DreamerV3 — one set of hyperparameters works across 150+ tasks, including collecting diamonds in Minecraft from scratch.DreamerV3：一套超参数适用于 150 多个任务，包括从零开始在《我的世界》中挖到钻石。
@misc{hafner2023mastering,
title = {{Mastering Diverse Domains through World Models}},
author = {Danijar Hafner and Jurgis Pasukonis and Jimmy Ba and Timothy Lillicrap},
year = {2023},
eprint = {2301.04104},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2301.04104},
}
Opens a 50,000-hour annotated egocentric video corpus with pipelines and checkpoints. Reports a log-linear law between pretraining hours and robot success, and 84.6% on RoboTwin 2.0 Clean2Random.开源 5 万小时带标注的第一视角视频语料及流程和权重。报告预训练时长与机器人成功率之间的对数线性规律，后训练后在 RoboTwin 2.0 Clean2Random 上达到 84.6%。
@misc{ma2026open,
title = {{Open-HumanWAM: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models}},
author = {Juncheng Ma and Jiaxin Li and Jiankai Tu and Bo Liang and Guodong Lv and Yufan Deng and Junzhe Tang and Liangzhu Yang and Chen Bian and Zicong Cao and Baohao Zhu and Jianxin Bi and Tengfei Yan and Yinyuan Zhang and Kewei Zhang and Xinglong Chen and Kaiqi Chen and Duomin Wang and Xiaotian Tang and Liang Tan and Li Li and Yuan Xu and Haosen Yang and Yuqi Wang and Jingfei Xia and Songyan Huang and Zhiyang Dou and Zhen Dong and Enze Xie and Tat-Seng Chua and Daquan Zhou},
year = {2026},
eprint = {2611.06817},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.06817},
}
Separates WAMs from world models, video generators and VLAs, then sorts work by what is generated (rendered, latent or none) and how it is built; concludes the field is moving toward generating less of the future.厘清 WAM 与世界模型、视频生成模型、VLA 的区别，并按生成内容（渲染、潜空间或不生成）和架构分类；结论是领域正走向少生成未来，只保留控制所需。
@misc{shen2026world,
title = {{World Action Models: A Survey}},
author = {Qiuhong Shen and Shihua Zhang and Yue Liao and Qi Li and Zhenxiong Tan and Shizun Wang and Shuicheng Yan and Xinchao Wang},
year = {2026},
eprint = {2606.20781},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.20781},
}
A family of omnimodal models that handles language, image, video, audio and action in one mixture-of-transformers, folding VLMs, video generators, world simulators and world-action models into a single framework.全模态世界模型家族，用一个 Mixture-of-Transformers 同时处理语言、图像、视频、音频和动作，把 VLM、视频生成、世界模拟器和世界动作模型并入同一框架。
@misc{nvidia2026cosmos,
title = {{Cosmos 3: Omnimodal World Models for Physical AI}},
author = {NVIDIA and : and Aditi and Niket Agarwal and Arslan Ali and Jon Allen and Martin Antolini and Adeline Aubame and Alisson Azzolini and Junjie Bai and Maciej Bala and Yogesh Balaji and Josh Bapst and Aarti Basant and Mukesh Beladiya and Mohammad Qazim Bhat and Zaid Pervaiz Bhat and Dan Blick and Vanni Brighella and Han Cai and Tiffany Cai and Eric Cameracci and Jiaxin Cao and Yulong Cao and Mark Carlson and Carlos Casanova and Ting-Yun Chang and Yan Chang and Yu-Wei Chao and Prithvijit Chattopadhyay and Roshan Chaudhari and Chieh-Yun Chen and Junyu Chen and Ke Chen and Qizhi Chen and Wenkai Chen and Xiaotong Chen and Yu Chen and An-Chieh Cheng and Click Cheng and Xiu Chia and Jeana Choi and Chaeyeon Chung and Wenyan Cong and Yin Cui and Magdalena Dadela and Nalin Dadhich and Wenliang Dai and Joyjit Daw and Alperen Degirmenci and others},
year = {2026},
eprint = {2606.02800},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.02800},
}
A foundation world model pretrained on 44k hours of egocentric human video, using continuous latent actions as proxies for missing labels, then post-trained on small robot datasets.在 4.4 万小时第一视角人类视频上预训练的基础世界模型，用连续潜动作代替缺失的动作标签，再用小规模机器人数据后训练。
@misc{gao2026dreamdojo,
title = {{DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos}},
author = {Shenyuan Gao and William Liang and Kaiyuan Zheng and Ayaan Malik and Seonghyeon Ye and Sihyun Yu and Wei-Cheng Tseng and Yuzhu Dong and Kaichun Mo and Chen-Hsuan Lin and Qianli Ma and Seungjun Nah and Loic Magne and Jiannan Xiang and Yuqi Xie and Ruijie Zheng and Dantong Niu and You Liang Tan and K. R. Zentner and George Kurian and Suneel Indupuru and Pooya Jannaty and Jinwei Gu and Jun Zhang and Jitendra Malik and Pieter Abbeel and Ming-Yu Liu and Yuke Zhu and Joel Jang and Linxi "Jim" Fan},
year = {2026},
eprint = {2602.06949},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.06949},
}
World Models世界模型Dexterous hand灵巧手human-videodexterouslatent-action
An open research stack that splits world-action models into composable modules, then runs controlled studies on what to inherit from video priors and how world and action learning interact.开放研究栈，把世界动作模型拆成可组合模块，并通过对照实验研究该从视频先验继承什么、世界学习与动作学习如何相互作用。
@misc{wang2026openwam,
title = {{OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining}},
author = {Yuran Wang and Siqiao Huang and Mingleyang Li and Chenhao Zhang and Jiaqi Liang and Weiyang Jin and Yue Chen and Xuemin Chi and Donghao Zhou and Qize Yu and Yu-Kai Wang and Yuhan Rui and Shenzhe Yao and Zhen Yuan and Zhenhao Shen and Kefei Zhu and Zijie Zhu and Ning Gao and Xiaowei Chi and Guanqi He and Shanghang Zhang and Hao Dong and Lin Shao and Hang Zhao},
year = {2026},
eprint = {2609.07398},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.07398},
}
A JEPA world model trained across 12 robot embodiments; its imagination error follows a power law in compute, so model quality can be predicted beyond the fitted scale.跨 12 种机器人形态训练的 JEPA 世界模型；其想象误差随算力服从幂律，因此可以预测超出拟合规模时的模型质量。
@misc{zholus2026robojepa,
title = {{RoboJEPA: Scaling Robotic Latent World Models}},
author = {Artem Zholus and Nicolas Beltran-Velez and Jianhao Yuan and Sarath Chandar and Tushar Nagarajan and Daniel Severo and Koustuv Sinha and Michal Drozdzal and Adriana Romero Soriano and Jeannette Bohg and Nicolas Ballas and Mahmoud Assran},
year = {2026},
eprint = {2610.10515},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2610.10515},
}
Post-trains a VLA (GigaBrain-0.5, 10k+ h of robot data) with RL conditioned on a video world model's predictions (RAMP); reports ~30% over the RECAP baseline on laundry folding, box packing and espresso making.在 GigaBrain-0.5（预训练于 1 万多小时机器人操作数据）基础上，用以视频世界模型预测为条件的强化学习（RAMP）做后训练；论文报告在叠衣服、装箱和制作浓缩咖啡等难任务上比 RECAP 基线提升约 30%。
@misc{team2026gigabrain,
title = {{GigaBrain-0.5M*: a VLA That Learns From World Model-Based Reinforcement Learning}},
author = {GigaBrain Team and Boyuan Wang and Bohan Li and Chaojun Ni and Guan Huang and Guosheng Zhao and Hao Li and Jie Li and Jindi Lv and Jingyu Liu and Lv Feng and Mingming Yu and Peng Li and Qiuping Deng and Tianze Liu and Xinyu Zhou and Xinze Chen and Xiaofeng Wang and Yang Wang and Yifan Li and Yifei Nie and Yilong Li and Yukun Zhou and Yun Ye and Zhichao Liu and Zheng Zhu},
year = {2026},
eprint = {2602.12099},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2602.12099},
}
VLA that reasons in discrete motion tokens (a chain of dynamics) shared across embodiments; predicted futures are scored against what actually happened, so action-free human videos give RL signal. Sim and real tests.VLA 用跨本体共享的离散运动 token 链（Chain-of-Dynamics）表达动作思考；预测每条链导致的未来并与真实结果比对来强化，使没有动作标注的人类视频也能提供监督。在仿真与真机上验证人到机器人的迁移。
@misc{kim2026thinkworld,
title = {{ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning}},
author = {Chaehyun Kim and Chi-Pin Huang and Jisoo Kim and Min-Hung Chen and Yu-Chiang Frank Wang and Seungryong Kim and Fu-En Yang},
year = {2026},
eprint = {2610.37666},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37666},
}
@misc{wang2026pdrvla,
title = {{PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations}},
author = {Hanqing Wang and Zhaotian Cai and Ruichuan An and Bohan Zeng and Tao Zhang and Yilin Chen and Xu Jiang and Kaiquan Chen and Chongyang Wu and Haoran Li and Fan Yang and Mingan Lin and Wentao Zhang and Zenan Zhou and Youzheng Wu},
year = {2026},
eprint = {2610.37381},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37381},
}
Rather than distilling per-motion experts or brute-forcing one RL policy, trains a humanoid tracking policy jointly with one differentiable world model of shared dynamics, fit from simulator logs (CoRL 2026 oral).SuperMimic 不再蒸馏逐动作专家或用大算力硬训单一 RL 策略，而是把人形机器人动作跟踪策略与一个从仿真日志拟合的、共享动力学的可微世界模型联合训练（CoRL 2026 口头报告）。
@misc{liao2026supermimic,
title = {{SuperMimic: Scaling Humanoid Whole-Body Control with One Differentiable World Model}},
author = {Qiayuan Liao and Yiyang Shao and Junfeng Long and Kevin Zakka and Roberto Horowitz and Pieter Abbeel and Koushil Sreenath},
year = {2026},
eprint = {2610.28480},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.28480},
}
An egocentric world model covering both walking and manipulating, trained on ~10k hours of human video plus ~3k hours of robot data under one shared kinematic action representation.同时覆盖移动与操作的第一人称世界模型，在约 1 万小时人类视频和约 3 千小时机器人数据上训练，用统一的运动学表示描述不同本体的动作。
@misc{liu2026ropedia,
title = {{Ropedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human Experience}},
author = {Tianqi Liu and Hanfei Zhao and Qifeng Liang and Hui En Pang and Jiarui Xu and Sayang Mu and Penghao Qiu and YingSheng Zhang and Jun Wang and Ziang Cao and Mingze Sun and Yukang Cao and Fangzhou Hong and Zhaoxi Chen and Ziwei Liu},
year = {2026},
eprint = {2610.26855},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.26855},
}
World Models世界模型egocentrichuman-datacross-embodimentloco-manipulation
Paper论文 2026 · Sber Robotics Center ★new新 2026-10-23
One world model that is also the policy, across 20 robot families mapped to a shared joint-space action format; reports strong WorldArena, RoboTwin 2.0 and LIBERO-plus results and a one-step distilled humanoid policy.同一个世界模型兼作策略，用共享的关节坐标动作空间覆盖 20 类机器人；论文报告其在 WorldArena、RoboTwin 2.0、LIBERO-plus 上表现突出，并蒸馏出可实时运行的单步人形机器人策略。
@misc{sidorov2026greenwam,
title = {{GreenWAM: A World Foundation Model for Robotics}},
author = {Gennady Sidorov and Valerii Kornilov and Denis Grankin and Polina Fedotova and Aleksey Postnikov},
year = {2026},
eprint = {2610.18774},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18774},
}
Rewinds an action-conditioned world model to key decision points to try alternative actions, turning outcomes into action-level supervision; real data keeps correcting the model. Reports +10.7–11.9 pts on 3 real tasks.把动作条件世界模型当作反事实引擎：回到关键决策点尝试其他动作并对比结果，得到动作级监督而无需单独的价值模型；真机经验再通过 LoRA 持续修正世界模型。报告在 3 个真实任务上比最强基线高 10.7–11.9 个百分点。
@misc{luo2026rollback,
title = {{Rollback: Experience Backtracking in World Models for Policy Improvement}},
author = {Jiayi Luo and Yang Zhang and Liliang Chen and Nan Wang and Chen Gao and Boxiang Qiu and Zhuoxuan Li and Baorui Peng and Chuankang Li and Yuxiang Yan and Si Liu and Qingyun Sun and Shuicheng Yan and Maoqing Yao and Jianxin Li and Guanghui Ren},
year = {2026},
eprint = {2610.36068},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36068},
}
CoRL 2026 Oral. DreamZero: a World Action Model on a pretrained video diffusion backbone, jointly predicting video and actions, so it learns skills from heterogeneous robot data without repetitive demonstrations.CoRL 2026 Oral。DreamZero：基于预训练视频扩散骨干的世界动作模型，联合预测视频与动作，可从异构机器人数据学习技能，无需重复示范。
@misc{ye2026world,
title = {{World Action Models are Zero-shot Policies}},
author = {Seonghyeon Ye and Yunhao Ge and Kaiyuan Zheng and Shenyuan Gao and Sihyun Yu and George Kurian and Suneel Indupuru and You Liang Tan and Chuning Zhu and Jiannan Xiang and Ayaan Malik and Kyungmin Lee and William Liang and Nadun Ranawaka and Jiasheng Gu and Yinzhen Xu and Guanzhi Wang and Fengyuan Hu and Avnish Narayan and Johan Bjorck and Jing Wang and Gwanghyun Kim and Dantong Niu and Ruijie Zheng and Yuqi Xie and Jimmy Wu and Qi Wang and Ryan Julian and Danfei Xu and Yilun Du and Yevgen Chebotar and Scott Reed and Jan Kautz and Yuke Zhu and Linxi "Jim" Fan and Joel Jang},
year = {2026},
eprint = {2602.15922},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.15922},
}
CoRL 2026 Spotlight. Model-based RL that uses a chunk world model as a differentiable training-time evaluator, targeting compounding error, inference latency and BPTT instability.CoRL 2026 Spotlight。将分块世界模型用作训练期可微评估器的模型式强化学习，针对累积误差、推理延迟与 BPTT 不稳定问题。
CoRL 2026 Spotlight. Shows world models are a stealthy data-poisoning entry point: compromised policies can be deployed even when the ground-truth training data looks safe.CoRL 2026 Spotlight。展示世界模型可成为隐蔽的数据投毒入口：即使真实训练数据看似安全，部署的策略仍可能被破坏。
@misc{rathbun2026targeting,
title = {{Targeting World Models to Compromise Robot Learning Pipelines}},
author = {Ethan Rathbun and Ahmed Agha and Saaduddin Mahmud and Christopher Amato and Alina Oprea and Eugene Bagdasarian},
year = {2026},
eprint = {2606.09499},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.09499},
}
Reuses a pretrained video diffusion model to extract compact latent actions (1,200 to 4 tokens per frame) that act as a shared interface across humans, robots and environments, including zero-shot action transfer.复用预训练视频扩散模型提取紧凑潜动作（每帧 1,200 压到 4 个 token），作为人、机器人和不同环境之间的共享接口，并支持零示教动作迁移。
@misc{kim2026dreamx,
title = {{DreamX: Video Diffusion Model is a Universal Latent Action Learner}},
author = {Jisoo Kim and Chaehyun Kim and Junsu Kim and Jiho Park and Seongchan Kim and Eunbeen Hong and Soowon Son and Fu-En Yang and Seungryong Kim},
year = {2026},
eprint = {2611.05377},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.05377},
}
LingBot-VA: an autoregressive diffusion framework that learns frame prediction and policy together, with shared video-action latents, closed-loop rollout and asynchronous inference.LingBot-VA：同时学习帧预测和策略的自回归扩散框架，具有视频动作共享潜空间、闭环推演和异步推理。
@misc{li2026causal,
title = {{Causal World Modeling for Robot Control}},
author = {Lin Li and Qihang Zhang and Yiming Luo and Shuai Yang and Ruilin Wang and Fei Han and Mingrui Yu and Zelin Gao and Nan Xue and Xing Zhu and Yujun Shen and Yinghao Xu},
year = {2026},
eprint = {2601.21998},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2601.21998},
}
Asks whether WAMs need future imagination at test time: keeps video co-training during training but skips video generation at inference, separating the two effects in a controlled comparison.追问世界动作模型在测试时是否需要想象未来：训练时保留视频联合训练，推理时跳过视频生成，以对照实验区分两者的作用。
@misc{yuan2026fast,
title = {{Fast-WAM: Do World Action Models Need Test-time Future Imagination?}},
author = {Tianyuan Yuan and Zibin Dong and Yicheng Liu and Hang Zhao},
year = {2026},
eprint = {2603.16666},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.16666},
}
An action-centered WAM that predicts actions from the observation and makes video generation optional, so motion accuracy depends less on forecast quality and inference is cheaper.以动作为中心的世界动作模型：由观测预测动作，视频生成变为可选，动作精度不再依赖视频预测质量，推理开销也更低。
@misc{ye2026gigaworld,
title = {{GigaWorld-Policy: An Efficient Action-Centered World--Action Model}},
author = {Angen Ye and Boyuan Wang and Chaojun Ni and Guan Huang and Guosheng Zhao and Hao Li and Hengtao Li and Jie Li and Jindi Lv and Jingyu Liu and Min Cao and Peng Li and Qiuping Deng and Wenjun Mei and Xiaofeng Wang and Xinze Chen and Xinyu Zhou and Yang Wang and Yifan Chang and Yifan Li and Yukun Zhou and Yun Ye and Zhichao Liu and Zheng Zhu},
year = {2026},
eprint = {2603.17240},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.17240},
}
Uses semantically coherent action events rather than fixed-length chunks as the unit of learning, to resolve the timescale mismatch between language, vision and control.用语义完整的动作事件代替固定长度动作块作为学习单元，以解决语言、视觉和控制之间的时间尺度不匹配。
@misc{li2026wall,
title = {{WALL-WM: Carving World Action Modeling at the Event Joints}},
author = {Shalfun Li and Victor Yao and Charles Yang and Truth Qu and Regis Cheng and Ryan Yu and Howard Lu and Newton Von and Vincent Chen and Yohann Tang and Maeve Zhang and Ellie Ma and Gody Li and Starrick Liu and Sage Yang and Lorien Shu and J. W. Gao and Ethan Chen and Colin Ye and Yu Sun and Elise Mon and PS Zhang and Neo Li and Lily Li and James Wang and Ping Yang and Chris Pan and Lucy Liang and Hang Su and Roy Gan and Hao Wang and Qian Wang},
year = {2026},
eprint = {2606.01955},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.01955},
}
Adapts Cosmos models for real-time interactive simulation: streams observations in response to actions, or renders them from an external simulator's state. Shown on manipulation, video transfer and multi-view AV.将 Cosmos 模型用于实时交互仿真的统一框架：根据动作流式生成观测，或依据外部仿真器的物理状态生成观测。示例涵盖操作、可控视频迁移和多视角自动驾驶。
@misc{nvidia2026interactive,
title = {{Interactive World Simulation with Cosmos for Physical AI}},
author = {NVIDIA and : and Anis Ahmad and Jon Allen and Malik Aqeel Anwar and Jesse Archer and Maciej Bala and Piotr Barejko and Aarti Basant and Dan Blick and Taras Bolzhelarskyi and Henry Che and Jin Chen and Yukang Chen and Tae Eun Choe and Yuzhu Dong and Nat Duca and Liang Feng and Francesco Ferroni and Xiao Fu and Mohit Gambhir and Vivek Goel and Rony Goldenthal and Aryaman Gupta and Ali Hassani and Nathan Hayes-Roth and Yong He and Michael Honke and Shengyu Huang and Wei Huang and DeLesley Hutchins and Suneel Indupuru and Pooya Jannaty and Jingyi Jin and Janmesh Kalra and Ashna Khetan and Xin Kong and Aleksandr Korovko and Hui Kuang and Zhengfei Kuang and George Kurian and Jean-Francois Lafleche and Jayjun Lee and Justin Leung and William Lew and Xiang Li and Zhaoshuo Li and Chen-Hsuan Lin and Ming-Yu Liu and Xiangyu Lu and others},
year = {2026},
eprint = {2611.06627},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2611.06627},
}
One autoregressive model that is both a VLA policy and an image-predicting world model; the two help each other on LIBERO, and an attention mask hiding earlier actions curbs error build-up in action chunks.单个自回归模型同时充当 VLA 策略与预测未来图像的世界模型，二者在 LIBERO 上相互促进；论文还发现自回归生成动作块会累积误差，用屏蔽先前动作的注意力掩码可以缓解。
@misc{cen2025worldvla,
title = {{WorldVLA: Towards Autoregressive Action World Model}},
author = {Jun Cen and Chaohui Yu and Hangjie Yuan and Yuming Jiang and Siteng Huang and Jiayan Guo and Xin Li and Yibing Song and Hao Luo and Fan Wang and Deli Zhao and Hao Chen},
year = {2025},
eprint = {2506.21539},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.21539},
}
A flow-based world foundation model unifying Text2World, Image2World and Video2World, released at 2B and 14B, with Cosmos-Transfer2.5 for Sim2Real translation.基于流的世界基础模型，统一 Text2World、Image2World 和 Video2World，发布 2B 和 14B 两种规模，并配有用于 Sim2Real 转换的 Cosmos-Transfer2.5。
@misc{nvidia2025world,
title = {{World Simulation with Video Foundation Models for Physical AI}},
author = {NVIDIA and : and Arslan Ali and Junjie Bai and Maciej Bala and Yogesh Balaji and Aaron Blakeman and Tiffany Cai and Jiaxin Cao and Tianshi Cao and Elizabeth Cha and Yu-Wei Chao and Prithvijit Chattopadhyay and Mike Chen and Yongxin Chen and Yu Chen and Shuai Cheng and Yin Cui and Jenna Diamond and Yifan Ding and Jiaojiao Fan and Linxi Fan and Liang Feng and Francesco Ferroni and Sanja Fidler and Xiao Fu and Ruiyuan Gao and Yunhao Ge and Jinwei Gu and Aryaman Gupta and Siddharth Gururani and Imad El Hanafi and Ali Hassani and Zekun Hao and Jacob Huffman and Joel Jang and Pooya Jannaty and Jan Kautz and Grace Lam and Xuan Li and Zhaoshuo Li and Maosheng Liao and Chen-Hsuan Lin and Tsung-Yi Lin and Yen-Chen Lin and Huan Ling and Ming-Yu Liu and Xian Liu and Yifan Lu and Alice Luo and others},
year = {2025},
eprint = {2511.00062},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2511.00062},
}
Diffuser — treats planning as denoising whole trajectories, so goals and rewards can steer sampling; an early link between diffusion and control.Diffuser：把规划视为对整条轨迹去噪，目标与奖励可以引导采样；扩散模型与控制结合的早期代表作。
@misc{janner2022planning,
title = {{Planning with Diffusion for Flexible Behavior Synthesis}},
author = {Michael Janner and Yilun Du and Joshua B. Tenenbaum and Sergey Levine},
year = {2022},
eprint = {2205.09991},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2205.09991},
}
@misc{hafner2020mastering,
title = {{Mastering Atari with Discrete World Models}},
author = {Danijar Hafner and Timothy Lillicrap and Mohammad Norouzi and Jimmy Ba},
year = {2020},
eprint = {2010.02193},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2010.02193},
}
PlaNet — learns a latent dynamics model (the RSSM later reused by Dreamer) from images and plans actions directly in latent space with CEM.PlaNet：从图像中学习潜空间动力学模型（即后来 Dreamer 沿用的 RSSM），并用 CEM 直接在潜空间中规划动作。
@misc{hafner2018learning,
title = {{Learning Latent Dynamics for Planning from Pixels}},
author = {Danijar Hafner and Timothy Lillicrap and Ian Fischer and Ruben Villegas and David Ha and Honglak Lee and James Davidson},
year = {2018},
eprint = {1811.04551},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/1811.04551},
}
Adds a 257M latent world model to a pretrained VLA so predicted future features condition the action expert, with no pixel generation or video pretraining; reports 98.7% on LIBERO and real-robot gains over π0.5.在预训练 VLA 上加一个 2.57 亿参数的潜空间世界模型，用预测的未来表征去条件化动作专家，无需像素生成或大规模视频预训练；报告 LIBERO 成功率 98.7%，真机任务上明显优于 π0.5。
@misc{wang2026wave,
title = {{WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction}},
author = {Haomin Wang and Anqiong Ji and Ke Wang and Zhongshan Liu and Ru Ying and Ran Zheng and Zhipeng Zhang},
year = {2026},
eprint = {2610.18568},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.18568},
}
Between direct action prediction and full video prediction: VLDA learns a compact latent 'Delta' of task-relevant state change and uses it to seed action denoising. Built on Baidu's open LoongForge.介于直接预测动作与完整预测未来画面之间：VLDA 学习一个紧凑的隐变量 Delta 表示与任务相关的状态变化，并用它引导动作去噪；实现于百度开源训练框架 LoongForge。
@misc{zhang2026vision,
title = {{Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation}},
author = {Lingsen Zhang and Wei Li and Jie He and Henghua Zhang and Kai Mo and Shiyong Li and Yanpeng Wang and Rui Shao},
year = {2026},
eprint = {2610.25874},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25874},
}
Splits the VLA vs WAM debate into backbone, future loss and inference machinery: a foresight loss predicts frozen future features, no video generation. Reports beating FastWAM on LIBERO-Plus (64.5% vs 55.3%).把 VLA 与世界动作模型（WAM）之争拆成骨干初始化、未来预测损失和推理时是否保留预测模块三个选择：用预测冻结未来特征的前瞻损失，训练与推理都不生成视频；报告在 LIBERO-Plus 上 64.5%，FastWAM 为 55.3%。
@misc{yu2026vision,
title = {{Vision-Language-Action or World-Action? Just Action Models}},
author = {Yonghao Yu and Lang Huang and Runyi Li and Toshihiko Yamasaki},
year = {2026},
eprint = {2610.37219},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.37219},
}
Rates VLA and world-action models on five axes (capability, training and deployment cost, verified research value, reproducibility), not success rate alone; withholds overall scores when evidence is missing.多维度评测 VLA 与世界-动作模型（WAM）：除任务能力外，还衡量下游训练成本、部署成本、经本地验证的研究价值和复现证据，按不同算力预算给出 0–100 画像；证据缺失时明确标注、不给总分，并开源评分代码与机器可读证据。
@misc{ma2026roboseed,
title = {{RoboSeed: Multidimensional Evaluation of Vision-Language-Action and World-Action Models: Capability, Resource Affordability, and Research Evidence}},
author = {Xin Ma and Furong Chen and Xinlong Man and Bo Zhang and Yalong Guo and Haolan Lei and Chaochen Cui and Hongyuan Zhang and Mulin Chen and Yi Wang and Lap-pui Chau and Chuang Yang},
year = {2026},
eprint = {2610.40026},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.40026},
}
JEPA-style VLA pretraining: future frames are only a latent target and the student sees the current observation alone, so it avoids leakage and learns dynamics rather than pixel changes; then fine-tune an action head.JEPA 式 VLA 预训练：未来帧只作为潜空间监督目标，学生网络只看当前观测，避免信息泄漏，学动力学而非像素变化；之后再微调动作头。
@misc{sun2026vla,
title = {{VLA-JEPA: Enhancing Vision-Language-Action Model with Latent World Model}},
author = {Jingwen Sun and Wenyao Zhang and Zekun Qi and Shaojie Ren and Zezhi Liu and Hanxin Zhu and Guangzhong Sun and Xin Jin and Zhibo Chen},
year = {2026},
eprint = {2602.10098},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.10098},
}
World-language-action model: one autoregressive backbone predicts text subtasks, subgoal images and actions; world prediction can be off at inference or on for test-time scaling. Reports 92.94% on RoboTwin 2.0 Clean.世界-语言-动作模型：单一自回归主干同时预测文本子任务、子目标图像与动作；推理时可关闭世界预测，也可开启做测试时扩展。报告在 RoboTwin 2.0 Clean 上成功率 92.94%。
@misc{yang2026world,
title = {{World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis}},
author = {Yi Yang and Zhihong Liu and Siqi Kou and Yiyang Chen and Yanzhe Hu and Jianbo Zhou and Boyuan Zhao and Zhijie Wei and Xiao Xia and Xueqi Li and Pengfei Liu and Zhijie Deng},
year = {2026},
eprint = {2606.05979},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.05979},
}
A teacher sees future force/torque and optical flow; its features are distilled into a student using only current observations, so the policy anticipates contact. Tested on five real contact-rich tasks.教师网络看到未来的力/力矩与光流，再把动力学特征蒸馏给只用当前观测的学生网络，使策略在部署时能预判接触。在五个真实接触丰富任务上测试。
@misc{tang2026forceldm,
title = {{ForceLDM: A Force-aware Latent Dynamics Model for Contact-Rich Manipulation}},
author = {Peiyuan Tang and Yao Li and Xiaodong Zhang and Wuyang Zhang and Qin Xia and Yanyong Zhang and Zijiang Yang},
year = {2026},
eprint = {2611.03294},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03294},
}
5B world action model pretrained on 605 h of RGB-D UMI data that predicts RGB, metric depth and actions. Adds RoboTwin-UMI; reports 85% average real success, +24.2 points over the best baseline.在 605 小时 RGB-D UMI 数据上预训练的 5B 世界动作模型，同时预测 RGB、度量深度和动作。并提出 RoboTwin-UMI 基准；报告真机平均成功率 85%，比最强基线高 24.2 个百分点。
@misc{yuan2026spatial,
title = {{Spatial-UMI-Dream: Scaling World Action Models with Metric Geometry for UMI Manipulation}},
author = {Weiduo Yuan and Ze Yuan and Sihan Liu and Jionghao Wang and Liang Heng and Zhenyu Zhao and Yue Wang},
year = {2026},
eprint = {2611.07229},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07229},
}
Slow system plans sparse long-horizon intent, fast system refines dense motion, linked by a latent Schrödinger bridge. Reports 45.8% lower latency than Cosmos-Policy and 70.8% success on six bimanual tasks.慢系统生成稀疏的长时程意图，快系统细化高频动作，二者以潜空间薛定谔桥耦合。报告延迟比 Cosmos-Policy 低 45.8%，六个双臂灵巧任务成功率 70.8%。
@misc{chen2026lima,
title = {{LiMA: Bridging Long-term Imagination to Real-time Dexterous Manipulation via Asynchronous Diffusion}},
author = {Ning Chen and Yankai Fu and Junkai Zhao and Qianpu Sun and Guocai Yao and Pengwei Wang and Zhongyuan Wang and Shanghang Zhang},
year = {2026},
eprint = {2609.28431},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.28431},
}
Trains a world model of 3D hand-point and scene motion, then aligns policy latents to it; at test time only the RGB policy runs, with no depth or future frames. Reports 49.4% → 60.5% on 11 DexJoCo tasks.先训练预测 3D 手部点与场景变化的世界模型，再把策略潜表示对齐到它；部署时只保留 RGB 策略，无需深度或未来预测。在 11 个 DexJoCo 任务上成功率从 49.4% 提升到 60.5%。
@misc{chen2026dexforesight,
title = {{DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight}},
author = {Zhongxi Chen and Yi Yang and Xinyi Chen and Mingzhe Ye and Yaping Li and Zhihong Liu and Congsheng Xu and Haonan Wang and Zhijie Deng},
year = {2026},
eprint = {2611.11166},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11166},
}
Visuo-tactile world model trained only on humans handling GelSight sensors (4.9 h React dataset). Reports that contact dynamics transfer to robot rollouts more reliably than visual appearance does.只用人手持 GelSight 传感器的交互数据（4.9 小时 React 数据集）训练视触觉世界模型。报告在机器人回放中接触动态比视觉外观更容易迁移，部分任务的相机视角预测甚至不如直接复制帧。
@misc{ma2026acts,
title = {{ACTS: Learning Visuo-Tactile World Models from Human Physical Interaction}},
author = {Yuxiang Ma and Weihan Xu and Pengfei Ye and Anna Mei and Yuxin Ray Song and Tingle Li and Haonan Chen and Edward H. Adelson},
year = {2026},
eprint = {2610.24285},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.24285},
}
128 simulated tactile tasks in three regimes (Precise, Recognition, Dynamic) with 12 perturbations, plus a tactile world model that treats touch as an evolving state and refines actions with predicted contact.128 个仿真触觉任务，分精确、识别、动态三类，含 12 种扰动；并提出动作条件的触觉世界模型，把触觉当作随时间演化的状态，用预测的触觉信息修正动作。
@misc{shi2026tacmind,
title = {{TacMind: Benchmarking and Modeling Tactile Interaction States for Robotic Manipulation}},
author = {Jiaxin Shi and Zibin Wu and Yichi Zhang and Jianghua Duan and Siyu Zhu and Weihao Yuan},
year = {2026},
eprint = {2611.00455},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.00455},
}
Learns visuo-tactile latents end-to-end inside a diffusion world-action model not frozen encoders; multi-task cross-modal losses avoid feature collapse. Reports up to 50% gains on real bolt screwing and nut mating.在扩散式世界-动作模型中端到端学习视触觉潜表示，而非使用冻结的预训练编码器；借助跨模态多任务学习避免特征坍缩。在真实拧螺栓、螺母配合等任务上报告最高 50% 的提升。
A survey of video world models in robotics - where they replace or complement physics simulators (e.g. deformable objects), their applications, and open challenges - with a maintained paper list.机器人视频世界模型综述：它们在哪些场景可替代或补充物理仿真器（如可变形物体）、主要应用与开放挑战，并附持续更新的论文列表。
@misc{mei2026robotic,
title = {{Robotic Video World Models: A Survey of Applications, Research Challenges, Future Directions}},
author = {Zhiting Mei and Tenny Yin and Ola Shorinwa and Apurva Badithela and Zhonghe Zheng and Joseph Bruno and Madison Bland and Lihan Zha and Asher Hancock and Jaime Fernández Fisac and Philip Dames and Anirudha Majumdar},
year = {2026},
eprint = {2601.07823},
archivePrefix = {arXiv},
primaryClass = {eess.SY},
url = {https://arxiv.org/abs/2601.07823},
}
World action model that jointly predicts video, motion and contact wrench, outputting time-aligned motion–wrench chunks as control references (hybrid motion-force style) for force-sensitive manipulation.一个同时预测未来视频、运动与接触力旋量的世界动作模型，借鉴力位混合控制思路，输出时间对齐的运动-力旋量片段作为控制参考，面向仅靠视觉无法感知接触与负载的力敏感操作任务。
@misc{ding2026cw,
title = {{CW-WAM: A Structured Contact-Wrench World Action Model for Force-Sensitive Manipulation}},
author = {Houzhu Ding and Xinda Qi and Luqi Wang and Ming Yang},
year = {2026},
eprint = {2610.30117},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30117},
}
World action model that brings scene-level 3D flow (tracked 3D points) inside the policy rather than as a separate module or extra head, so 4D geometry and visual semantics are modelled together with actions.世界动作模型：把场景级三维流（跟踪的三维点）纳入策略内部，而不是作为独立模块或额外输出头，让 4D 几何、视觉语义与动作一起建模。
@misc{tian2026uniflowwam,
title = {{UniFlowWAM: A Unified World-Action Model with Visual Semantics and 4D Geometry}},
author = {Jialin Tian and Zhuofan Li and Xinkai Wang and Bokai Lin and Chenyi Wang and Xinyu Zhan and Cewu Lu and Lixin Yang},
year = {2026},
eprint = {2610.30991},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30991},
}
FRED trains a world action model to predict better one-step future video features via self-distillation plus history context; reports large gains on long-horizon, out-of-domain and human-video-only tasks.FRED 通过自蒸馏并引入历史观测，让世界动作模型在单步去噪中预测出更好的未来视频表征；报告在长时程、域外环境以及只有人类视频示范的任务上提升明显。
@misc{jang2026self,
title = {{Self-Distilling Future Video Representations for World Action Models}},
author = {Huiwon Jang and Suhyeok Jang and Hyeonmo Kang and Seungku Kim and Jinwoo Shin and Sihyun Yu},
year = {2026},
eprint = {2610.30320},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30320},
}
World Models世界模型world-action-modelvideo-diffusionself-distillationhuman-to-robotreal
AW-0 jointly generates code, future video and actions in one mixture-of-transformers; code structures planning and becomes reusable skills. Reports gains over VLA and WAM baselines on RoboTwin 2.0 and real two-arm tasks.AW-0 在同一个混合 Transformer 中联合生成代码、未来视频和机器人动作三路输出：代码既组织多步预测与控制，又沉淀为可复用技能库；报告在 RoboTwin 2.0 和真机双臂任务上优于 VLA 与 WAM 基线。
@misc{liu2026agentic,
title = {{Agentic World Action Models}},
author = {Zhuoyang Liu and Dantong Niu and Botong Zhang and Boning Shao and Jing Wang and Ruijie Zheng and Yuvan Sharma and Runyu Lu and Jiaming Liu and Renrui Zhang and Roei Herzig and Yutong Bai and Ken Goldberg and Danfei Xu and Yuke Zhu and Guanzhi Wang and Linxi "Jim" Fan and Trevor Darrell},
year = {2026},
eprint = {2610.30136},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.30136},
}
SkeLAM splits latent actions into a robot-skeleton-grounded kinematic part and a visual residual, so world-model rollouts follow robot controls; on DROID reports 19.5% lower end-effector error than DreamDojo.SkeLAM 把潜在动作拆成由机器人骨架监督的运动学部分和保留其余视觉动态的残差部分，使世界模型生成更贴合机器人控制指令；报告在 DROID 上末端轨迹误差比 DreamDojo 低 19.5%，预测成功率与真机评测一致。
@misc{lv2026skeleton,
title = {{Skeleton-Grounded Latent Action Model for Robot World Models}},
author = {Qi Lv and Jiayi Lin and Yuxin Jiang and Jianzong Wang and Xiang Deng and Mike Shou Zheng},
year = {2026},
eprint = {2610.31474},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.31474},
}
World Models世界模型latent-actionvideo-world-modeldroidpolicy-evaluation
Paper论文 2026 · Alibaba DAMO Academy new新 2026-10-26
A 16B model interleaving video prediction and language reasoning for embodied understanding and long-horizon visual planning, plus a world-action extension for control; reports 51.5% on 20 RoboCasa365 tasks.16B 统一具身模型，交替进行视频预测与语言推理，兼顾具身理解与长程视觉规划，并扩展出用于机器人控制的世界-动作版本；另提出 Rynn-VP、Rynn-GEU 评测，报告在 20 个 RoboCasa365 组合任务上成功率 51.5%。
@misc{zhu2026rynnunify,
title = {{RynnUnify: Towards Unified Embodied Brain}},
author = {Minghao Zhu and Bohan Hou and Kun Xiang and Zixuan Wang and Yinzhou Tang and Yifei Yang and Zhikai Wang and Tianyi Zhang and Kehan Li and Haoyu Zhao and Siteng Huang and Leigang Qu and Sicong Leng and Dongchi Huang and Mingxiu Chen and Shijian Lu and Yue Wang and Xiaodan Liang and Zhenyu Zeng and Deli Zhao and Chengju Liu and Jianfei Yang and Xin Li},
year = {2026},
eprint = {2610.32965},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32965},
}
Action-conditioned world model that predicts multi-view RGB and tactile images together in one video space, post-trained for state consistency; used for closed-loop policy evaluation and GRPO updates.动作条件的视触觉世界模型：在统一的视频空间中同时预测多视角 RGB 与触觉图像，并经一致性后训练保证与真实任务状态吻合；可用于闭环策略评估和 GRPO 策略更新。
@misc{xing2026physical,
title = {{Physical Canvas: A Vision Tactile World Model For Robot Policy Evaluation And Learning}},
author = {Zebin Xing and Yuhang Zheng and Yupeng Zheng and Yichen Zhang and Zichen Li and Shangqing Zhou and Yujie Zang and Yuchen Liu and Zhongpu Xia and Haoran Li and Qichao Zhang},
year = {2026},
eprint = {2610.28613},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.28613},
}
Auxiliary training task for world action models: an EMA teacher sees the clean frame, the online model sees a perturbed one and aligns its action features. Reports better robustness for four WAMs.为世界-动作模型加一个自监督辅助任务：EMA 教师看原始画面，在线模型看外观扰动后的画面并对齐其动作表征；报告在 RoboTwin 2.0、LIBERO-Plus 等基准上提升四种 WAM 对视觉扰动的鲁棒性，且不损失原有性能。
@misc{zou2026pica,
title = {{PICA-WAM: Visual Perturbation Invariance via Cross-Appearance Self-Alignment in World Action Models}},
author = {Yude Zou and Yifei Yao and Guanbo Jiao and Jianming Ma and Guanjie Zheng and Xing Gao},
year = {2026},
eprint = {2610.26813},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.26813},
}
World action model that acts in one diffusion step: 'imagination tokens' replace random-noise initialisation as an action prior. Reports ~3× speedup over direct-policy WAMs, 40 ms latency on an RTX 5090.单步扩散即可生成动作的世界动作模型：用“想象 token”代替随机噪声初始化，为动作提供先验。报告比直接策略式 WAM 快约 3 倍，编译后在 RTX 5090 上延迟低至 40 毫秒，覆盖自动驾驶与具身控制。
@misc{zhang2026oswam,
title = {{OSWAM: An Effective World Action Model with One-Step Action Generation}},
author = {Guowen Zhang and Liyi Chen and Xu Hu and Pengfei Wang and Jingfan Chen and Chenhang He and Lei Zhang},
year = {2026},
eprint = {2610.35850},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.35850},
}
Train the video world model's autoencoder for temporal predictability, not just reconstruction, with a past-conditioned decoder. Reports up to 5× cheaper inference, 7× faster training and lower long-horizon / OOD error.视频世界模型的潜空间不应只为重建而训练：加入时间可预测性目标并配合以历史为条件的解码器。报告推理成本最多降低 5 倍、训练快 7 倍以上，长时域与分布外误差降低 15–24%，与真实策略评测更一致。
Dexterous-manipulation world model that first predicts object motion and deformation in metric 3D from a hand trajectory, then renders video. Reports 27–40% lower moved-point error than PointWorld.灵巧操作世界模型：先根据手部轨迹在度量 3D 空间中显式预测物体运动与形变，再据此渲染视频。报告在 HOT3D、OakInk2、PhysTwin 上运动点误差比 PointWorld 低 27–40%，两步因果生成大幅提速。
@misc{lee2026dexplicit,
title = {{DeXplicit: Making Physical State Explicit in Dexterous Manipulation World Models}},
author = {Adam Lee and Mutian Xu and Yukang Cao and Shulin Tian and Ziwei Liu},
year = {2026},
eprint = {2610.30781},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.30781},
}
Why action-conditioned video models ignore actions: expert demos carry too little action influence. AIM actively collects data to raise it; 500 such episodes + 3k demos match a 300k-demo model on controllability.解释动作条件视频世界模型为何忽略动作：专家示教的“动作影响力”太低，模型容量再大也无法弥补。AIM 主动采集数据提高这一指标；报告 500 条主动采集数据加 3000 条示教即可在末端可控性上追平 30 万条专家数据。
@misc{bateman2026training,
title = {{Training Controllable World Models via Action Influence Maximization}},
author = {Samuel M. Bateman and Tenny Yin and Chongyi Zheng and Lei Huang and Brian Wang and Benjamin Eysenbach and Jaime Fernández Fisac and Dhruv Shah},
year = {2026},
eprint = {2610.27387},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.27387},
}
World action model for moving targets: forecasts the object's future masks, steers a video model with a Trajectory ControlNet, and distills the pipeline to one-step video-action inference to cut latency.面向动态目标的世界动作模型：先预测目标物体的未来掩码轨迹，用 Trajectory ControlNet 引导视频模型生成与轨迹一致的未来，再通过解耦蒸馏实现一步视频-动作推理以降低延迟。
@misc{geng2026trajwam,
title = {{TrajWAM: An Efficient Trajectory-Guided World Action Model for Dynamic Robotic Manipulation}},
author = {Haohan Geng and Sha Zhang and Weishuo Sun and Guanting Ye and Wenhao Yu and Yuang Xie and Qicheng Zeng and Peng Ye and Jiajun Deng and Jianmin Ji},
year = {2026},
eprint = {2610.25763},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25763},
}
Compact (<100M) world-action model: predicts frozen-DINOv3 tokens at two future horizons and decodes action chunks, with no robot or video pretraining. Reports beating billion-parameter WAMs on RoboTwin 2.0.参数不到 1 亿的紧凑世界动作模型：在冻结 DINOv3 特征空间中预测两个未来时刻的 token 并直接回归动作块，无需机器人或视频预训练；报告在 RoboTwin 2.0 上超过十亿参数级 WAM。
@misc{liu2026gawm,
title = {{GAWM: A Compact Latent World-Action Model for Multi-Task Robot Manipulation}},
author = {Yang Liu and Xiaoli Liu and Liuzhao Li and Yuduo Song and Muyuan Liu and Xiang Gao},
year = {2026},
eprint = {2610.37091},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.37091},
}
Turns a pretrained video diffusion transformer into a policy by encoding each action step as one patch in the video latent space; at test time it outputs an action chunk in one DDIM step (84 ms), no video generation.把预训练视频扩散 Transformer 改造成策略：每个时间步的动作编码为视频潜空间中的一个 patch，联合训练视频、视频+动作与纯动作三种扩散目标；测试时一步 DDIM 生成动作块（84 毫秒），无需生成视频。
@misc{collins2026tapestry,
title = {{TAPESTRY: A Single Backbone for Video Generation and Robot Control}},
author = {Jeremy A. Collins and Seungjae Lee and Rachanon Wachakorn and Krishnan Srinivasan and Animesh Garg and Vitor Campagnolo Guizilini and Paarth Shah},
year = {2026},
eprint = {2610.36823},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36823},
}
Cross-embodiment video world model conditioned on RynnLAM's 3D latent actions, pretrained on 23 datasets (~553M frames) of human, robot and sim video; its generated data improves LIBERO and real-robot policies.以 RynnLAM 的三维潜在动作为统一条件接口的跨本体视频世界模型，在 23 个数据集（约 5.5 亿帧）的人类、机器人和仿真视频上预训练；生成的数据提升了 LIBERO 与真机策略成功率。
@misc{zhao2026rynnworld,
title = {{RynnWorld-Latent: A Cross-Embodiment World Model Conditioned on Universal Latent Actions}},
author = {Haoyu Zhao and Yinzhou Tang and Zixuan Wang and Minghao Zhu and Dongchi Huang and Xingyue Zhao and Qin Zhao and Kehan Li and Siteng Huang and Xin Li and Zhenyu Zeng and Mingxiu Chen and Deli Zhao and Zhongyu Li},
year = {2026},
eprint = {2610.36791},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36791},
}
Adds future 3D point-motion prediction as an auxiliary task to a world action model (video/motion/action MoT); inference stays one visual pass. Reports gains over Fast-WAM with few demos and under shift.在世界动作模型中加入未来三维点运动预测作为辅助目标（视频/运动/动作三专家 MoT），推理仍只需一次视觉前向；报告在少量示范和分布偏移下明显优于 Fast-WAM（RoboCasa-GR1、LIBERO、LIBERO-Plus）。
@misc{hu2026trace,
title = {{Trace-WAM: Learning to Act via 3D Motion Prediction for World Action Models}},
author = {Xu Hu and Changjie Wu and Linzhe Shi and Hang Zhang and Chen Yang and Lingjun Zhang and Guowen Zhang and Mu Xu and Junran Peng and Qing Li and Zhaoxiang Zhang},
year = {2026},
eprint = {2610.36840},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.36840},
}
@misc{kim2026adapting,
title = {{Adapting Video Models for Robot Control via Motion-guided Asymmetric Prediction}},
author = {Taeyoung Kim and Junyoung Sung and Heeseung Kwon and Minseop Kim and Jinwoo Shin},
year = {2026},
eprint = {2610.36624},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.36624},
}
Shows a fixed world model's uncertainty can break when the deployment policy changes; reweights conformal calibration by a learned state-action ratio. Reports 90% sets covering 55.0% vs 90.9% on real Franka data.指出部署策略一变，固定世界模型的不确定性校准就可能失效；用域分类器学到的状态-动作密度比做加权共形校准，并给出覆盖率保证（部分用 Lean 4 形式化验证）。报告在真实 Franka 轨迹上，普通 90% 共形集合仅覆盖 55.0%，加权后为 90.9%。
@misc{mittal2026calibrated,
title = {{Calibrated for Which Policy? Conformal Uncertainty for Fixed World Models Under Intervention}},
author = {Ayush Mittal and Dhruv Gupta},
year = {2026},
eprint = {2610.35877},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2610.35877},
}
Training-free physics check for video models: an MLLM states history-based expectations and PhysTrace marks when the video breaks them; adds IntPhys2-Timing (reports up to 70.7% hit rate).无需训练的视频模型物理一致性评测：由多模态大模型根据历史帧给出物理预期，再定位视频中违背预期的时间段并给出可解释描述；新增带违规区间标注的 IntPhys2-Timing（报告命中率最高 70.7%），面向依赖视频模型的世界动作模型。
@misc{yoo2026phystrace,
title = {{PhysTrace: Interpretable Temporal Evaluation of Physical Consistency in Video Generative Models}},
author = {Taekoan Yoo and Wonkyung Jung and Kyunghun Kim and Kyeongbo Kong},
year = {2026},
eprint = {2610.31429},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2610.31429},
}
World action model without video generation: predicts future features in a vision foundation model's latent space as policy subgoals. Reports 98.6% on LIBERO and up to 24× lower latency than pixel WAMs.不生成视频的世界动作模型：在视觉基础模型的特征空间中训练潜动作模型，预测未来观测特征作为“潜在视觉子目标”来条件化动作生成。报告 LIBERO 98.6%、RoboTwin 91.22%，每个动作块推理 187 ms，比像素级 WAM 延迟低最多 24 倍。
@misc{chen2026lawam,
title = {{LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies}},
author = {Jialei Chen and Kai Wang and Kang Chen and Shuaihang Chen and Feng Gao and Wenhao Tang and Zhiyuan Li and Weilin Liu and Zhuyu Yao and Boxun Li and Yuanbo Xu and Chao Yu},
year = {2026},
eprint = {2606.15768},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.15768},
}
World model that jointly denoises future video and 3D point trajectories in one Mixture-of-Transformers, trained on 1.44M samples; adds WorldFlowBench, reports 54% lower ADE than the best baseline on SpatialCog.同时预测未来视频与查询点 3D 轨迹的生成式世界模型：3D 运动以与视频网格对齐的位移 token 表示，由统一的 Mixture-of-Transformers 在 rectified flow 目标下联合去噪；在 144 万样本上训练，新建 WorldFlowBench，报告 SpatialCog 上 ADE 比最强外部基线降低 54.2%。
@misc{lu2026worldflow,
title = {{WorldFlow: Joint Video and 3D Motion Prediction via Large-Scale Pretraining}},
author = {Yuhao Lu and Jiongxu Chen and Hexiang Wei and Jiajin Wu and Juelin Li and Mingjian Liang and Fan Yang and Zhuowei Liang and Zihao Liu and Zhennan Wang and Sheng Wang and Lin Shao},
year = {2026},
eprint = {2610.39278},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.39278},
}
World Models世界模型video-world-model3d-trajectoriesrectified-flowmixture-of-transformersbenchmark
ID-WM pairs a high-level model that samples reachable transitions with a low-level world model used as an inverse-dynamics solver. Reports a 135% gain over single-level world-model MPC on long-horizon tasks.ID-WM 将状态可达性与局部控制解耦：高层模型在状态流形上采样可行转移，低层世界模型充当逆动力学求解器，把子目标落实为精确动作。报告比单层世界模型 MPC 提升 135%，比最优策略与扩散规划器提升 38.5%。
@misc{zhou2026hierarchical,
title = {{Hierarchical World Models with Implicit Dynamics}},
author = {Gaoyue Zhou and Yvonne Wu and Zichen Jeff Cui and Nicolas Ballas and Mido Assran and Lerrel Pinto and Yann LeCun},
year = {2026},
eprint = {2610.42146},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.42146},
}
Separates video and action generalization in WAMs: on RoboTwin 2.0 unseen tasks, adding action-free videos lifts success from 25.07% to 55.90%, and one labeled demo per task to 70.70%.拆解世界动作模型（WAM）的视频泛化与动作泛化：在 RoboTwin 2.0 的分布外任务上，加入无动作标签视频使成功率由 25.07% 升到 55.90%，每任务再加一条带标签演示升到 70.70%。WorldBridge 全数据设定报告 96.23%。
@misc{cen2026worldbridge,
title = {{WorldBridge: Bridging Video and Action for Task Generalization in World Action Models}},
author = {Jun Cen and Yubo Guo and Hao Li and Liangkai Li and Yaoxu Lyu and Meng Wei and Zeqi Xiao and Sizhe Yang and Yuqiang Yang and Yifan Ye and Shanghang Zhang and Duo Zheng},
year = {2026},
eprint = {2610.40748},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.40748},
}
GenSight puts future video generation and dense perception (depth, embodiment segmentation, 2D flow) into one video diffusion transformer with a shared prediction space.GenSight 把未来视频生成与稠密感知（深度、本体分割、二维光流）放进同一个视频扩散 Transformer，共用一个预测空间。
@misc{feng2026seeing,
title = {{Seeing and Imagining as One: Unified Video Generation and 4D Perception for Embodied World Modeling}},
author = {Haitang Feng and Zhuoman Liu and Zhennan Wang and Sheng Wang and Juelin Li and Chengzhi Zhao and Hexiang Wei and Lin Shao},
year = {2026},
eprint = {2610.39232},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.39232},
}
Learns latent actions from action-free video, then aligns them with physical motion in post-training. Comes with NavAnywhere: 17.5M observations from 15 sources across ground robots, people and drones.先从无动作标注的视频中学习潜在动作，再在后训练阶段对齐真实运动。配套数据集 NavAnywhere 含 17.5M 条视觉观测，来自 15 个来源，覆盖地面机器人、人和无人机。
@misc{nie2026opennwm,
title = {{OpenNWM: A Generalist Navigation World Model with Latent Action Pretraining}},
author = {Dujun Nie and Ruiqi Song and Zixuan Wang and Hanqing Wang and Yifan Han and Baiyong Ding and Wenzhao Zheng and Hangbin Wu and Long Chen},
year = {2026},
eprint = {2611.01844},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.01844},
}
Is a failing world action model limited by proposals or scoring? On Push-T, simulator costs in place of learned scores add 12.2 points of coverage; simulator states in proposals show no detected gain.世界动作模型失败时，问题出在动作提议还是轨迹评分？在 Push-T 上，把学习的评分换成仿真器代价可提升 12.2 个百分点的覆盖率，而在提议阶段换用仿真器状态未见增益。提供可复现的诊断流程。
A streaming multi-view causal world model. Its 3D Gaussian reconstruction feeds renderings back as scene context for long rollouts; the model targets closed-loop simulation, training and world action models.流式多视角因果世界模型：用 3D 高斯重建的渲染结果作为场景上下文来维持长时生成，面向闭环仿真、闭环训练和世界动作模型应用。
@misc{zhu2026xiaomi,
title = {{Xiaomi Auto World Model 2.0: Causal World Models for Real-World Driving and Embodied Intelligence}},
author = {Zhenxin Zhu and Zhijian Huang and Cheng Chi and Yuru Wang and Jianhua Wu and Jiayi Guan and Yonggang Li and Mingfei Tu and JiaKui Hu and Tao Xu and Xianqi Wang and Hao Li and Yuechuan Pu and Qizeng Jia and Ruiling Yang and Hongxiao Yu and Xuhan Sheng and Jian Mao and Ruiqiang Zhang and Yu Zhu and Guang Chen and Hangjun Ye and Bing Wang},
year = {2026},
eprint = {2611.01677},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2611.01677},
}
An autoregressive multiview world model running at 10.1 FPS that adapts to held-out robots in 100 to 200 gradient steps; its authors release weights and code and note that fine local calibration still trails.自回归多视角世界模型，运行 10.1 FPS，100 至 200 步梯度即可适配未见平台；作者开放权重和代码，并说明细粒度局部标定仍落后于同类。
@misc{davies2026primesim,
title = {{PrimeSim: A Generalist World Model Simulator from Large-Scale Cross-Embodiment Demonstrations}},
author = {Travis Davies and Wentao Ma and Lihao Guo and Tengfei Ma and Xiaowen Chang and Jiafeng Xu and Shilong Wang and Yiqi Deng and Shaowen He and Xiaoqi Li and Qian Zhu and Ran Cheng and Hao Dong},
year = {2026},
eprint = {2611.04870},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04870},
}
A 14B action-controllable video model trained on three million manipulation clips, with DPO-style post-training that suppresses unphysical results such as penetration, plus the EZSbench benchmark.14B 动作可控视频模型，基于 300 万条操作片段训练，用 DPO 式后训练抑制穿模等非物理现象，并提出 EZSbench 基准。
@misc{chen2026abot,
title = {{ABot-PhysWorld: Interactive World Foundation Model for Robotic Manipulation with Physics Alignment}},
author = {Yuzhi Chen and Ronghan Chen and Dongjie Huo and Yandan Yang and Dekang Qi and Haoyun Liu and Tong Lin and Shuang Zeng and Junjin Xiao and Xinyuan Chang and Feng Xiong and Xing Wei and Zhiheng Ma and Mu Xu},
year = {2026},
eprint = {2603.23376},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2603.23376},
}
Performs unseen manipulation tasks by following a human video placed in context, with no parameter update; it turns robot trajectories into matched human videos to get paired training data.通过跟随上下文中的人类视频执行未见过的操作任务，无需更新参数；并把机器人轨迹自动转成匹配的人类视频，以获得成对训练数据。
@misc{zhou2026zero,
title = {{Zero-WAM: In-Context World-Action Modeling from Human Videos for Open-Ended Task Generalization}},
author = {Jiaming Zhou and Qihang Zhang and Gangwei Xu and Cunxin Fan and Yujie Zhao and Ruilin Wang and Yiming Luo and Shuai Yang and Xing Zhu and Yujun Shen and Junwei Liang and Yinghao Xu},
year = {2026},
eprint = {2608.26103},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.26103},
}
A three-stream world action model under a UniDiffuser formulation; one model covers policy, world modeling, video generation and inverse dynamics, and it scales to video-only and cross-embodiment data.基于 UniDiffuser 的三流世界动作模型，一个模型同时覆盖策略、世界建模、视频生成和逆动力学，并可利用纯视频和跨形态数据。
@misc{team2026motubrain,
title = {{Motubrain: An Advanced World Action Model for Robot Control}},
author = {Motubrain Team and Chendong Xiang and Fan Bao and Haitian Liu and Hengkai Tan and Hongzhe Bi and James Li and Jiabao Liu and Jingrui Pang and Kiro Jing and Louis Liu and Mengchen Cai and Rongxu Cui and Ruowen Zhao and Runqing Wang and Shuhe Huang and Yao Feng and Yinze Rong and Zeyuan Wang and Jun Zhu},
year = {2026},
eprint = {2604.27792},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.27792},
}
A latent world-action model that adds future-aware latent queries to a VLA-style policy and trains them with a future-informed branch, without generating future frames.潜空间世界动作模型，在 VLA 式策略中加入面向未来的潜查询，并用利用未来信息的分支训练，全程不生成未来帧。
@misc{luo2026being,
title = {{Being-H0.7: A Latent World-Action Model from Egocentric Videos}},
author = {Hao Luo and Wanpeng Zhang and Yicheng Feng and Sipeng Zheng and Haiweng Xu and Chaoyi Xu and Ziheng Xi and Yuhui Fu and Zongqing Lu},
year = {2026},
eprint = {2605.00078},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.00078},
}
A WAM combining a video expert, an action expert, a frozen VLM and training-only 4D geometric distillation, pretrained on a heterogeneous corpus with over 20K hours of open data.结合视频专家、动作专家、冻结 VLM 和仅训练时使用的 4D 几何蒸馏的世界动作模型，在含 2 万多小时开放数据的异构语料上预训练。
@misc{miao2026internw,
title = {{InternW0-$Δ$: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data}},
author = {Xingyu Miao and Zizun Li and Baole Fang and Kaiwen Song and Tenghui Wang and Hanxue Zhang and Yating Wang and Xudong Li and Yuping He and Xueyuan Wei and Chao Gao and Xijie Yang and Yingxiang Xu and Kerui Ren and Wenqi Guo and Jianjun Zhou and Xinzhe Wang and Weiguang Zhao and Ni Yang and Zetao Cai and Yufei Xue and Hengjie Li and Zeyu He and Yuanzhen Zhou and Rong Fu and Jianyang Zhang and Siwei Cui and Fuxian Huang and Yunsong Zhou and Xing Gao and Yifei Yao and Qiaojun Yu and Kailin Li and Ming Zhou and Mu Huang and Xinyue Li and Wenze Cui and Bingqi Jiang and Xueyue Zhu and Junting Dong and Haoyu Guo and Tao Lu and Mulin Yu and Bowen Zhou and Bin Zhao and Tianfan Xue and Weinan Zhang and Chunhua Shen},
year = {2026},
eprint = {2609.31394},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.31394},
}
Blames WAM motion errors on video bases not tuned for manipulation and on naive action add-ons, then continues pretraining the video model on manipulation clips with event-level captions.认为世界动作模型的运动错误来自未针对操作优化的视频底座和粗糙的动作模块，因此在带事件级描述的操作视频上继续预训练视频模型。
@misc{guo2026video,
title = {{Video Prediction Policy 2: Predict Better, Act Better}},
author = {Yanjiang Guo and Haodong Yan and Zhide Zhong and Zhongru Zhang and Qingyuan Yang and Qingzhou Lu and Xiaoyu Chen and Yen-Jen Wang and Shuying Deng and Chenghan Yang and Puzhen Yuan and Chenxin Liu and Tun Ban and Xiang Zhu and Yichen Liu and Kun Feng and Haoang Li and Jianyu Chen},
year = {2026},
eprint = {2610.10270},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.10270},
}
A language-conditioned video world model using language as a shared action interface across manipulation, driving, indoor navigation and human-to-robot transfer.语言条件视频世界模型，把自然语言作为操作、驾驶、室内导航和人到机器人迁移的统一动作接口。
@misc{zhang2026qwen,
title = {{Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation}},
author = {Jie Zhang and Xiaoyue Chen and Anzhe Chen and Dayiheng Liu and Deqing Li and Gengze Zhou and Hale Yin and Haoqi Yuan and Haoyang Li and Jiahao Li and Jiazhao Zhang and Jingren Zhou and Kaiyuan Gao and Kun Yan and Lihan Jiang and Ningyuan Tang and Pei Lin and Qihang Peng and Shengming Yin and Tianhe Wu and Tianyi Yan and Xiao Xu and Yan Shu and Yanran Zhang and Ye Wang and Yi Wang and Yilei Chen and Yixian Xu and Yiyang Huang and Yuxiang Chen and Zekai Zhang and Zhendong Wang and Zixing Lei and Zhixuan Liang and Zihao Liu and Zikai Zhou and Chenxu Lv and Xiong-Hui Chen and Chenfei Wu},
year = {2026},
eprint = {2606.17030},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.17030},
}
A 38-billion-parameter multimodal autoregressive model that treats embodied generation as an extension of image and video generation, keeping the prior of the pretrained world foundation model.380 亿参数的多模态自回归模型，把具身生成视为图像和视频生成的延伸，以保留预训练世界基础模型的先验。
@misc{li2026xiaomi,
title = {{Xiaomi-Robotics-U0: Unified Embodied Synthesis with World Foundation Model}},
author = {Xinghang Li and Jun Guo and Qiwei Li and Long Qian and Hang Lai and Yueze Wang and Hongyu Yan and Jiahang Cao and Xi Chen and Jingen Qu and Jiaxi Song and Nan Sun and Hanye Zhao and Futeng Liu and Wanli Peng and Heyun Wang and Yunhong Wang and Caoyu Xia and Jack Zhao and Diyun Xiang and Hangjun Ye and Heng Qu and Huaping Liu and Jason Li},
year = {2026},
eprint = {2607.11643},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.11643},
}
A training framework that improves the physical consistency of video world simulators by focusing supervision on physics-informative regions at pixel and semantic levels.训练框架，把监督集中在具有物理信息的区域，在像素和语义两个层面提升视频世界模拟器的物理一致性。
@misc{zhang2026physisforcing,
title = {{PhysisForcing: Physics Reinforced World Simulator for Robotic Manipulation}},
author = {Peiwen Zhang and Yufan Deng and Shangkun Sun and Juncheng Ma and Duomin Wang and Jonas Du and Zilin Pan and Ye Huang and Hao Liang and Songyan Huang and Ruihua Zhang and Enze Xie and Ming-Yu Liu and Daquan Zhou},
year = {2026},
eprint = {2606.28128},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.28128},
}
A robustness study that compares world action models against VLAs under perturbation, to test the claim that explicit future prediction gives better generalization.在扰动下对比世界动作模型与 VLA 的鲁棒性研究，用于检验显式未来预测能带来更好泛化这一说法。
@misc{zhang2026do,
title = {{Do World Action Models Generalize Better than VLAs? A Robustness Study}},
author = {Zhanguang Zhang and Zhiyuan Li and Behnam Rahmati and Rui Heng Yang and Yintao Ma and Amir Rasouli and Sajjad Pakdamansavoji and Yangzheng Wu and Lingfeng Zhang and Tongtong Cao and Feng Wen and Xinyu Wang and Xingyue Quan and Yingxue Zhang},
year = {2026},
eprint = {2603.22078},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.22078},
}
A robotics-oriented WAM survey that separates WAMs from world models, model-based RL and reactive VLAs, with a taxonomy of representations, architectures, training and data.面向机器人的世界动作模型综述，厘清其与世界模型、基于模型的强化学习和反应式 VLA 的区别，并按表示、架构、训练和数据分类。
@misc{lu2026world,
title = {{World-Action Models for Robot Learning and Control: A Survey}},
author = {Zuxing Lu and Hongjia Zhai and Guanzhi Wang and Huajian Zeng and Jiaqi Yang and Jingyu Liu and Lei Cheng and Yuantai Zhang and Yuheng Qiu and Zezhou Cheng and Ivan Laptev and Danfei Xu and Benjamin Riviere and Giuseppe Loianno and Eric Xing and Xingxing Zuo},
year = {2026},
eprint = {2609.16074},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2609.16074},
}
A concise tutorial rather than a survey: defines world models and WAMs for robotics and compares spatial intelligence models, JEPA and Cosmos in a single frame.简明教程而非综述：界定机器人领域的世界模型和世界动作模型，并在统一视角下比较空间智能模型、JEPA 和 Cosmos。
@misc{zhang2026from,
title = {{From World Models to World Action Models: A Concise Tutorial for Robotics}},
author = {Xiaoxiong Zhang and Xiong Zeng and Wei Zhang},
year = {2026},
eprint = {2607.00836},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.00836},
}
Finds that WAM futures can hallucinate training-domain content under visual shift, and uses DINOv3 as the shared representation for future prediction and history retrieval to keep actions robust.发现视觉偏移下世界动作模型预测的未来会幻想出训练域内容，于是用 DINOv3 作为未来预测和历史检索的共享表示，以保持动作鲁棒。
@misc{wang2026st,
title = {{ST-WAM: Semantic-Temporal World Action Model for Robust Manipulation under Visual Distribution Shifts}},
author = {Mingxin Wang and Bin Hu and Bin Qian and Kaitao Jiang and Haoning Wu and Feng Yan and Bowen Jing and Ruiyang Hao and Enyi Wang and Kangning Niu and Yandan Yang and Mu Xu and Yan Wang and Houde Liu and Tianlun Li},
year = {2026},
eprint = {2607.28993},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.28993},
}
A closed-loop video simulator retrained on thousands of hours of real robot data, with a state expert, a world judge that scores rollouts, and 25-frame rollouts in 2.3 seconds on one H100.闭环视频世界模拟器，用数千小时真机数据重新训练，加入状态专家和为推演打分的世界评判器，在单张 H100 上 2.3 秒生成 25 帧推演。
@misc{qiu2026ge,
title = {{GE-Sim 2.0: A Roadmap Towards Comprehensive Closed-loop Video World Simulators for Robotic Manipulation}},
author = {Boxiang Qiu and Liliang Chen and Yue Liao and Nan Wang and Lintao Wang and Jiayi Luo and Wenzhi Zhao and Shengcong Chen and Di Chen and Ye Li and Chen Gao and Shuicheng Yan and Si Liu and Maoqing Yao and Guanghui Ren},
year = {2026},
eprint = {2605.27491},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2605.27491},
}
Video world model that first predicts where and how objects will respond to contact, then feeds that into object features and temporal attention; reports 3.29 points above PhysisForcing on RBench.视频世界模型：先预测物体对接触的响应位置和方式，再反馈到物体特征与时序注意力；报告在 RBench 上比 PhysisForcing 高 3.29 分。
@misc{liu2026reactworld,
title = {{ReactWorld: Structuring Interaction Responses for Embodied Video Generation}},
author = {Hanwei Liu and Huijia Zhang and Hanzhang Wang and Guangyu Liu and Hanbei Zhan and Linjie Hu and Bin Xiao and Xuan Zhou and Dong Zhang},
year = {2026},
eprint = {2611.06120},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.06120},
}
Predicts how nearby vehicles respond to each candidate action and feeds those consequences to a typed decision model; reports replayed NGSIM collisions falling from 48.2% to 2.8%.预测每个候选动作下周围车辆的反应，再把后果交给类型化决策模型比较；报告在回放的 NGSIM 片段中碰撞率从 48.2% 降至 2.8%。
@misc{qu2026jev,
title = {{Jev with Foresight: Safe Language-Guided Autonomous Driving via a Language World Model}},
author = {Yansong Qu and Jiancong Chen and Samuel Labi},
year = {2026},
eprint = {2611.05448},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05448},
}
Compares how a pretrained LeWorldModel and the simulator respond to small action changes; their most sensitive directions differ by a median of about 36–42° at the smallest scale, so evaluation should be scale-specific.比较预训练 LeWorldModel 与仿真器对微小动作变化的响应；最小尺度下最敏感方向的中位夹角相差约 36–42°，说明评估应按尺度进行。
Builds a robot foundation model on video instead of language: an open video planner trained on internet-scale human video makes zero-shot plans that are turned into robot actions. Model and dataset are released.以视频而非语言为主干构建机器人基础模型：在互联网规模人类视频上训练开源视频规划器，零样本生成视频计划再转为机器人动作；模型与数据集均已开源。
@misc{chen2025large,
title = {{Large Video Planner Enables Generalizable Robot Control}},
author = {Boyuan Chen and Tianyuan Zhang and Haoran Geng and Caiyi Zhang and Peihao Li and Kiwhan Song and William T. Freeman and Jitendra Malik and Pieter Abbeel and Russ Tedrake and Vincent Sitzmann and Yilun Du},
year = {2025},
eprint = {2512.15840},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.15840},
}
Do world action models need video generation? ImageWAM conditions a flow-matching action expert on image-editing KV caches instead. Reports beating VLA baselines at 1/6 the FLOPs and 1/4 the latency of video WAMs.追问世界动作模型是否真需要视频生成：ImageWAM 改用图像编辑模型去噪时的 KV 缓存来条件化流匹配动作专家；报告优于 VLA 基线，计算量为视频 WAM 的 1/6、延迟为 1/4。
@misc{zhang2026imagewam,
title = {{ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?}},
author = {Yuyang Zhang and Wenyao Zhang and Zekun Qi and He Zhang and Haitao Lin and Jingbo Zhang and Yao Mu and Xiaokang Yang and Wenjun Zeng and Xin Jin},
year = {2026},
eprint = {2606.19531},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.19531},
}
Lightweight world action model: compact video backbone with future supervision in a downsampled latent, and an action expert that pools several layers to predict chunks in one pass. Reports 0.44B trainable params, 72 ms.轻量世界动作模型：紧凑视频主干在降采样潜空间中做未来监督，动作专家融合多层特征一次前向预测动作块；报告可训练参数仅 0.44B，推理延迟 72 毫秒。
@misc{li2026light,
title = {{Light-WAM: Efficient World Action Models with State-Fusion Action Decoding}},
author = {Ziang Li and Dongzhou Cheng and Yibin Wang and Shiyue Wang and Xiaoyang Xu and Lingxuan Weng and Juan Wang and Jiaqi Wang},
year = {2026},
eprint = {2606.08242},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.08242},
}
World model that predicts 3D traces of key points (objects, tools, hands) instead of pixels or actions, pretrained from video without action labels. Trace-conditioned policies are reported competitive with π0-style VLAs.不预测像素也不预测动作，而是预测物体、工具、手等关键点的 3D 轨迹，可用无动作标签的视频预训练。报告以轨迹为条件的策略可与 π0 类 VLA 相当。
@misc{lee2026scalable,
title = {{$μ_0$: A Scalable 3D Interaction-Trace World Model}},
author = {Seungjae Lee and Yoonkyo Jung and Jusuk Lee and Jonghun Shin and Amir Hossein Shahidzadeh and Yao-Chih Lee and H. Jin Kim and Jia-Bin Huang and Furong Huang},
year = {2026},
eprint = {2606.13769},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.13769},
}
Steers a video world model toward rare but plausible bad outcomes (e.g. task failures) by optimizing its initial noise with a VLM objective, to evaluate and improve policies. Driving and manipulation.通过优化视频世界模型的初始噪声（以 VLM 为目标），把想象引向罕见但合理的不利结果（如任务失败），用于评估和改进策略；覆盖自动驾驶与操作。
@misc{seo2026stressdream,
title = {{StressDream: Steering Video World Models for Robust Policy Evaluation and Improvement}},
author = {Junwon Seo and Sushant Veer and Ran Tian and Wenhao Ding and Apoorva Sharma and Karen Leung and Edward Schmerling and Marco Pavone and Andrea Bajcsy},
year = {2026},
eprint = {2606.00267},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.00267},
}
Probe study of which video world model latents carry action information: temporal video pretraining matters, pixel reconstruction quality does not. V-JEPA-style encoders give the best trade-off.探测哪些视频世界模型的潜变量含有动作信息：关键在于时序视频预训练，而非像素重建质量；V-JEPA 类编码器取得最佳权衡。
@misc{yeom2026what,
title = {{What Makes Video World Model Latents Action-Relevant: Prediction over Reconstruction}},
author = {Jewon Yeom and Hanseul Kim and Jeongjae Park and Sungmok Jung and Jaejin Lee and Taesup Kim},
year = {2026},
eprint = {2606.07687},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.07687},
}
Small world action model trained from scratch on frozen V-JEPA 2.1 features, predicting future features only in training. Reports SOTA on RoboTwin 2.0 among policies without embodied pretraining.在冻结的 V-JEPA 2.1 特征上从零训练的紧凑世界动作模型，仅在训练时预测未来特征。报告在无具身预训练的策略中 RoboTwin 2.0 最优，并修正了 LIBERO-Plus 的指令。
@misc{clemente2026la,
title = {{La-JEPA: Compact Robot Policies with Latent World Modeling}},
author = {Mateo Clémente and George Z. Xue and Amin Abyaneh and Liheng Ma and Rui Heng Yang and Tongtong Cao and Zhanguang Zhang and Yingxue Zhang},
year = {2026},
eprint = {2611.08156},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08156},
}
Generative world model of 3D particle and rigid-body motion covering rigid, deformable, cloth and granular materials. Reports 31% lower error than prior learned dynamics; planning searches its latent noise.生成式 3D 粒子与刚体运动世界模型，统一覆盖刚体、可变形体、布料与颗粒材料。报告预测误差比此前最好的学习动力学低 31%，并通过在潜噪声空间搜索来做规划。
@misc{ren2026gpwm,
title = {{GPWM: Modeling the World as 3D Particle Motion}},
author = {Robert Ren and Matthew Bronars and He Zhu and Yikai Tang and Jianing Zhang and Sergey Zakharov and Pavel Tokmakov and Katerina Fragkiadaki},
year = {2026},
eprint = {2611.08338},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08338},
}
World action model that predicts oriented bounding-box trajectories of task objects instead of future video frames: much cheaper, and reported to predict actions better.世界动作模型不再预测未来视频帧，而是预测任务相关物体的有向包围盒轨迹：计算量大幅下降，并报告动作预测更好。
@misc{yu2026box,
title = {{Box-WAM: Box Trajectories as Compact State Representations for World Action Models}},
author = {Qinxi Yu and Wenzhou Ding and Gloria X. Wang and Albert J. Zhai and Shenlong Wang},
year = {2026},
eprint = {2610.17526},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.17526},
}
Recalibrates DROID into DROID-3D with metric depth and multi-view extrinsics, then trains a video world model that predicts RGB and depth. Depth supervision reportedly also improves RGB by 1.48 dB PSNR.把 DROID 重新标定为带度量深度与多视角外参的 DROID-3D，并训练同时预测 RGB 与深度的视频世界模型。报告深度监督还让 RGB 预测提升 1.48 dB PSNR。
@misc{bardhan2026depthworld,
title = {{DepthWorld: 3D World Model for Robot Manipulation}},
author = {Jai Bardhan and Josef Sivic and Vladimir Petrik},
year = {2026},
eprint = {2610.08780},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.08780},
}
Factorizes an interaction into action, camera and embodiment so one human video can be re-rendered as many robot rollouts. Generated data helps on RoboCasa GR1 and a real IRON humanoid.把一次交互分解为动作、相机与本体三个因素，使一段人类视频可重新生成为多种机器人视角的数据。生成数据在 RoboCasa GR1 与真实 IRON 人形上带来提升。
@misc{chen2026anyworld,
title = {{AnyWorld: Factorized Egocentric World Models for Cross-Embodiment Generalization}},
author = {Cheng Chen and Jerry Bai and Jiacheng Wei and Boyu Chen and Xiaoji Zheng and Fan Wu and Minghao Yang and Tianrun Chen and Ruibo Li and Xiaoyu Yue and Xiaoyang Guo and Yixiao Ge and Guosheng Lin and Fayao Liu},
year = {2026},
eprint = {2608.29242},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.29242},
}
Adds episodic memory tokens to video-world-model policies so actions can depend on events no longer in view. Reports LIBERO-Mem success up from 5% to 42.5% and 75–80% on real memory tasks.为视频世界模型策略加入情景记忆 token，使动作能依赖已不在视野中的事件。报告 LIBERO-Mem 成功率从 5% 提升到 42.5%，真机记忆类任务约 75–80%。
@misc{jiang2026memoryvam,
title = {{MemoryVAM: Integrating Memory into Video Action Model for Robot Manipulation}},
author = {Yuxin Jiang and Chang Yu and Yunuo Chen and Xiang Feng and Yin Yang and Nishank Gite and Chenfanfu Jiang},
year = {2026},
eprint = {2606.20679},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.20679},
}
Zero-shot long-horizon manipulation: a VLM proposes and checks several generated video plans, actions follow object or hand flow, and execution is monitored with local corrections such as poking.零样本长程操作：VLM 提出并验证多条生成的视频计划，动作由物体流或手部流转换而来，执行中持续监控并做戳推等局部纠正。
@misc{fu2026novaplan,
title = {{NovaPlan: Zero-Shot Long-Horizon Manipulation via Closed-Loop Video Language Planning}},
author = {Jiahui Fu and Junyu Nan and Lingfeng Sun and Hongyu Li and Jianing Qian and Benjamin Yang and Yilun Du and Jennifer L. Barry and Kris Kitani and George Konidaris},
year = {2026},
eprint = {2602.20119},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.20119},
}
World Models世界模型video-generationplanningzero-shotlong-horizonreal
World model for deformable objects that couples a differentiable MPM simulator with networks that infer per-particle elasticity from video and correct the simulator's residual errors. Real manipulation sequences.可变形物体世界模型：可微 MPM 仿真器配合两个网络，一个从视觉推断逐粒子弹性，一个学习修正仿真器的残差误差。在真实操作序列上验证。
@misc{yin2026physcore,
title = {{PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics}},
author = {Haocheng Yin and Shuohan Tao and Yongsheng Chen and Lu Gan},
year = {2026},
eprint = {2607.20653},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.20653},
}
World Models世界模型deformablempmresidual-learningmaterial-identificationreal
Controlled study of which world-prediction target best transfers egocentric human data to robots: pixels transfer weakly, DINO features give up to 4x OOD generalization, 3D flow gives +20–30% in domain.对照研究哪种世界预测目标最能把第一人称人类数据迁移到机器人：像素迁移较弱，DINO 特征带来最高 4 倍的分布外泛化，3D 运动流在域内提升 20–30%。
@misc{li2026egowam,
title = {{EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human Data}},
author = {Baoyu Li and Xinchen Yin and Mengying Lin and Yixin Zhang and Danfei Xu},
year = {2026},
eprint = {2607.08436},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.08436},
}
Trains an action-conditioned video world model purely from autonomous robot play instead of human demos, capturing failure-rich contact. Reports up to 40% better policy evaluation and +65% real success after RL in it.完全用机器人自主“玩耍”数据（而非人类演示）训练动作条件视频世界模型，覆盖更多接触与失败情形。报告策略评估提升最高 40%，在其中做强化学习后真机成功率提升 65%。
@misc{yin2026playworld,
title = {{PlayWorld: Learning Robot World Models from Autonomous Play}},
author = {Tenny Yin and Zhiting Mei and Zhonghe Zheng and Miyu Yamane and David Wang and Jade Sceats and Samuel M. Bateman and Lihan Zha and Apurva Badithela and Ola Shorinwa and Anirudha Majumdar},
year = {2026},
eprint = {2603.09030},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2603.09030},
}
World Models世界模型video-world-modelself-playpolicy-evaluationrlreal
One video-diffusion backbone serves as both a video-action policy and an action-conditioned simulator that scores progress, trained on ~27,300 hours of robot, UMI and human data; samples and ranks actions at test time.同一个视频扩散主干既是视频-动作策略，也是可预测任务进度的动作条件模拟器；训练数据约 2.73 万小时（真机、UMI、人类视频等），推理时采样并排序候选动作。
@misc{zhou2026wm,
title = {{$τ_0$-WM: A Unified Video-Action World Model for Robotic Manipulation}},
author = {Pengfei Zhou and Shengcong Chen and Di Chen and Jiaxu Wang and Rongjun Jin and Bingwen Zhu and Yike Pan and Songen Gu and Kuanning Wang and Shufeng Nan and Xingyu Qiu and Chenhao Qiu and Pu Yang and Yunuo Cai and Jianxiong Gao and Yifan Li and Yanwei Fu and Xiangyu Yue and Zhi Chen and Jianlan Luo},
year = {2026},
eprint = {2606.01027},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.01027},
}
World Models世界模型video-action-modelsimulatortest-time-computelarge-scalereal
RL post-training of VLAs inside a video world model, with keyframe-initialised rollouts and world-model–policy co-evolution to keep hallucination in check. Reports strong LIBERO results and real-robot gains.在视频世界模型中对 VLA 做强化学习后训练，通过关键帧初始化的短推演和世界模型-策略协同演化来抑制幻觉。报告 LIBERO 表现优异并在多个真机平台上提升。
@misc{jiang2026wovr,
title = {{WoVR: World Models as Reliable Simulators for Post-Training VLA Policies with RL}},
author = {Zhennan Jiang and Shangqing Zhou and Yutong Jiang and Zefang Huang and Mingjie Wei and Yuhui Chen and Tianxing Zhou and Zhen Guo and Hao Lin and Quanlu Zhang and Yu Wang and Haoran Li and Dongbin Zhao and Chao Yu},
year = {2026},
eprint = {2602.13977},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.13977},
}
A large pretrained world model plans subtasks and generates goal images; a VLA follows them. Reports the same VLA rising from 14% to 69% in novel scenes with these generated goals.用大规模预训练世界模型做高层规划，把任务拆成子任务并生成目标图像，由底层 VLA 跟随执行。报告同结构 VLA 在新场景中的成功率借此从 14% 提升到 69%。
@misc{long2026scaling,
title = {{Scaling World Model for Hierarchical Manipulation Policies}},
author = {Qian Long and Yueze Wang and Jiaxi Song and Junbo Zhang and Peiyan Li and Wenxuan Wang and Yuqi Wang and Haoyang Li and Shaoxuan Xie and Guocai Yao and Hanbo Zhang and Xinlong Wang and Zhongyuan Wang and Xuguang Lan and Huaping Liu and Xinghang Li},
year = {2026},
eprint = {2602.10983},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.10983},
}
Action-conditioned world-action model that also learns from failed rollouts, predicting future video and task progress for the action actually taken; progress scores can rank sampled actions.以实际执行动作为条件的世界动作模型，把失败轨迹也作为监督，预测未来视频与任务进度；进度分可在推理时为候选动作排序。
@misc{peng2026fact,
title = {{FACT: Failure-Aware Causal Training for World-Action Models}},
author = {Quanquan Peng and Yutong Liang and Rui Yan and Nicklas Hansen and Xiaolong Wang},
year = {2026},
eprint = {2608.10232},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2608.10232},
}
World Models世界模型Robot arm机械臂world-action-modelfailure-dataprogress-estimationbimanualreal
RL inside a learned diffusion world model without a simulator: the big model rolls out trajectories while a light latent surrogate supplies gradients. Beats PPO on sample efficiency on Push-T; quadruped task too.不用物理仿真器，直接在学习到的扩散世界模型中训练策略：大模型负责生成轨迹，轻量潜在代理模型提供一阶梯度。在 Push-T 上样本效率明显优于 PPO，并在四足任务上验证。
@misc{amigo2026coupled,
title = {{Coupled Local and Global World Models for Efficient First Order RL}},
author = {Joseph Amigo and Rooholla Khorrambakht and Nicolas Mansard and Ludovic Righetti},
year = {2026},
eprint = {2602.06219},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2602.06219},
}
World Models世界模型Quadruped / legged四足 / 足式rldiffusion-world-modelfirst-order-gradientspush-tquadruped
Hierarchical MPC on latent visual world models at two time scales, linked by latent subgoals and learned macro-actions. Reports 70% vs 0% on real Franka pick-and-place from one goal image, with up to 3x less compute.在两个时间尺度的潜在视觉世界模型上做分层 MPC，以潜在子目标和学习到的宏动作衔接。报告真实 Franka 单目标图抓放成功率 70%（单层规划为 0%），计算量最多少 3 倍。
@misc{zhang2026hierarchical,
title = {{Hierarchical Planning with Latent World Models}},
author = {Wancong Zhang and Basile Terver and Artem Zholus and Soham Chitnis and Harsh Sutaria and Mido Assran and Randall Balestriero and Amir Bar and Yann LeCun and Nicolas Ballas},
year = {2026},
eprint = {2604.03208},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2604.03208},
}
World Models世界模型Robot arm机械臂planninglatent-world-modelhierarchical-mpcmacro-actionsreal
Video world model that renders the robot's own motion from its URDF and controller, so the network only predicts how the scene responds. Transfers to unseen robots and can animate retargeted human demos.视频世界模型把机器人自身运动交给控制器和 URDF 渲染，网络只需预测场景的响应。可迁移到未见过的机器人本体，并能由重定向的人类演示生成机器人操作视频。
@misc{kim2026robot,
title = {{Robot-Factored World Models via Robot Rendering}},
author = {Byungjun Kim and Taeksoo Kim and Hyunsoo Cha and Hanbyul Joo},
year = {2026},
eprint = {2607.22535},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.22535},
}
World Models世界模型video-world-modelrobot-renderingurdfcross-embodimenthuman-video
Learns STRIPS-style operators (preconditions, add/delete effects) over latent predicates from image transitions, then plans classically from start and goal images. Beats visual-rollout and latent-search methods.直接从图像转移数据学习 STRIPS 式符号算子（前提条件、增删效果）和潜在二值谓词，再从起始图与目标图用经典规划求解。优于视觉推演和潜在图搜索等基线。
RoboImagine post-trains autoregressive robot world models by comparing candidate futures, scored by a frozen multimodal judge plus pixel/perceptual terms. Reports +1.26 dB PSNR on RT-1 over continued SFT.RoboImagine 通过比较多条候选未来来后训练自回归机器人世界模型：冻结的多模态评估器给出语义偏好，像素与感知层面的一致性项约束视觉误差。在 RT-1 上报告 PSNR 比继续监督微调高 1.26 dB。
@misc{wu2026learning,
title = {{Learning Robotic World Models with Semantic and Visual Feedback}},
author = {Hao Wu and Yuqi Li and Yizhou Zhao and Kun Wang and Xiao Luo and Yu Cheng and Bo An and Yingli Tian and Laszlo A. Jeni and Ming-Hsuan Yang},
year = {2026},
eprint = {2611.07850},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07850},
}
World Models世界模型world-modelpost-trainingvideo-predictionvlm-judgegrpo
Distills intermediate denoising features of a video diffusion model into an action model during training and drops the video model at test time. Reports 95.9% on LIBERO and about 21x faster inference than a WAM baseline.训练时把视频扩散模型的中间去噪表示蒸馏进动作模型，测试时完全丢弃视频模型。在 LIBERO 上报告 95.9% 平均成功率，推理比代表性世界动作模型快约 21 倍。
@misc{zhang2026w,
title = {{W2A: Distilling World Priors into Action Models for Real-Time Robotic Control}},
author = {Xiaopeng Zhang and Yongjin Mu and Yueyang Weng and Chengzhi Zhao and Yanjie Li},
year = {2026},
eprint = {2611.02600},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02600},
}
World Models世界模型world-action-modeldistillationvideo-diffusionlibero
FAWAM uses 6-axis force/torque in perception, jointly predicts future actions and wrenches, and corrects actions online against the predicted wrench. Reports +36.25 pts over vision-only baselines on real tasks.FAWAM 在感知、预测与闭环执行三层使用六维力/力矩：联合预测未来动作与末端力旋量，并以预测的力轨迹为参考在线修正动作。真机任务上比纯视觉基线成功率高 36.25 个百分点。
@misc{he2026fawam,
title = {{FAWAM: Force-Aware World Action Models for Closed-Loop Contact-Rich Manipulation}},
author = {Haotian He and Zeyu Yan and Qipeng Liu and Ning Guo and Wenzhao Lian},
year = {2026},
eprint = {2606.08555},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.08555},
}
Object-centric world-action model that both takes object masks as prompts and predicts future masks, reducing language ambiguity in clutter and background bias in RGB prediction.以物体为中心的世界动作模型：既把目标物体掩码作为提示输入，又预测未来掩码，从而减少杂乱场景中的语言歧义和 RGB 预测中的背景偏差。
@misc{yu2026maskwam,
title = {{MaskWAM: Unifying Mask Prompting and Prediction for World-Action Models}},
author = {Hanyang Yu and Haitao Lin and Jingbo Zhang and Wenyao Zhang and Chenghao Gu and Heng Li and Ping Tan},
year = {2026},
eprint = {2606.13515},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2606.13515},
}
World Models世界模型world-action-modelmasksobject-centricmixture-of-transformers
3D world model that encodes each object's point cloud with an SDF-supervised vector-neuron VAE and predicts per-object SE(3) increments applied analytically, so shapes do not drift over long rollouts.3D 世界模型：用 SDF 监督的向量神经元 VAE 编码每个物体的点云，并预测逐物体的 SE(3) 运动增量后解析地施加，长时推演中物体形状不会漂移。
@misc{weng2026ord,
title = {{ORD-WM: Learning Object-Centric Rigid Dynamics for 3D World Modeling in Robotic Manipulation}},
author = {Yueyang Weng and Xiaopeng Zhang and Yuxi Ou and Yongjin Mu and Chengzhi Zhao and Yanjie Li},
year = {2026},
eprint = {2610.23156},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.23156},
}
Defines Temporal Ratio, how much an action head attends to future latents vs the current frame; it predicts compositional generalization of video-action models and drives an inference-time adaptive guidance method.提出 Temporal Ratio：动作头对未来潜变量相对当前帧的注意力依赖程度；它能预测视频动作模型的组合泛化能力，并据此设计推理时的自适应引导方法。
@misc{mishra2026understanding,
title = {{Understanding and Mitigating the Video-Action Generalization Gap via Temporal Ratio}},
author = {Utkarsh A. Mishra and Yongxin Chen and Danfei Xu and Yang Liu and Xi Chen and Jiayuan Mao},
year = {2026},
eprint = {2607.08127},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2607.08127},
}
World Models世界模型world-action-modelvideo-action-modelgeneralizationanalysis
Steers a frozen world-action model with unlabeled human videos at test time: a small memory inside the model adapts via self-supervised video prediction, after meta-training on paired human-robot data.测试时用无标注的人类视频引导冻结的世界动作模型：模型内部的小型记忆通过自监督视频预测进行适配；此前用成对人-机数据做元训练使其对控制有用。
@misc{feng2026wam,
title = {{WAM-TTT: Steering World-Action Models by Watching Human Play at Test Time}},
author = {Yusen Feng and Bingchen Han and Jiangran Lyu and Kai Liu and Yixin Zheng and Yuxuan Wan and Weiheng Liu and Sun Han and Ruiqin Li and Yulong Zhang and Fangfu Liu and Xuesong Shi and Libin Liu and Yizhou Wang and Zhizheng Zhang and He Wang},
year = {2026},
eprint = {2607.06988},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2607.06988},
}
World Models世界模型world-action-modeltest-time-traininghuman-videosteering
SKIP generates only task-relevant keyframes (approach, contact, grasp, release) with a sparse video diffusion model, then fills the gaps with an action-conditioned interpolator. 4.16x faster dense rollouts on LIBERO.SKIP 只用稀疏视频扩散模型生成与任务相关的关键帧（接近、接触、抓取、释放），再用动作条件插值器补全中间帧。在 LIBERO 上生成稠密推演快 4.16 倍。
@misc{he2026skip,
title = {{SKIP: Sparse Keyframe Interpolation Paradigm for Efficient Embodied World Models}},
author = {Ziheng He and Yixiang Chen and Ning Yang and Zhanqian Wu and Qisen Ma and Yuan Xu and Jiabing Yang and Peiyan Li and Xiangnan Wu and Xiaofeng Wang and Zheng Zhu and Jing Liu and Nianfeng Liu and Yan Huang},
year = {2026},
eprint = {2606.00664},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.00664},
}
World Models世界模型world-modelvideo-generationkeyframesefficiencylibero
Learned dynamics that ignore control changes can hide out-of-distribution actions. A support-conditioned control-sensitivity regularizer keeps predictions responsive, improving OOD detection and safer planning.对控制变化不敏感的学习动力学会掩盖分布外动作。基于数据支撑的控制敏感性正则保持模型对控制的响应，从而提升分布外检测并使闭环规划更安全。
@misc{yu2026sensitivity,
title = {{Sensitivity Shaping for Latent Modeling}},
author = {Hongzhan Yu and Chenghao Li and Ruipeng Zhang and Henrik Christensen and Sicun Gao},
year = {2026},
eprint = {2606.14585},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2606.14585},
}
Detects robot anomalies from external cameras only by estimating dynamics and predicting states. On 240 real crane trajectories: +26% anomaly classification, and 88% accuracy locating the degrading motor.仅凭外部相机，通过估计动力学参数与预测状态来检测机器人异常。在 240 条真实起重机轨迹上异常分类提升 26%，定位退化电机的准确率 88%。
@misc{beddow2026adaptive,
title = {{Adaptive World-Models for Anomaly Recognition (AWARE): A Framework for Robot Dynamics Prediction and Introspection}},
author = {Luke Beddow and George Mavroghenis and Rodrigo Chacon Quesada and Kamil Dreczkowski and Cong Sun and Jiankai Wang and Oscar Kwong-Fai Pang and Antoine Cully},
year = {2026},
eprint = {2610.32742},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.32742},
}
World Models世界模型anomaly-detectionworld-modelcamera-onlysystem-identification
Training-free speedup for world-action models: reuses denoising residuals across consecutive inference chunks at the same step, since they correlate during smooth motion. Up to 2.5x with a Fast-WAM backbone.无需训练的世界动作模型加速方法：平滑运动时相邻推理块在同一去噪步的残差高度相关，因此跨块缓存复用。以 Fast-WAM 为骨干最高提速 2.5 倍。
DreamFree: a WAM that skips future-video generation. An encoder learns to predict a frozen video diffusion model's features from the current frame; actions come from those. Reports 11 Hz control with a 14B backbone.DreamFree：无需生成未来视频的世界动作模型。编码器从当前观测预测冻结视频扩散模型的表征，动作模块据此出动作；报告以 14B 视频扩散骨干实现 11Hz 实时闭环控制。
@misc{jo2026future,
title = {{Future Representation Distillation for Imagination-free World Action Models}},
author = {Jaehyeong Jo and Sangwon Jang and Sihyun Yu and Seonghyeon Ye and Yekyung Nah and Hojung Jung and Kimin Lee and Sung Ju Hwang},
year = {2026},
eprint = {2611.04527},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04527},
}
World Models世界模型world-action-modelvideo-diffusiondistillationreal-time
Paper论文 2026 · National University of Singapore new新 2026-11-10
Represents object motion as dense 3D trajectories over the whole surface, occluded parts included, for rigid and deformable objects. Adds a 1,210-hour flow dataset and a world model predicting it from image + task.把物体运动表示为覆盖整个表面（含遮挡部分）的稠密 3D 轨迹，适用于刚体与可形变物体。发布 1210 小时的全物体流数据集，并训练从图像和任务描述预测该流的世界模型。
@misc{chen2026objectworld,
title = {{ObjectWorld: Scaling Object-Centric World Models with A Large-Scale 3D Full-Object Flow Dataset}},
author = {Haonan Chen and Sicheng Isabella Lu and Haitian Li and Jiajin Wu and Hexiang Wei and Jiawei Zhang and Xin Fei and Yiwen Hou and Chenrui Tie and Chongkai Gao and Sheng Wang and Lin Shao},
year = {2026},
eprint = {2611.11085},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11085},
}
200K+ navigation-planning samples annotated from camera trajectories up to instructions; RL post-training inside a video world model, used as a cheap stand-in environment, reportedly helps on unseen scenes.包含 20 万余条导航规划样本的数据集，场景多样，标注从相机运动轨迹到高层语义指令；并把视频世界模型当作低成本交互环境做强化学习后训练，报告称在未见场景上有明显提升。
@misc{feng2026embodiedworld,
title = {{EmbodiedWorld-200K: Toward Embodied World Agents via Embodied-Planning Dataset and Interactive World Models}},
author = {X. Feng and J. Tang and Z. Lin and L. Chen and S. Shao and Q. Lu and K. Huang},
year = {2026},
eprint = {2610.25959},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2610.25959},
}
Two coupled JEPAs share a latent space: one proposes goal-conditioned latent intentions, the other predicts consequences, closing an intention-action-consequence loop. Evaluated on image-goal navigation in Habitat MP3D.两个耦合的 JEPA 共享隐空间：一个按目标提出隐式“意图”，另一个预测动作后果，形成意图-动作-后果闭环，让世界模型直接参与构造规划；在 Habitat MP3D 图像目标导航上评测。
@misc{zhou2026injepa,
title = {{InJePA: Intention-First Latent Planning with Coupled JEPA World Models for Visual Navigation}},
author = {Siming Zhou and Bowen Ma and Ruiran Du and Pengcheng Zheng and Sheng Zheng and Dongshen Han and Zhijie Li and Wei Dong and Fan Mo and Chaoning Zhang},
year = {2026},
eprint = {2610.23896},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.23896},
}
Dual-system video-action model for VLN: a slow system turns history and instruction into a compact latent plan, a fast one turns plan + latest view into local trajectories; no video is generated at inference.面向视觉语言导航的双系统视频-动作模型：慢系统根据历史观测与指令生成紧凑的隐式规划，快系统结合该规划与最新观测输出局部轨迹；推理时不生成未来视频。代码与权重称将在该仓库发布。
@misc{wang2026havana,
title = {{HAVANA: A Hierarchical Video-Action Model for Vision-Language Navigation}},
author = {Dongzhihan Wang and Yi Du and Chen Wang and Xiuxian Li and Liang Xu},
year = {2026},
eprint = {2610.24122},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.24122},
}
RepresentNav fuses several vision foundation models into a policy-learned space and predicts the next latent before acting (JEPA-style); reports 83.8% on VLNVerse fine-grained (+20.05 pts), real-robot tests.RepresentNav 用一个小型 Transformer 融合多个视觉基础模型各层特征，构建由导航目标自行塑造的表示空间，并在每步动作前预测下一时刻的潜在表示（JEPA 式）；报告 VLNVerse 细粒度成功率 83.8%（领先 20.05 个百分点），并在真机上验证。
@misc{chen2026see,
title = {{See What I Need, Dream What I Want: A Navigation Foundation Model on a Self-Constructed Representation Space}},
author = {Jiawei Chen and Yinxi Wang and Zihan Yang and Meijun Gao and Yaowei Liu and Shuming Zhang and Zecheng Li and Xiangju Mi and Yuxuan Liu and Runzhong Huang and Tianlin Li and Xun Zhang and Zhengyang Wang and Xu Miao and Tong Yang},
year = {2026},
eprint = {2610.31762},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.31762},
}
Benchmark of 800 human-rated robot/hand videos from 10 generators to test MLLM judges on physical adherence and instruction alignment, plus a 12K-video training set and RoboJudge-9B (Pearson 0.719 with humans).评测多模态大模型能否评判具身视频生成：800 条人工标注视频（10 个生成模型，夹爪/人手/灵巧手），考察物理合理性与指令一致性；另提供 1.2 万条训练视频及 RoboJudge-9B（与人类评分 Pearson 0.719）。
@misc{ma2026robojudge,
title = {{RoboJudge: Multimodal Language Models as Judges for Embodied Video Generation}},
author = {Siyuan Ma and Yu Qi and Cheng Ye and Ruotian Peng and Xinyi Xu and Zitiantao Lin and Zedong Cai and Yijian Huang and Weiyang Liu and Siyuan Wang and Yandong Wen},
year = {2026},
eprint = {2610.27782},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2610.27782},
}
Evaluates video world models by whether generated manipulation can be converted into actions and executed in physics simulation, not only whether it looks realistic.评估视频世界模型时不只看画面是否逼真，而是把生成的操作转成动作，在物理仿真中验证能否执行。
@misc{jiang2026robowm,
title = {{RoboWM-Bench: A Benchmark for Evaluating World Models in Robotic Manipulation}},
author = {Feng Jiang and Yang Chen and Kyle Xu and Yuchen Liu and Haifeng Wang and Zhenhao Shen and Jasper Lu and Shengze Huang and Yuanfei Wang and Chen Xie and Ruihai Wu},
year = {2026},
eprint = {2604.19092},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2604.19092},
}
@misc{nvidia2025cosmos,
title = {{Cosmos World Foundation Model Platform for Physical AI}},
author = {NVIDIA and : and Niket Agarwal and Arslan Ali and Maciej Bala and Yogesh Balaji and Erik Barker and Tiffany Cai and Prithvijit Chattopadhyay and Yongxin Chen and Yin Cui and Yifan Ding and Daniel Dworakowski and Jiaojiao Fan and Michele Fenzi and Francesco Ferroni and Sanja Fidler and Dieter Fox and Songwei Ge and Yunhao Ge and Jinwei Gu and Siddharth Gururani and Ethan He and Jiahui Huang and Jacob Huffman and Pooya Jannaty and Jingyi Jin and Seung Wook Kim and Gergely Klár and Grace Lam and Shiyi Lan and Laura Leal-Taixe and Anqi Li and Zhaoshuo Li and Chen-Hsuan Lin and Tsung-Yi Lin and Huan Ling and Ming-Yu Liu and Xian Liu and Alice Luo and Qianli Ma and Hanzi Mao and Kaichun Mo and Arsalan Mousavian and Seungjun Nah and Sriharsha Niverty and David Page and Despoina Paschalidou and Zeeshan Patel and Lindsey Pavao and others},
year = {2025},
eprint = {2501.03575},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2501.03575},
}
A self-supervised video encoder, post-trained with a little robot data, plans zero-shot pick-and-place by latent prediction.自监督视频编码器经少量机器人数据后训练，通过潜空间预测实现零样本抓放规划。
@misc{assran2025v,
title = {{V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning}},
author = {Mido Assran and Adrien Bardes and David Fan and Quentin Garrido and Russell Howes and Mojtaba and Komeili and Matthew Muckley and Ammar Rizvi and Claire Roberts and Koustuv Sinha and Artem Zholus and Sergio Arnaud and Abha Gejji and Ada Martin and Francois Robert Hogan and Daniel Dugas and Piotr Bojanowski and Vasil Khalidov and Patrick Labatut and Francisco Massa and Marc Szafraniec and Kapil Krishnakumar and Yong Li and Xiaodong Ma and Sarath Chandar and Franziska Meier and Yann LeCun and Michael Rabbat and Nicolas Ballas},
year = {2025},
eprint = {2506.09985},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2506.09985},
}
VPP — fine-tunes Stable Video Diffusion on robot and human videos and feeds its one-pass predictive features to a diffusion policy; reports 4.33 average length on CALVIN ABC→D and real Franka / dexterous-hand results.VPP：在机器人与人类操作视频上微调 Stable Video Diffusion，只做一次前向提取"预测性"视觉特征再交给扩散策略；论文报告 CALVIN ABC→D 平均完成长度 4.33，并有 Franka 与灵巧手真机实验。
@misc{hu2024video,
title = {{Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations}},
author = {Yucheng Hu and Yanjiang Guo and Pengchao Wang and Xiaoyu Chen and Yen-Jen Wang and Jianke Zhang and Koushil Sreenath and Chaochao Lu and Jianyu Chen},
year = {2024},
eprint = {2412.14803},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2412.14803},
}
UVA — one shared latent for video and actions with two light diffusion heads; masked training lets the same model act as policy, video generator and forward/inverse dynamics model, while keeping policy inference fast.UVA：视频与动作共享一个潜表示，配两个轻量扩散头解码；通过掩码训练，同一模型既能当策略，也能做视频生成和正/逆动力学，且策略推理仍然较快。
Genie Envisioner (AgiBot) — a video-diffusion world model (GE-Base) plus a flow-matching action decoder (GE-Act), an action-conditioned neural simulator (GE-Sim) and the EWMBench benchmark in one platform.智元 Genie Envisioner：一个平台集成视频扩散世界模型 GE-Base、流匹配动作解码器 GE-Act、动作条件神经仿真器 GE-Sim，以及世界模型评测基准 EWMBench。
@misc{liao2025genie,
title = {{Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation}},
author = {Yue Liao and Pengfei Zhou and Siyuan Huang and Donglin Yang and Shengcong Chen and Yuxin Jiang and Yue Hu and Jingbin Cai and Si Liu and Jianlan Luo and Liliang Chen and Shuicheng Yan and Maoqing Yao and Guanghui Ren},
year = {2025},
eprint = {2508.05635},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.05635},
}
Fine-tunes video world models on a robot to dream new demos, labels them with latent actions or an inverse dynamics model, and trains policies on them; a humanoid learns 20+ new behaviors from one-scene teleop data.把视频世界模型微调到目标机器人上“做梦”生成新示范，再用隐动作或逆动力学模型补上动作标签用于训练策略；人形机器人仅凭单一场景的遥操作数据学会 20 多种新行为。
@misc{jang2025dreamgen,
title = {{DreamGen: Unlocking Generalization in Robot Learning through Video World Models}},
author = {Joel Jang and Seonghyeon Ye and Zongyu Lin and Jiannan Xiang and Johan Bjorck and Yu Fang and Fengyuan Hu and Spencer Huang and Kaushil Kundalia and Yen-Chen Lin and Loic Magne and Ajay Mandlekar and Avnish Narayan and You Liang Tan and Guanzhi Wang and Jing Wang and Qi Wang and Yinzhen Xu and Xiaohui Zeng and Kaiyuan Zheng and Ruijie Zheng and Ming-Yu Liu and Luke Zettlemoyer and Dieter Fox and Jan Kautz and Scott Reed and Yuke Zhu and Linxi Fan},
year = {2025},
eprint = {2505.12705},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.12705},
}
Adds a few tokens to a VLA-style diffusion transformer and aligns them with latents of future observations, so the policy anticipates the future without a separate world model.只在 VLA 式扩散 Transformer 中加入少量 token，并让它们对齐未来观测的潜表示，使策略无需独立世界模型就能预判未来。
@misc{zheng2025flare,
title = {{FLARE: Robot Learning with Implicit World Modeling}},
author = {Ruijie Zheng and Jing Wang and Scott Reed and Johan Bjorck and Yu Fang and Fengyuan Hu and Joel Jang and Kaushil Kundalia and Zongyu Lin and Loic Magne and Avnish Narayan and You Liang Tan and Guanzhi Wang and Qi Wang and Jiannan Xiang and Yinzhen Xu and Seonghyeon Ye and Jan Kautz and Furong Huang and Yuke Zhu and Linxi Fan},
year = {2025},
eprint = {2505.15659},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2505.15659},
}
Uses a Veo-based generative system for robot policy evaluation across nominal performance, out-of-distribution generalization, and physical and semantic safety.用基于 Veo 的生成系统评估机器人策略，覆盖常规性能、分布外泛化以及物理和语义安全。
@misc{team2025evaluating,
title = {{Evaluating Gemini Robotics Policies in a Veo World Simulator}},
author = {Gemini Robotics Team and Krzysztof Choromanski and Coline Devin and Yilun Du and Debidatta Dwibedi and Ruiqi Gao and Abhishek Jindal and Thomas Kipf and Sean Kirmani and Isabel Leal and Fangchen Liu and Anirudha Majumdar and Andrew Marmon and Carolina Parada and Yulia Rubanova and Dhruv Shah and Vikas Sindhwani and Jie Tan and Fei Xia and Ted Xiao and Sherry Yang and Wenhao Yu and Allan Zhou},
year = {2025},
eprint = {2512.10675},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2512.10675},
}
Adds action conditioning to pretrained video generators to get learned world models for policy evaluation, as an alternative to costly real tests and hand-built simulation.给预训练视频生成模型加入动作条件，得到可用于策略评估的世界模型，替代昂贵的真机测试和手工搭建的仿真。
@misc{tseng2025scalable,
title = {{Scalable Policy Evaluation with Video World Models}},
author = {Wei-Cheng Tseng and Jinwei Gu and Qinsheng Zhang and Hanzi Mao and Ming-Yu Liu and Florian Shkurti and Lin Yen-Chen},
year = {2025},
eprint = {2511.11520},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2511.11520},
}
Predicts future states by propagating Gaussian primitives under robot actions with a latent DiT and 3D VAE, usable for representation learning and as a neural simulator.用潜空间 DiT 和 3D VAE 预测高斯基元在机器人动作下的传播，可用于表征学习，也可作为神经模拟器。
@misc{lu2025gwm,
title = {{GWM: Towards Scalable Gaussian World Models for Robotic Manipulation}},
author = {Guanxing Lu and Baoxiong Jia and Puhao Li and Yixin Chen and Ziwei Wang and Yansong Tang and Siyuan Huang},
year = {2025},
eprint = {2508.17600},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2508.17600},
}
A Transformer point cloud world model for interactions among rigid, deformable and flexible materials, trained from real perception data without elaborate scene reconstruction.Transformer 点云世界模型，可描述刚体、可变形和柔性材料之间的交互，直接用真实感知数据训练，无需复杂场景重建。
@misc{huang2025particleformer,
title = {{ParticleFormer: A 3D Point Cloud World Model for Multi-Object, Multi-Material Robotic Manipulation}},
author = {Suning Huang and Qianzhong Chen and Xiaohan Zhang and Jiankai Sun and Mac Schwager},
year = {2025},
eprint = {2506.23126},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2506.23126},
}
Names two failure modes of latent safe control, estimation gaps (safety facts not visible) and prediction gaps (failures not foreseeable), gives a diagnostic for each and mitigations.指出潜空间安全控制的两种失效：估计缺口（当前观测看不到安全关键量）与预测缺口（失败无法提前预见），并分别给出诊断指标和缓解方法。
@misc{kim2025how,
title = {{How Well Do Latent World Models Understand Partially Observable Safety Constraints?}},
author = {Matthew Kim and Kensuke Nakamura and Andrea Bajcsy},
year = {2025},
eprint = {2510.06492},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2510.06492},
}
World Models世界模型latent-world-modelsafetypartial-observabilityanalysis
A 3D-aware LLM with scene, object and action tokens, aligned with diffusion models so it can imagine goal images and point clouds before planning actions; trained on ~2M 3D-language-action pairs the authors assembled.具备三维感知的大语言模型，引入场景、物体与动作交互 token，并与扩散模型对齐，可在规划动作前“想象”目标图像和点云；训练数据为作者整理的约 200 万条三维-语言-动作样本。
@misc{zhen2024d,
title = {{3D-VLA: A 3D Vision-Language-Action Generative World Model}},
author = {Haoyu Zhen and Xiaowen Qiu and Peihao Chen and Jincheng Yang and Xin Yan and Yilun Du and Yining Hong and Chuang Gan},
year = {2024},
eprint = {2403.09631},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2403.09631},
}
Pretrained to generate future video on 38M clips, then fine-tuned to predict frames and actions together; reports 97.7% average success over 100+ real tabletop tasks and roughly double GR-1's success in unseen scenes.先在 3800 万段视频上预训练未来帧生成，再在机器人数据上微调，同时预测视频帧与动作轨迹；报告在 100 多个真机桌面任务上平均成功率 97.7%，在未见场景中成功率约为 GR-1 的两倍。
@misc{cheang2024gr,
title = {{GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulation}},
author = {Chi-Lam Cheang and Guangzeng Chen and Ya Jing and Tao Kong and Hang Li and Yifeng Li and Yuxiao Liu and Hongtao Wu and Jiafeng Xu and Yichu Yang and Hanbo Zhang and Minzhao Zhu},
year = {2024},
eprint = {2410.06158},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2410.06158},
}
Seer forecasts the next visual state and feeds it to an inverse dynamics model that outputs actions, trained end to end and pretrained on DROID; reports 4.28 avg. length on CALVIN ABC-D plus LIBERO and real-robot gains.Seer 先预测未来视觉状态，再由逆动力学模型据此输出动作，端到端训练并在 DROID 上预训练；报告 CALVIN ABC-D 平均任务长度 4.28，在 LIBERO-LONG 和真机任务上也有提升。
@misc{tian2024predictive,
title = {{Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation}},
author = {Yang Tian and Sizhe Yang and Jia Zeng and Ping Wang and Dahua Lin and Hao Dong and Jiangmiao Pang},
year = {2024},
eprint = {2412.15109},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2412.15109},
}
A recurrent graph network estimates object state and latent physics from visuo-tactile history; the learned dynamics model drives MPC for tasks like dense packing.循环图网络从视触觉历史估计物体状态与潜在物理属性，学到的动力学模型用于 MPC，完成密集装箱等任务。
@misc{ai2024robopack,
title = {{RoboPack: Learning Tactile-Informed Dynamics Models for Dense Packing}},
author = {Bo Ai and Stephen Tian and Haochen Shi and Yixuan Wang and Cheston Tan and Yunzhu Li and Jiajun Wu},
year = {2024},
eprint = {2407.01418},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2407.01418},
}
@misc{bruce2024genie,
title = {{Genie: Generative Interactive Environments}},
author = {Jake Bruce and Michael Dennis and Ashley Edwards and Jack Parker-Holder and Yuge Shi and Edward Hughes and Matthew Lai and Aditi Mavalankar and Richie Steigerwald and Chris Apps and Yusuf Aytar and Sarah Bechtle and Feryal Behbahani and Stephanie Chan and Nicolas Heess and Lucy Gonzalez and Simon Osindero and Sherjil Ozair and Scott Reed and Jingwei Zhang and Konrad Zolna and Jeff Clune and Nando de Freitas and Satinder Singh and Tim Rocktäschel},
year = {2024},
eprint = {2402.15391},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2402.15391},
}
Predicts future DINOv2 patch features instead of pixels, so a world model trained on offline data can plan to new visual goals at test time.预测未来的 DINOv2 图像块特征而非像素，使基于离线数据训练的世界模型能在测试时规划到新的视觉目标。
@misc{zhou2024dino,
title = {{DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning}},
author = {Gaoyue Zhou and Hengkai Pan and Yann LeCun and Lerrel Pinto},
year = {2024},
eprint = {2411.04983},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2411.04983},
}
GR-1 — a GPT-style model that predicts both actions and future frames; pre-training on Ego4D video prediction before robot fine-tuning lifts CALVIN unseen-scene success from 53.3% to 85.4% (as reported).GR-1：GPT 式模型同时预测动作与未来画面；先在 Ego4D 上做视频预测预训练、再用机器人数据微调，论文报告 CALVIN 未见场景成功率从 53.3% 提升到 85.4%。
@misc{wu2023unleashing,
title = {{Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulation}},
author = {Hongtao Wu and Ya Jing and Chilam Cheang and Guangzeng Chen and Jiafeng Xu and Xinghang Li and Minghuan Liu and Hang Li and Tao Kong},
year = {2023},
eprint = {2312.13139},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2312.13139},
}
Hierarchical world model for visual whole-body humanoid control: a high-level agent commands a low-level one, with no reward design or skill primitives.用于视觉全身人形控制的分层世界模型：高层智能体指挥低层智能体，不依赖奖励设计或技能基元。
@misc{hansen2024hierarchical,
title = {{Hierarchical World Models as Visual Whole-Body Humanoid Controllers}},
author = {Nicklas Hansen and Jyothir S and Vlad Sobal and Yann LeCun and Xiaolong Wang and Hao Su},
year = {2024},
eprint = {2405.18418},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2405.18418},
}
UniSim — an action-conditioned video model used as a simulator to train policies that transfer to the real world.UniSim：以动作为条件的视频生成模型充当仿真器，训练出的策略可迁移到真实世界。
@misc{yang2023learning,
title = {{Learning Interactive Real-World Simulators}},
author = {Sherry Yang and Yilun Du and Kamyar Ghasemipour and Jonathan Tompson and Leslie Kaelbling and Dale Schuurmans and Pieter Abbeel},
year = {2023},
eprint = {2310.06114},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2310.06114},
}
An implicit (decoder-free) world model planned with MPC; one set of hyperparameters works across many continuous-control tasks and scales to a multitask agent.无解码器的隐式世界模型配合 MPC 规划；同一组超参数适用于大量连续控制任务，并可扩展为多任务智能体。
IRIS — a discrete autoencoder turns frames into tokens and a Transformer predicts the next tokens; the agent learns in imagination with only ~2 hours of Atari gameplay.IRIS：离散自编码器把画面变成 token，Transformer 预测后续 token；智能体只用约 2 小时的 Atari 游戏经验便在想象中学会策略。
@misc{micheli2022transformers,
title = {{Transformers are Sample-Efficient World Models}},
author = {Vincent Micheli and Eloi Alonso and François Fleuret},
year = {2022},
eprint = {2209.00588},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2209.00588},
}
UniPi — casts decision making as text-conditioned video generation, then recovers actions from the generated frames with an inverse dynamics model.UniPi：把决策问题转化为文本条件的视频生成，再用逆动力学模型从生成的画面中恢复动作。
@misc{du2023learning,
title = {{Learning Universal Policies via Text-Guided Video Generation}},
author = {Yilun Du and Mengjiao Yang and Bo Dai and Hanjun Dai and Ofir Nachum and Joshua B. Tenenbaum and Dale Schuurmans and Pieter Abbeel},
year = {2023},
eprint = {2302.00111},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2302.00111},
}
@misc{wu2022daydreamer,
title = {{DayDreamer: World Models for Physical Robot Learning}},
author = {Philipp Wu and Alejandro Escontrela and Danijar Hafner and Ken Goldberg and Pieter Abbeel},
year = {2022},
eprint = {2206.14176},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2206.14176},
}
World Models世界模型Quadruped / legged四足 / 足式realquadrupedarmmodel-based-rl
TD-MPC — a task-oriented latent model used for short-horizon planning, with a learned value function covering the long-term return; the predecessor of TD-MPC2.TD-MPC：用面向任务的潜空间模型做短视距规划，由学习到的价值函数估计长期回报；是 TD-MPC2 的前身。
The first Dreamer — learns a latent world model from pixels and trains the policy entirely on imagined rollouts.第一代 Dreamer：从像素学习潜空间世界模型，并完全在想象的轨迹上训练策略。
@misc{hafner2019dream,
title = {{Dream to Control: Learning Behaviors by Latent Imagination}},
author = {Danijar Hafner and Timothy Lillicrap and Jimmy Ba and Mohammad Norouzi},
year = {2019},
eprint = {1912.01603},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/1912.01603},
}
The classic VAE + RNN agent that learns a compressed model of its environment and trains a controller inside the "dream".经典的 VAE + RNN 智能体，学习环境的压缩模型并在"梦境"中训练控制器。
Yu Zhou, Xinyuan Xia, Xuankun Yang, Yinghao Yang et al.
Abstract摘要
World action models jointly predict video and actions conditioned on language. Existing studies focus primarily on architectures and training mechanisms, typically treating language-video alignment acquired through video pretraining as an established foundation. However, introducing action learning may impair this alignment. Principled guidance on data and supervision design for maintaining language-video alignment during action learning remains limited. To investigate this, we curate a pool of approximately 47,000 hours of general videos, egocentric videos, UMI demonstrations, and robot …
@misc{zhou2026swam,
title = {{SWAM: Scaling World Action Models with Cross-Modal Alignment}},
author = {Yu Zhou and Xinyuan Xia and Xuankun Yang and Yinghao Yang and Wendi Chen and Bowen Liu and Jianhan Ma and Weicheng Duan and others},
year = {2026},
eprint = {2611.12870},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.12870},
}
George Nehma, Patrick Quinn, Madhur Tiwari, Camilo Riano-Rios
Abstract摘要
Soft powered descent landings are becoming evermore important in the field of space exploration and scientific advancement, because of their ability to deliver mission critical payloads, both highly accurately and safely. The high efficiency restriction on planetary landings due to the severely limited fuel supply results in the need for optimal guidance trajectories and control. Model based reinforcement learning is proving to be an important tool in simplifying the highly complex process of generating optimal guidance trajectories and control. In this work, we present DreamLander, a state …
@misc{nehma2026dreamlander,
title = {{DreamLander: A 6-DOF World Model based Reinforcement Learning Planetary Lander}},
author = {George Nehma and Patrick Quinn and Madhur Tiwari and Camilo Riano-Rios},
year = {2026},
eprint = {2611.12824},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12824},
}
Xinrui Zu, Michael Przystupa, Zhao Yang, Shujian Yu et al.
Abstract摘要
World models have increasingly been applied as a model-based control method for robotics tasks such as manipulation and locomotion. While early world model architectures relied on decoders to prevent mode collapse and learn a predictable latent space, recent works on world models, such as Joint Embedding Predictive Architecture or LeWorldModel, increasingly aim to learn embeddings without a grounding decoder, utilising only a latent forward dynamics predictor and a stabilising entropy term. In this paper, we set out to demonstrate that the further introduction of two additional latent losses …
@misc{zu2026lewm,
title = {{$Σ$-LeWM: Learning World Models via S-Information for Robot Control}},
author = {Xinrui Zu and Michael Przystupa and Zhao Yang and Shujian Yu and Kevin Sebastian Luck},
year = {2026},
eprint = {2611.12732},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2611.12732},
}
Hasaan Ahmad, Qizhen Ying, Yuxuan Sun, Tianjin Huang et al.
Abstract摘要
A world model predicts how a scene changes after an action, and a planner uses it to search for actions that reach a goal. If the camera moves or the controller changes, the same command moves the robot differently: when its commands are turned by 90 degrees, a frozen planner with the LeWM world model that succeeds in 77.5% of episodes succeeds in only 7.8%. Current fixes retrain the model, need training data, or forget a set-up when it returns. We propose Library of Hands (LoH), a training-free method that keeps the world model frozen and corrects only the map from commands to applied …
@misc{ahmad2026library,
title = {{Library of Hands: Fix the Hands, Not the World Model}},
author = {Hasaan Ahmad and Qizhen Ying and Yuxuan Sun and Tianjin Huang and Guangming Wang},
year = {2026},
eprint = {2611.12678},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.12678},
}
One-step stochastic world models avoid the mode averaging of deterministic feature predictors without the tens of network evaluations per step that diffusion and flow matching require. The simplest such objective is nearest-of-m matching (conditional implicit maximum likelihood estimation, IMLE): draw m candidate futures and pull only the one closest to the observed outcome. Recent work conjectures that a small m suffices for feature-space world models and that a strictly proper energy score should replace the hard minimum. We test both claims with one-step world models on frozen DINOv2/v3 …
@misc{song2026how,
title = {{How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens}},
author = {Yong Hwi Song and Keon Myung Lee},
year = {2026},
eprint = {2611.11887},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2611.11887},
}
Spatial generalization in robot world models depends on whether action representations capture the physical relationships that remain valid across changing spatial configurations. Yet different coordinates can describe the same interaction, whereas similar motions can produce different outcomes when local physical conditions change. How these distinctions affect learned prediction remains insufficiently understood. This study evaluates matched action representations defined in world, object, and surface coordinate frames using a controlled robotic peeling benchmark. Global pose and contact …
Sampling-based planners such as MPPI and CEM query a learned world model thousands of times per control step, so low-bit world models are an obvious way to make model-based control cheaper on a robot. With round-to-nearest activation quantization and no calibration, we find that planning can collapse, by up to 100 points of success, while the world model's latent one-step prediction error stays no larger than that of a stochastic rounding scheme that plans well. At the same bit width, what decides the outcome is not the size of the rounding error but whether it persists across planner …
Yunyi Gao, Yixing Shi, Yize Liu, Hai-Duong Tran et al.
Abstract摘要
Quadrotors inspecting buildings and tunnels must navigate unknown clutter with a forward-facing depth camera whose limited field of view constrains scene understanding and collision-avoiding navigation. Recent world-model agents learn navigation through imagined experience (e.g., AirDreamer), but remain data-intensive and often lack explicit geometric memory and awareness of what the camera can observe. We propose ScanDreamer, a map-grounded world model that constructs an online 2D scan-memory map from depth images and vehicle pose. Our world model, actor, and critic are conditioned on this …
@misc{gao2026scandreamer,
title = {{ScanDreamer: A Map-Grounded World Model for Quadrotor Navigation}},
author = {Yunyi Gao and Yixing Shi and Yize Liu and Hai-Duong Tran and Nhat-Hoang P. Nguyen and Shuai Wang and Yi Deng and Shengming Liu and others},
year = {2026},
eprint = {2611.11605},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11605},
}
Santosh Mohan Rajkumar, Sriram Narayanan, Samuel E. Otto, Debdipta Goswami
Abstract摘要
World models learn internal simulators of environment dynamics for prediction, planning, and control, yet learning them from limited, noisy, and partial observations remains challenging. While much progress has focused on pixel-based observations, many physical systems instead provide vector-valued measurements that reveal only part of the underlying state. We introduce K-PWM, a probabilistic world model for prediction and control under partial observation. K-PWM couples Koopman-bilinear latent dynamics with a nonlinear observation decoder and performs probabilistic latent-state inference …
@misc{rajkumar2026k,
title = {{K-PWM: Control-Oriented Structured Bilinear World Models under Partial Observation}},
author = {Santosh Mohan Rajkumar and Sriram Narayanan and Samuel E. Otto and Debdipta Goswami},
year = {2026},
eprint = {2611.11521},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11521},
}
Zihao Sheng, Zilin Huang, Yuhao Luo, Zhengyang Wan et al.
Abstract摘要
World models allow autonomous vehicles to imagine the consequences of candidate actions before execution, but their reliability depends on the futures represented in their training data. We identify a survivorship bias in driving world models: because the futures recorded in prevailing driving datasets almost always end safely, existing models often imagine benign outcomes even when a collision is imminent. We introduce a neutral world model (NWM) that learns both routine and safety-critical futures from a mixture of routine driving and real accident videos with a bounding-box-guided …
@misc{sheng2026breaking,
title = {{Breaking survivorship bias in driving world models}},
author = {Zihao Sheng and Zilin Huang and Yuhao Luo and Zhengyang Wan and Boyue Wang and Soyoung Ahn and David A. Noyce and Sikai Chen},
year = {2026},
eprint = {2611.11516},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11516},
}
Zhongxi Chen, Yi Yang, Xinyi Chen, Mingzhe Ye et al.
Abstract摘要
Dexterous manipulation requires rich spatial foresight over both the surrounding scene and the robot hand. Existing world-action models often acquire such predictive knowledge through explicit future prediction. We introduce DexForesight, a framework that learns spatial foresight implicitly by injecting dynamics-aware world representations into the latent space of a dexterous policy. During training, DexForesight learns an action-conditioned spatial world model from offline scene depth, kinematic hand geometry, and demonstration trajectories. The world model predicts 3D hand-point motion and …
@misc{chen2026dexforesight,
title = {{DexForesight: Grounding Dexterous Policies with Implicit Spatial Foresight}},
author = {Zhongxi Chen and Yi Yang and Xinyi Chen and Mingzhe Ye and Yaping Li and Zhihong Liu and Congsheng Xu and Haonan Wang and others},
year = {2026},
eprint = {2611.11166},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11166},
}
Generative world modeling from large-scale visual data offers a promising path toward scalable robot manipulation. This calls for an intermediate representation that can be supervised at scale and remains actionable for downstream control. We propose 3D full-object flow, an object-centric motion representation that assigns dense 3D trajectories to points across the complete surface of a target object, including its occluded regions. By extending visible-surface object flow to the full object geometry, this representation makes complete object motion explicit and applies uniformly to both …
@misc{chen2026objectworld,
title = {{ObjectWorld: Scaling Object-Centric World Models with A Large-Scale 3D Full-Object Flow Dataset}},
author = {Haonan Chen and Sicheng Isabella Lu and Haitian Li and Jiajin Wu and Hexiang Wei and Jiawei Zhang and Xin Fei and Yiwen Hou and others},
year = {2026},
eprint = {2611.11085},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.11085},
}
Dong Fang, Feixuan Huang, Yuanxin Zhong, Yuxiang Gao et al.
Abstract摘要
Dexterous manipulation with world action models (WAMs) relies on hand configuration and contact information, yet joint-state and whole-hand tactile observations may be partially or entirely missing. RGB images provide scene context and hand-object geometry but are insufficient to reliably recover these missing modalities. We further observe a correlation between hand state and tactile information. Motivated by this observation, we introduce BAT-WAM, a framework that bridges visual, joint-state, and tactile representations for WAMs. BAT-WAM learns separate codebooks and decoders for the …
@misc{fang2026bat,
title = {{BAT-WAM: Bridging Action and Touch in World-Action Models for Dexterous Hands}},
author = {Dong Fang and Feixuan Huang and Yuanxin Zhong and Yuxiang Gao and Qi Song and Yunlong Wang and Xiaosong Jia and Yu-Gang Jiang},
year = {2026},
eprint = {2611.10967},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10967},
}
Yinuo Jiang, Guangzhi Wang, Yiran Wang, Junfu Pu et al.
Abstract摘要
Latent world models enable visual control by planning action sequences in compact representation spaces, but a gap remains between how latent dynamics are learned and how they are used for planning. First, teacher-forced training does not expose the dynamics model to the recursively predicted latent states encountered during planning. Second, sampling-based planners typically initialize action search from goal-agnostic proposals, without exploiting the action-conditioned dynamics learned by the world model. We introduce Dynamics-Informed Action Learning (DIAL) to bridge these two gaps. During …
@misc{jiang2026dynamics,
title = {{Dynamics-Informed Action Learning for Goal-Conditioned Latent Planning}},
author = {Yinuo Jiang and Guangzhi Wang and Yiran Wang and Junfu Pu and Cheng Cheng and Honghao Chen and Cheng Cheng and Ying Shan},
year = {2026},
eprint = {2611.10660},
archivePrefix = {arXiv},
primaryClass = {cs.CV},
url = {https://arxiv.org/abs/2611.10660},
}
Autonomous robot teams deployed in remote, unstructured environments may face changes in the environment, hardware, and team composition that degrade previously effective policies. Shifting mission objectives can be handled by selecting a different solution from an existing Pareto set, but shifting deployment conditions can reorder the set itself, making previously dominated policies competitive or new policies necessary. However, evaluating these policies in new contexts while minimizing risk to the hardware remains an open problem. We argue that, although largely untested in this domain, …
@misc{santos2026multi,
title = {{Multi-Objective Adaptation After Deployment: A Position on World Models for Remote Robot Teams}},
author = {Jamie Santos and Kagan Tumer},
year = {2026},
eprint = {2611.10435},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10435},
}
Embedded AI is moving from perception to prediction. Robots now use world models$\unicode{x2014}$learned simulators that predict the outcome of candidate actions in a compressed latent space before any is executed. Every predicted step must be computed on board within a tight power and thermal budget, yet the energy these steps fundamentally require, and how far real hardware sits from it, have not been characterized. We derive lower bounds on the arithmetic operations and the CMOS dynamic energy of a latent state transition from information theory and Boolean circuit complexity. Profiling a …
@misc{ramkumar2026world,
title = {{The World Model Planning Accelerator (WMPA): Hardware-Algorithm Co-Design for Embedded AI in Robotics}},
author = {Prerana Ramkumar and Suhas Suresh Bharadwaj and Fazal Hameed},
year = {2026},
eprint = {2611.10316},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.10316},
}
Chen Chen, Omid Bagherieh, Naveen Nikamanth Appiah Balaji, Jesse T. Gonzalez et al.
Abstract摘要
Recent work on robot manipulation has focused on specialist policies trained on demonstrations. We show that generalist AI models can control manipulators more reliably than specialist policies such as vision-language-action models (VLAs) and world-action models (WAMs). We systematically analyze agentic manipulation policies (AMPs), in which a generalist AI model commands the manipulator directly, adapting its actions as the task unfolds. We present AMP1, a reference harness that runs unchanged across single-arm, bimanual, and mobile manipulation. AMP1 achieves state-of-the-art results on the …
@misc{chen2026agentic,
title = {{On Agentic Manipulation Policies}},
author = {Chen Chen and Omid Bagherieh and Naveen Nikamanth Appiah Balaji and Jesse T. Gonzalez and Negin Heravi and Fahad Islam and Xueting Li and Jean-Baptiste Regli and others},
year = {2026},
eprint = {2611.09892},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.09892},
}
Recently, world-action models (WAMs) have emerged as a powerful framework for robot manipulation. By jointly modeling action generation and future observation prediction, WAMs can act and anticipate the outcomes of the actions. However, existing WAMs typically rely on frozen pretrained encoders trained with reconstruction, which may not produce representations well suited for robot control. In this paper, we tackle the problem of learning physicallygrounded, control-relevant representations for WAMs. We propose MDF++, a visuo-tactile world-action model that learns latent representations of …
Robert Ren, Matthew Bronars, He Zhu, Yikai Tang et al.
Abstract摘要
We introduce GPWM, a Generative Particle World Model that predicts how 3D scenes evolve by generating the motion of particles and rigid bodies over time. Unlike video world models that generate scene appearance, GPWM operates directly in 3D physical space. A single representation and architecture cover rigid, deformable, cloth, and granular materials, in both passive physics and robotic interaction. GPWM is an object-aware multi-resolution transformer that combines flow matching with diffusion forcing to generate coherent physical trajectories far beyond its training horizon. We train and …
@misc{ren2026gpwm,
title = {{GPWM: Modeling the World as 3D Particle Motion}},
author = {Robert Ren and Matthew Bronars and He Zhu and Yikai Tang and Jianing Zhang and Sergey Zakharov and Pavel Tokmakov and Katerina Fragkiadaki},
year = {2026},
eprint = {2611.08338},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08338},
}
Mateo Cl\'emente, George Z. Xue, Amin Abyaneh, Liheng Ma et al.
Abstract摘要
World-action models adapt pretrained video generators, and vision-language-action models build on pretrained vision-language backbones; both tie policy architecture and inference cost to large transformers designed for generation rather than control. It remains unclear how much these generation-oriented choices affect control. In this work, we introduce Latent Action JEPA (La-JEPA), a compact world-action model whose policy transformer is trained from scratch in the feature space of a pretrained V-JEPA 2.1 encoder. Following Fast-WAM, La-JEPA learns to predict future features jointly with …
@misc{clemente2026la,
title = {{La-JEPA: Compact Robot Policies with Latent World Modeling}},
author = {Mateo Cl\'emente and George Z. Xue and Amin Abyaneh and Liheng Ma and Rui Heng Yang and Tongtong Cao and Zhanguang Zhang and Yingxue Zhang},
year = {2026},
eprint = {2611.08156},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08156},
}
Zihao Zhang, Xinyu Zhang, Bing Zeng, Xingyu Chen et al.
Abstract摘要
World-action models (WAMs) jointly model future traffic states and driving actions for end-to-end autonomous driving in intelligent transportation systems. However, their action learning remains dominated by expert-trajectory imitation, leaving unsafe alternatives close to expert trajectories insufficiently constrained. To improve WAM robustness in safety-critical scenarios, we propose NegWAM, a safety post-training framework for unified WAMs that learns explicit safety boundaries from near-expert failure trajectories. NegWAM constructs negative trajectories that are geometrically close to …
@misc{zhang2026reinforcing,
title = {{Reinforcing End-to-End Autonomous Driving Safety via Negative-Aware Planning in a Unified World-Action Model}},
author = {Zihao Zhang and Xinyu Zhang and Bing Zeng and Xingyu Chen and Zhenning Li and Bin Yu and Ziyuan Pu},
year = {2026},
eprint = {2611.08111},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.08111},
}
Xiyan Su, Finn Rasmus Sch\"afer, Hojun Lim, Frank Diermeyer et al.
Abstract摘要
World action models (WAMs) for autonomous driving couple the driving policy and the visual world model in a single generative network whose objectives are fixed post-training. Conditioning on a new driving objective requires retraining. We study energy guidance as a general mechanism and present GuidedWAM, a model-agnostic framework for injecting driving objectives into frozen WAMs at sampling time. Our framework constructs each objective as differentiable energies on the sampler's intermediate clean-action estimate, which needs only forward access to the frozen model, and is stabilized by …
@misc{su2026guidedwam,
title = {{GuidedWAM: Training-Free Energy Guidance for World Action Models in Autonomous Driving}},
author = {Xiyan Su and Finn Rasmus Sch\"afer and Hojun Lim and Frank Diermeyer and Johannes Betz and Markus Lienkamp},
year = {2026},
eprint = {2611.07912},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07912},
}
Robot world models predict action-conditioned future visual states to support robotic planning and decision making. However, visual reconstruction does not fully capture task progress, semantic evaluation alone provides limited constraints on fine-grained visual errors, and homogeneous predictions limit comparative feedback. We present RoboImagine, a framework for post-training autoregressive robotic world models through candidate-future comparisons with semantic and visual feedback. Diverse search promotes diversity among candidate trajectories under shared visual observations, action …
@misc{wu2026learning,
title = {{Learning Robotic World Models with Semantic and Visual Feedback}},
author = {Hao Wu and Yuqi Li and Yizhou Zhao and Kun Wang and Xiao Luo and Yu Cheng and Bo An and Yingli Tian and others},
year = {2026},
eprint = {2611.07850},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07850},
}
Weiduo Yuan, Ze Yuan, Sihan Liu, Jionghao Wang et al.
Abstract摘要
The Universal Manipulation Interface (UMI) enables scalable demonstration collection independently of a specific robot. However, its moving, close-range gripper cameras entangle camera and object motion without a stable external view, leaving the metric geometry needed for precise manipulation implicit in RGB. We introduce Spatial-UMI-Dream, a 5B spatial World Action Model (WAM) pretrained on 605 hours of RGB-D UMI data through two-stage training -- world-only adaptation followed by joint world-action learning -- to jointly predict future RGB, fisheye-aligned metric depth, and actions. …
@misc{yuan2026spatial,
title = {{Spatial-UMI-Dream: Scaling World Action Models with Metric Geometry for UMI Manipulation}},
author = {Weiduo Yuan and Ze Yuan and Sihan Liu and Jionghao Wang and Liang Heng and Zhenyu Zhao and Yue Wang},
year = {2026},
eprint = {2611.07229},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.07229},
}
Jun Liu, Dajie Sun, Zhenglun Kong, Zhenjiang Fan et al.
Abstract摘要
World-action models increasingly apply test-time scaling, spending extra inference computation on candidate selection and adaptive allocation to improve control. Such gains are credible only when candidates carry exploitable signal, nominal compute savings become real speedup, and evaluation stays reliable. We audit all three on Cosmos Policy and RoboCasa with a premise-first, pre-registered methodology. On clean data, a learned value selector is 1.66 pp worse than random; a hindsight oracle gains only 1.40 pp, with an episode-level interval spanning zero; and 82% of states show identical …
@misc{liu2026when,
title = {{When Does Test-Time Search Help World Action Models? An Empirical Audit of Candidate Selection and Simulator Integrity}},
author = {Jun Liu and Dajie Sun and Zhenglun Kong and Zhenjiang Fan and Peiyan Dong and Fan Yang and Wenbin Zhang and Geng Yuan and others},
year = {2026},
eprint = {2610.41970},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.41970},
}
A robot action can satisfy shape and range constraints yet carry the wrong semantics: axes may be permuted, signs inverted, rotations rescaled, or gripper polarity reversed. We study learned state-action-effect relations as offline contracts for detecting these mismatches and ranking their likely fields. Clean held-out transitions calibrate alarms; a fixed inverse vocabulary supports closed-set localization. On ten LIBERO Spatial tasks comprising 432 source episodes, a capacity-controlled comparison at 2.5% empirical clean false-positive rate gives 46.1% recall for a ridge forward residual, …
Joint-Embedding Predictive Architectures (JEPAs) support planning by comparing predicted latent states with an encoded goal. For this comparison to guide temporal progress, states closer in time to the goal should receive lower costs. However, training representations for prediction and distribution regularization does not directly enforce this ordering. We examine this mismatch in flat LeWorldModel (LeWM) by analyzing how its Euclidean goal cost changes with temporal separation. Our analysis identifies conditions under which states at different long offsets receive similar average costs, …
Large language models understand passengers' driving commands, but as driving agents they reason in generated text for seconds and must infer each action's consequences implicitly from the scene. We present Jev-AD, which pairs Jev, a typed decision model that answers predefined questions with probabilities, not generated text, with a language world model that predicts for every candidate action how nearby vehicles move over the next two seconds. Code turns these predictions into foreseen consequences, which Jev compares instead of inferring them from coordinates. In simulated highway merging …
@misc{qu2026jev,
title = {{Jev with Foresight: Safe Language-Guided Autonomous Driving via a Language World Model}},
author = {Yansong Qu and Jiancong Chen and Samuel Labi},
year = {2026},
eprint = {2611.05448},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05448},
}
Although existing candidate trajectory evaluators have substantially improved end-to-end planning accuracy, their underlying mechanism remains suboptimal. Model-based evaluators typically make scoring future-aware by explicitly predicting candidate-conditioned future states over the planning horizon, which we term rollout. However, rollout incurs high inference cost and accumulates prediction errors. We argue that planning does not require reconstructing a full future scene for each candidate, but only future-relevant information for reliable comparison. We therefore propose RFWorld, a …
@misc{wang2026trajectory,
title = {{Trajectory Evaluation via Rollout-Free World Model for End-to-end Autonomous Driving}},
author = {Junzhe Wang and Jinqing Zhang and Qingjie Liu},
year = {2026},
eprint = {2611.05072},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05072},
}
Xingyu Wang, Yanjiang Guo, Zhijing Ding, Jiaxin Fang et al.
Abstract摘要
World Action Models (WAMs) improve robotic manipulation by jointly modeling visual dynamics and robot actions. However, most existing WAMs model dense short-horizon visual dynamics, capturing local interactions but providing limited long-range task context. Extending prediction horizons increases uncertainty and computation while entangling task-level evolution with fine-grained local dynamics, obscuring their dependencies and distinct roles in action generation. We propose HiLo-WAM, which decouples long-horizon task evolution from local visual dynamics while jointly modeling their …
@misc{wang2026hilo,
title = {{HiLo-WAM: Decoupling Long-Horizon Task Evolution and Local Dynamics in World Action Models}},
author = {Xingyu Wang and Yanjiang Guo and Zhijing Ding and Jiaxin Fang and Zeyu Lai and Jiayu Dong and Weikang Wan and Jianyu Chen and others},
year = {2026},
eprint = {2611.05023},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.05023},
}
Travis Davies, Wentao Ma, Lihao Guo, Tengfei Ma et al.
Abstract摘要
A robot world model that serves as a simulator must do more than generate plausible videos: it must faithfully follow commanded actions, maintain consistency across camera views, and sustain coherent long-horizon rollouts. We present PrimeSim, an action-conditioned, multiview world model simulator that addresses these challenges through progressive midtraining and posttraining. Starting from a pretrained video generator, we midtrain on 30,110 hours of robot demonstrations spanning 22 platforms. To improve action following, we introduce a unified cross-embodiment action representation and …
@misc{davies2026primesim,
title = {{PrimeSim: A Generalist World Model Simulator from Large-Scale Cross-Embodiment Demonstrations}},
author = {Travis Davies and Wentao Ma and Lihao Guo and Tengfei Ma and Xiaowen Chang and Jiafeng Xu and Shilong Wang and Yiqi Deng and others},
year = {2026},
eprint = {2611.04870},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04870},
}
Ju Dong, Xinyu Xiao, Yiming Zhou, Tianshuo Yang et al.
Abstract摘要
Humanoid loco-manipulation requires coordinated body motion and hand interaction over time, and a plausible-looking command sequence can still drop a held object or leave an interaction incomplete. We present action-conditioned latent foresight, a latent world-action model (WAM) that evaluates proposed command sequences before execution. An action generator proposes joint motion-hand command sequences from vision, language, and the robot state. An action-conditioned endpoint predictor estimates the visual features each sequence prefix reaches and measures progress toward a …
@misc{dong2026coordinating,
title = {{Coordinating Humanoid Loco-Manipulation through Action-Conditioned Latent Foresight}},
author = {Ju Dong and Xinyu Xiao and Yiming Zhou and Tianshuo Yang and Xuerong Liu and Lei Zhang and Jian Chen and Haocheng Zhao and others},
year = {2026},
eprint = {2611.04839},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04839},
}
Qinglin Yang, Jiepeng Wang, Yuxing Huang, Haibin Huang et al.
Abstract摘要
Recent advances in video modeling have driven substantial progress in world action models (WAMs). Existing approaches exploit pretrained video priors by decoding actions from latent representations or inferring them from generated videos. The underlying link is causal: robot actions produce changes in the scene, and video observations of these changes provide evidence for inferring the actions that produced them. However, existing training formulations do not by themselves ensure that actions and their predicted visual consequences are consistent, potentially limiting the use of pretrained …
@misc{yang2026cycle,
title = {{Cycle WAM: Coupling Action and Video through Cycle Consistency}},
author = {Qinglin Yang and Jiepeng Wang and Yuxing Huang and Haibin Huang and Kai Xu and Chi Zhang and Ruizhen Hu and Xuelong Li},
year = {2026},
eprint = {2611.04711},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04711},
}
Jaehyeong Jo, Sangwon Jang, Sihyun Yu, Seonghyeon Ye et al.
Abstract摘要
World Action Models (WAMs) use the predictive knowledge of pretrained video diffusion models for robot control through future video prediction. However, iterative video denoising is costly, and inaccurate visual predictions can compromise action prediction. Our key idea is to make this predictive knowledge available through visual representations computed from the current observation, enabling action prediction without generating future videos. We introduce DreamFree, a WAM that repurposes a pretrained video diffusion model as a predictive visual encoder through representation distillation. …
@misc{jo2026future,
title = {{Future Representation Distillation for Imagination-Free World Action Models}},
author = {Jaehyeong Jo and Sangwon Jang and Sihyun Yu and Seonghyeon Ye and Yekyung Nah and Hojung Jung and Kimin Lee and Sung Ju Hwang},
year = {2026},
eprint = {2611.04527},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04527},
}
Jimmy Wang, Yifan Chou, Kaifeng Zhang, Yixuan Wang et al.
Abstract摘要
Predicting how the world evolves in response to robot actions is central to policy learning, evaluation, and planning. However, many action-conditioned world models predict future observations in 2D image or video space without explicitly modeling the underlying 3D scene, making spatial consistency across viewpoints and over long-horizon interactions challenging. We introduce 3D World Simulator, an action-conditioned multimodal world modeling framework that jointly predicts appearance, geometry, and tactile observations from a unified 3D representation. Our approach combines a hybrid …
@misc{wang2026hybrid,
title = {{Hybrid Particle-Voxel Dynamics for Multimodal 3D World Simulation}},
author = {Jimmy Wang and Yifan Chou and Kaifeng Zhang and Yixuan Wang and Binghao Huang and Tao Yu and Saleh Nabi and Yunzhu Li},
year = {2026},
eprint = {2611.04492},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.04492},
}
World models are increasingly used to support prediction and planning in embodied agents, but the knowledge is usually established before deployment and then carried into later encounters. Continual operation faces a different problem. A familiar environment may change, and the agent must use previous experience without allowing old expectations to override current physical evidence. This study asks what repeated physical experience must contribute for a spatial world model to become progressively more useful rather than simply more persistent. A physics-grounded dynamic maze separates …
World action models (WAMs) jointly generate actions and imagined futures, but a plausible prediction need not be realized by its co-generated action. We introduce WISE (\emph{World Action Inference Scaling through Evaluation}), a pre-execution selector that scores each sampled rollout on action validity and temporal regularity, inverse-dynamics consistency between the proposed action and the imagined motion, and instruction-conditioned task progress, and executes the highest-scoring candidate while keeping the WAM frozen. With four candidates per decision, WISE raises closed-loop success from …
@misc{ahmed2026seeing,
title = {{Seeing Through the Illusion: Test-Time Scaling for World Action Models}},
author = {Cheikh Ahmed and Mahdi Mostajabdaveh and Milad Yazdani and Zirui Zhou},
year = {2026},
eprint = {2611.03585},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03585},
}
A robot that acts around people needs a world model of the human: a belief over the person's body state and a dynamics model that predicts where that state goes next. This abstract argues that two published components form exactly these two halves, and specifies how to couple them. KNOWN [1] reconstructs the 3D human body while estimating both aleatoric and epistemic uncertainty in real time (11.6 ms per frame); its aleatoric term identifies unreliable observations, its epistemic term identifies inputs from low-density regions of the training data, and refinement driven by the epistemic term …
Peiyuan Tang, Yao Li, Xiaodong Zhang, Wuyang Zhang et al.
Abstract摘要
Contact-rich manipulation requires fine-grained control and complex physical interactions with the environment, posing significant challenges for robot policy learning. Vision-only policies struggle with such tasks, as contact states and interaction forces are difficult to infer from visual observations alone. Existing methods mitigate this issue by incorporating force or tactile sensing. However, they often treat these signals as auxiliary observations without fully exploiting their role in modeling future interaction dynamics. To address this, we propose ForceLDM, a force-aware latent …
@misc{tang2026forceldm,
title = {{ForceLDM: A Force-aware Latent Dynamics Model for Contact-Rich Manipulation}},
author = {Peiyuan Tang and Yao Li and Xiaodong Zhang and Wuyang Zhang and Qin Xia and Yanyong Zhang and Zijiang Yang},
year = {2026},
eprint = {2611.03294},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.03294},
}
We reimplement the dynamics model of the Robotic World Model (RWM) and test the uncertainty estimate that its follow-up, RWM-U, uses to penalize reward in offline policy optimization. Our implementation matches the released model's gradients. On held-out data, training on the model's own rollouts beats one-step training at the preregistered long horizon for all ten episodes, but loses on short one-step windows. RWM-U's ensemble disagreement ranks rollout errors better than forecast step (Spearman rho = 0.90 at 100 steps), yet its spread substantially understates error: one-sigma coverage on …
@misc{singh2026right,
title = {{Right Ranking, Wrong Scale: Reproducing the Robotic World Model's Dynamics Model and Testing the Uncertainty of Its Follow-up}},
author = {Joyjeet Singh},
year = {2026},
eprint = {2611.02699},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2611.02699},
}