comments
read the original abstract

Test-time scaling improves foundation-model inference by spending additional computation, but robot control requires deciding whether extra compute is useful before executing an action. World Action Models (WAMs) make this decision natural: each rollout exposes both an action chunk and predicted future observations. We propose \methodgated, a training-free selective test-time scaling framework for WAMs. We first instantiate \method, a fixed-budget Best-of-$N$ selector that ranks sampled rollouts by cross-view depth reprojection consistency of their predicted futures, computed with a frozen geometry foundation model. \methodgated\ adds a lightweight action--future consistency gate that invokes \method\ only when the initial rollout appears internally inconsistent. Across five benchmark--backbone settings on RoboCasa, LIBERO Long, and RoboTwin~2.0, fixed-budget \method\ improves $N{=}8$ task success in every setting, e.g., raising the RoboCasa group average from $66.3\%$ to $68.4\%$ with Cosmos Policy and from $80.8\%$ to $82.5\%$ with X-WAM. With gating enabled, \methodgated\ recovers on average $74.8\%$ of the always-on success gain while triggering additional sampling on only $26.2\%$ of decision points. Offline diagnostics show that cross-view reprojection is a strong task-label-free selector, and we identify false low-score selections as a failure mode that helps explain why performance can saturate or degrade as $N$ increases.

Figures

Figures reproduced from arXiv: 2607.17454 by Boyuan Zheng, Hui Shen, Kunxiao Gao, Minkyoung Cho, Yulong Cao, Zesen Zhao, Z.Morley Mao.

Figure 1
Figure 1. Figure 1: The action–future gate decides whether the initial WAM rollout is reliable enough to execute directly; when it triggers, GeoBoN [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: False low-score selection rate as the candidate budget [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Representative false low-score selection. The candidate futures are visually near-identical, but the geometric evaluator assigns one [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

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

29 extracted references · 18 linked inside Pith

  1. [1]

    Mo- tus: A unified latent action world model.arXiv preprint arXiv:2512.13030, 2025

    Hongzhe Bi, Hengkai Tan, Shenghao Xie, Zeyuan Wang, Shuhe Huang, Haitian Liu, Ruowen Zhao, Yao Feng, Chen- dong Xiang, Yinze Rong, Hongyan Zhao, Hanyu Liu, Zhizhong Su, Lei Ma, Hang Su, and Jun Zhu. Mo- tus: A unified latent action world model.arXiv preprint arXiv:2512.13030, 2025

    Pith/arXiv arXiv 2025

  2. [2]

    AHA- W AM: Asynchronous horizon-adaptive world-action model- ing with observation-guided context routing.arXiv preprint arXiv:2606.09811, 2026

    Jisong Cai, Long Ling, Shiwei Chu, Zhongshan Liu, Jiayue Kang, Zhixuan Liang, Wenjie Xu, Yinan Mao, Weinan Zhang, Xiaokang Yang, Ru Ying, Ran Zheng, and Yao Mu. AHA- W AM: Asynchronous horizon-adaptive world-action model- ing with observation-guided context routing.arXiv preprint arXiv:2606.09811, 2026

    Pith/arXiv arXiv 2026

  3. [3]

    RoboTwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic ma- nipulation.arXiv preprint arXiv:2506.18088, 2025

    Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Zixuan Li, Qiwei Liang, Xianliang Lin, Yiheng Ge, Zhenyu Gu, Weiliang Deng, Yubin Guo, Tian Nian, Xuan- bing Xie, Qiangyu Chen, Kailun Su, Tianling Xu, Guodong Liu, Mengkang Hu, Huan-ang Gao, Kaixuan Wang, Zhixuan Liang, Yusen Qin, Xiaokang Yang, Ping Luo, and Yao Mu. RoboTwin 2.0: A scalable...

    Pith/arXiv arXiv 2025

  4. [4]

    RoVer: Robot reward model as test-time verifier for vision- language-action model.arXiv preprint arXiv:2510.10975, 2025

    Mingtong Dai, Lingbo Liu, Yongjie Bai, Yang Liu, Zhouxia Wang, Rui Su, Chunjie Chen, Liang Lin, and Xinyu Wu. RoVer: Robot reward model as test-time verifier for vision- language-action model.arXiv preprint arXiv:2510.10975, 2025

    arXiv 2025

  5. [5]

    Two-frame motion estimation based on polynomial expansion

    Gunnar Farneback. Two-frame motion estimation based on polynomial expansion. InScandinavian Conference on Image Analysis, pages 363–370. Springer, 2003

    2003

  6. [6]

    X-WAM: Unified 4D world action modeling from video priors with asynchronous denoising.arXiv preprint arXiv:2604.26694, 2026

    Jun Guo, Qiwei Li, Peiyan Li, Zilong Chen, Nan Sun, Yifei Su, Heyun Wang, Yuan Zhang, Xinghang Li, and Huap- ing Liu. X-WAM: Unified 4D world action modeling from video priors with asynchronous denoising.arXiv preprint arXiv:2604.26694, 2026

    Pith/arXiv arXiv 2026

  7. [7]

    Verifier-Free Test-Time Sampling for Vision-Language-Action Models

    Suhyeok Jang, Dongyoung Kim, Changyeon Kim, Youngsuk Kim, and Jinwoo Shin. Verifier-Free Test-Time Sampling for Vision-Language-Action Models. InInternational Confer- ence on Learning Representations, 2026

    2026

  8. [8]

    Cosmos Policy: Fine- tuning video models for visuomotor control and planning

    Moo Jin Kim, Yihuai Gao, Tsung-Yi Lin, Yen-Chen Lin, Yunhao Ge, Grace Lam, Percy Liang, Shuran Song, Ming-Yu Liu, Chelsea Finn, and Jinwei Gu. Cosmos Policy: Fine- tuning video models for visuomotor control and planning. arXiv preprint arXiv:2601.16163, 2026

    Pith/arXiv arXiv 2026

  9. [9]

    RoboMonkey: Scaling test-time sampling and verification for vision-language-action models

    Jacky Kwok, Christopher Agia, Rohan Sinha, Matt Foutter, Shulu Li, Ion Stoica, Azalia Mirhoseini, and Marco Pavone. RoboMonkey: Scaling test-time sampling and verification for vision-language-action models. InProceedings of The 9th Conference on Robot Learning, pages 3200–3217. PMLR, 2025

    2025

  10. [10]

    Efficient-WAM: A 1B-parameter world-action model with low-cost future imagi- nation.arXiv preprint arXiv:2606.10040, 2026

    Jiajun Li, Tiecheng Guo, Yifan Ye, Rongyu Zhang, Xiaowei Chi, Qianpu Sun, Ying Li, Yunfan Lou, Yan Huang, Zhihe Lu, Meng Guo, and Shanghang Zhang. Efficient-WAM: A 1B-parameter world-action model with low-cost future imagi- nation.arXiv preprint arXiv:2606.10040, 2026

    Pith/arXiv arXiv 2026

  11. [11]

    Causal World Modeling for Robot Control.arXiv preprint arXiv:2601.21998, 2026

    Lin Li, Qihang Zhang, Yiming Luo, Shuai Yang, Ruilin Wang, Fei Han, Mingrui Yu, Zelin Gao, Nan Xue, Xing Zhu, Yujun Shen, and Yinghao Xu. Causal World Modeling for Robot Control.arXiv preprint arXiv:2601.21998, 2026

    Pith/arXiv arXiv 2026

  12. [12]

    Light-W AM: Efficient world action models with state-fusion action decoding.arXiv preprint arXiv:2606.08242, 2026

    Ziang Li, Dongzhou Cheng, Yibin Wang, Shiyue Wang, Xi- aoyang Xu, Lingxuan Weng, Juan Wang, and Jiaqi Wang. Light-W AM: Efficient world action models with state-fusion action decoding.arXiv preprint arXiv:2606.08242, 2026

    Pith/arXiv arXiv 2026

  13. [13]

    LIBERO: Benchmarking knowl- edge transfer for lifelong robot learning

    Bo Liu, Yifeng Zhu, Chongkai Gao, Yihao Feng, Qiang Liu, Yuke Zhu, and Peter Stone. LIBERO: Benchmarking knowl- edge transfer for lifelong robot learning. InAdvances in Neural Information Processing Systems, 2023

    2023

  14. [14]

    World Action Verifier: Self-improving world models via forward-inverse asymmetry.arXiv preprint arXiv:2604.01985, 2026

    Yuejiang Liu, Fan Feng, Lingjing Kong, Weifeng Lu, Jinzhou Tang, Kun Zhang, Kevin Murphy, Chelsea Finn, and Yilun Du. World Action Verifier: Self-improving world models via forward-inverse asymmetry.arXiv preprint arXiv:2604.01985, 2026

    Pith/arXiv arXiv 2026

  15. [15]

    MotuBrain: An advanced world action model for robot control.arXiv preprint arXiv:2604.27792, 2026

    MotuBrain Team, Chendong Xiang, Fan Bao, Haitian Liu, Hengkai Tan, Hongzhe Bi, James Li, Jiabao Liu, Jingrui Pang, Kiro Jing, Louis Liu, Mengchen Cai, Rongxu Cui, Ruowen Zhao, Runqing Wang, Shuhe Huang, Yao Feng, Yinze Rong, Zeyuan Wang, and Jun Zhu. MotuBrain: An advanced world action model for robot control.arXiv preprint arXiv:2604.27792, 2026

    Pith/arXiv arXiv 2026

  16. [16]

    RoboCasa: Large-scale simulation of everyday tasks for generalist robots.arXiv preprint arXiv:2406.02523, 2024

    Soroush Nasiriany, Abhiram Maddukuri, Lance Zhang, Adeet Parikh, Aaron Lo, Abhishek Joshi, Ajay Mandlekar, and Yuke Zhu. RoboCasa: Large-scale simulation of everyday tasks for generalist robots.arXiv preprint arXiv:2406.02523, 2024

    Pith/arXiv arXiv 2024

  17. [17]

    Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models.arXiv preprint arXiv:2605.07514, 2026

    Bo-Kai Ruan, Teng-Fang Hsiao, Ling Lo, and Hong-Han Shuai. Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models.arXiv preprint arXiv:2605.07514, 2026

    Pith/arXiv arXiv 2026

  18. [18]

    Scaling LLM test-time compute optimally can be more effec- tive than scaling model parameters

    Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. Scaling LLM test-time compute optimally can be more effec- tive than scaling model parameters. InInternational Confer- ence on Learning Representations, 2025

    2025

  19. [19]

    SANTS: A state- adaptive scheduler for world action models.arXiv preprint arXiv:2605.27947, 2026

    Yirui Sun, Guangyu Zhuge, Keliang Liu, Jie Gu, Xinyu Bing, Zhongxue Gan, and Chunxu Tian. SANTS: A state- adaptive scheduler for world action models.arXiv preprint arXiv:2605.27947, 2026

    Pith/arXiv arXiv 2026

  20. [20]

    Jianyuan Wang, Minghao Chen, Shangzhan Zhang, Nikita Karaev, Johannes Sch ¨onberger, Patrick Labatut, Piotr Bo- janowski, David Novotny, Andrea Vedaldi, and Christian Rupprecht. VGGT- Ω. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, 2026

    2026

  21. [21]

    When to Trust Imag- ination: Adaptive action execution for world action models

    Rui Wang, Yue Zhang, Jiehong Lin, Kuncheng Luo, Jianan Wang, Zhongrui Wang, and Xiaojuan Qi. When to Trust Imag- ination: Adaptive action execution for world action models. arXiv preprint arXiv:2605.06222, 2026

    Pith/arXiv arXiv 2026

  22. [22]

    Le, Ed H

    Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V . Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. InInternational Conference on Learning Representations, 2023

    2023

  23. [23]

    Chi, Quoc V

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V . Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models. InAdvances in Neural Information Pro- cessing Systems, 2022. 7

    2022

  24. [24]

    Griffiths, Yuan Cao, and Karthik Narasimhan

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. InAdvances in Neural Information Processing Sys- tems, 2023

    2023

  25. [25]

    World Action Models are Zero-Shot Policies.arXiv preprint arXiv:2602.15922, 2026

    Seonghyeon Ye, Yunhao Ge, Kaiyuan Zheng, Shenyuan Gao, Sihyun Yu, George Kurian, Suneel Indupuru, You Liang Tan, Chuning Zhu, Jiannan Xiang, Ayaan Malik, Kyungmin Lee, William Liang, Nadun Ranawaka, Jiasheng Gu, Yinzhen Xu, Guanzhi Wang, Fengyuan Hu, Avnish Narayan, Johan Bjorck, Jing Wang, Gwanghyun Kim, Dantong Niu, Ruijie Zheng, Yuqi Xie, Jimmy Wu, Qi ...

    Pith/arXiv arXiv 2026

  26. [26]

    Fast-WAM: Do world action models need test-time future imagination?arXiv preprint arXiv:2603.16666, 2026

    Tianyuan Yuan, Zibin Dong, Yicheng Liu, and Hang Zhao. Fast-WAM: Do world action models need test-time future imagination?arXiv preprint arXiv:2603.16666, 2026

    Pith/arXiv arXiv 2026

  27. [27]

    Efros, Eli Shechtman, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep fea- tures as a perceptual metric. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018

    2018

  28. [28]

    C3ache: Accelerating world action models with cross infer- ence chunk cache.arXiv preprint arXiv:2606.08962, 2026

    Weisen Zhao, Lam Nguyen, Zhicong Lu, and Yuzhang Shang. C3ache: Accelerating world action models with cross infer- ence chunk cache.arXiv preprint arXiv:2606.08962, 2026

    Pith/arXiv arXiv 2026

  29. [29]

    τ0-WM: A unified video-action world model for robotic manipulation.arXiv preprint arXiv:2606.01027, 2026

    Pengfei Zhou, Shengcong Chen, Di Chen, Jiaxu Wang, Rongjun Jin, Bingwen Zhu, Yike Pan, Songen Gu, Kuan- ning Wang, Shufeng Nan, Xingyu Qiu, Chenhao Qiu, Pu Yang, Yunuo Cai, Jianxiong Gao, Yifan Li, Yanwei Fu, Xi- angyu Yue, Zhi Chen, and Jianlan Luo. τ0-WM: A unified video-action world model for robotic manipulation.arXiv preprint arXiv:2606.01027, 2026. 8

    Pith/arXiv arXiv 2026