Jason / Works Embodied AIZero to One
Works
没主意?快捷入口
End-to-End VLA · Plate Nº 180

Toward Embodied AGI: A Review of Embodied AI and the Road Ahead

12 min read · 4045 字 · ⭐⭐⭐ · 长篇结构化

这是一份面向零基础读者的结构化研究笔记。本文只把公开论文文本和 arXiv 元数据能支持的结论写成事实;本站没有本地训练、仿真或真机复现实验,因此不会把论文中的路线图判断写成本站 E4 结果。

一句话讲什么(TL;DR)

这篇综述提出 Embodied AGI 的五级路线图 L1-L5,并用四个能力维度评估当前差距:omnimodal capabilities、humanoid cognitive abilities、real-time responsiveness、generalization。作者认为当前 Embodied AI 大致处在 L1-L2,距离 L3+ 还缺多模态、类人认知、实时响应和开放泛化能力,并提出一个 L3+ robotic brain 的概念框架。

如果只记一个直觉:它不是某个机器人模型论文,而是像自动驾驶 L1-L5 一样,给具身智能从“会做单一任务”到“类人开放任务”画阶段地图。

所以这一节是想说:这篇论文给 Embodied AGI 提供了评估尺子,而不是提供一个已完成系统。

这是个什么场景

AGI 常被理解为通用人工智能,但很多讨论停在文本或数字世界。Embodied AGI 进一步要求智能体有身体,能在真实世界中感知、行动、互动和学习。机器人不是只要答对问题,还要在物理环境里完成开放任务。

论文提出一个 pragmatic definition:Embodied AGI 是一种 Embodied AI,具备 human-like interaction capabilities,并能以 human-level proficiency 完成 diverse, open-ended real-world tasks。这个定义把 AGI 的“通用”落到真实世界任务和人类互动上。

从 AI 到 Embodied AGI

文本推理
  │
  ▼
多模态理解
  │
  ▼
具身行动
  │
  ▼
开放真实任务 + 类人交互 + 长期学习
四个能力维度

1. Omnimodal capabilities
2. Humanoid cognitive abilities
3. Real-time responsiveness
4. Generalization

这篇论文适合在读了很多单篇机器人模型后回看。它帮我们问:这些模型到底处在什么阶段?只是能抓一个物体,还是能跨任务、跨环境、实时响应、像人一样和社会关系互动?

所以这一节是想说:Embodied AGI 把“会做任务”提升到“能在人类世界中通用行动”。

Toward Embodied AGI — 场景示意:这论文要解决的现实问题
Plate Nº IToward Embodied AGI — 场景示意:这论文要解决的现实问题

之前的人怎么做的,为什么不够好

已有 Embodied AI 进展很多,例如 LLM、VLM、VLA、omnimodal models、机器人学习和强化学习。但论文认为这些能力仍然不足以达到 L3+。现有系统往往只能处理有限任务、有限环境或有限模态。

比如很多 VLA 能在某些机器人任务上表现强,但它们常常不是全模态;很多模型能理解图像和语言,但不具备精确实时动作执行;很多机器人能完成单一或组合任务,但不能可靠处理开放任务和人类社会互动。

论文指出现有架构和主流学习范式,包括 supervised learning 和 reinforcement learning,在 acquiring human-like behaviors 和 robust generalization 方面仍不足。也就是说,继续堆现有范式不一定直接通向 L3-L5。

所以这一节是想说:现有 Embodied AI 有明显进步,但离类人通用具身智能还有结构性差距。

这篇论文的新想法

论文最核心的新想法是五级路线图。受自动驾驶分级启发,它把 Embodied AGI 分为 L1 到 L5:

  • L1:Single-task completion,能稳定完成单一明确任务。
  • L2:Compositional task completion,能组合简单技能完成复合任务。
  • L3:Conditional general-purpose task completion,能在一定条件下跨任务、跨环境、实时适应。
  • L4:Highly general-purpose robots,能对广泛未见任务保持强泛化。
  • L5:All-purpose robots,开放任务中的类人全能目标。

第二个新想法是四维评估:全模态能力、类人认知能力、实时响应能力、泛化能力。路线图不是只看任务成功率,而是看机器人是否能处理更多模态、是否有类人学习和社会理解、是否实时、是否能面对开放环境。

L1-L5 和四维能力的关系

┌──────────┬────────────┬────────────┬────────────┬────────────┐
│ 等级     │ 模态       │ 认知       │ 实时       │ 泛化       │
├──────────┼────────────┼────────────┼────────────┼────────────┤
│ L1       │ 单任务输入 │ 任务内规则 │ 低闭环     │ 任务内     │
│ L2       │ 多技能组合 │ 计划分解   │ 局部闭环   │ 同类任务   │
│ L3+      │ omnimodal  │ 类人记忆   │ 强实时     │ 跨任务类   │
└──────────┴────────────┴────────────┴────────────┴────────────┘

所以这一节是想说:本文把“通用具身智能”拆成等级和能力维度,使讨论可比较。

它分几步做的(方法)

第 1 步:定义 Embodied AGI

论文先给出实用定义:Embodied AGI 要具备类人交互能力,并以人类水平熟练完成多样开放真实任务。这个定义很重要,因为它避免只用“模型参数大”或“benchmark 分数高”来讨论 AGI。

第 2 步:提出 L1-L5 路线图

作者借鉴自动驾驶分级,把具身智能划分为五个等级。L1 类似单功能辅助,L2 能组合任务,L3 开始具备条件通用性,L4 是高度通用,L5 是开放任务全能。

Embodied AGI levels

L1: 单一任务
L2: 组合任务
L3: 条件通用任务
L4: 高度通用机器人
L5: 全用途机器人

这种分级的好处是能防止“会做一个新任务就叫 AGI”的夸大表述。每一级都有任务范围、泛化和能力要求。

第 3 步:用四个维度评估当前状态

四个维度分别是 omnimodal capabilities、humanoid cognitive abilities、real-time responsiveness、generalization。论文认为当前 Embodied AI 大致处在 L1-L2,原因是全模态处理、类人认知、实时响应和开放泛化都还不够。

L3+ 需要的四件事

全模态输入输出
类人认知和社会理解
快速准确的实时行动
跨环境跨任务泛化

第 4 步:讨论 L3-L5 的关键构件

论文进一步讨论 L3+ 需要什么:更全面的 omnimodal model structure,更强的 streaming input 和 duplex interaction,更接近人类的 social connection understanding、自我意识、procedural memory、memory reconsolidation,以及 physical-oriented training。

这些构件不一定是最终答案,而是作者提出的概念框架,用来描述未来 robotic brain 应该满足什么目标。

第 5 步:提出 L3+ robotic brain 概念框架

论文最后提出一个 conceptual framework,包括 omnimodal model structure 和 training paradigm。它强调 streaming、cross-modal alignment、omnimodal understanding、physical-oriented training 等方向。

所以这一节是想说:论文方法是“定义目标 -> 分级 -> 评估差距 -> 提出 L3+ 构件”。

Toward Embodied AGI — 方法示意:核心 pipeline
Plate Nº IIToward Embodied AGI — 方法示意:核心 pipeline

关键数字

本文最重要的数字是五级路线图和四个能力维度。五级是 L1-L5,四维是 omnimodal capabilities、humanoid cognitive abilities、real-time responsiveness、generalization。

论文还给出一个关键判断:current state of Embodied AI development between Levels 1 and 2 (L1-L2)。这不是严格实验分数,而是作者基于当前文献和能力维度的综述性判断。

此外,论文提到 advancing to L3 requires handling substantially different task categories and exhibiting robust real-time responsiveness。也就是说,L3 的门槛不是只在同一任务族内泛化,而是跨任务类别和实时响应能力都要提升。

所以这一节是想说:L1-L5 和四维能力是本文的主记忆点。

实验结果说明了什么

这篇论文没有训练一个新模型,也没有做统一机器人实验。它的“结果”是路线图和能力差距分析。作者认为,当前很多系统能达到 L1,部分接近 L2,但尚未达到 L3+。

这个判断对读其他论文很有帮助。比如一篇 VLA 在抓取任务上泛化不错,可能只是 L1 到 L2;一篇导航模型能跨任务和跨本体,但如果没有类人认知或开放任务能力,也不能直接叫 L4 或 L5。

论文还提醒我们,实时响应不是可选项。一个机器人如果每次思考都很慢,即使推理结果正确,也难以在人类环境中交互。未来 L3+ 需要真正支持 multiplexed、omnimodal、streaming 的处理方式。

所以这一节是想说:论文的结果是给现有 Embodied AI 降温,并指出 L3+ 的具体缺口。

术语表

  • Embodied AGI:具备身体并能在真实世界中完成开放任务的通用智能。
  • L1-L5:作者提出的五级 Embodied AGI 路线图。
  • Omnimodal:不只视觉语言,还包含音频、触觉、力、热、运动等更多模态。
  • Humanoid cognitive abilities:类人认知能力,如社会理解、自我意识、程序记忆等。
  • Real-time responsiveness:实时响应能力,能快速准确行动和双向交互。
  • Generalization:跨环境、跨任务、跨情境泛化。
  • Robotic brain:作者提出的 L3+ 机器人认知架构概念。

所以这一节是想说:本文术语是为了衡量“离类人通用机器人还有多远”。

局限和边界

第一,L1-L5 是作者提出的路线图,不是行业统一标准。它有启发性,但后续可能被修正。

第二,当前状态在 L1-L2 的判断是综述性判断,不是通过统一 benchmark 严格测量得出。

第三,Embodied AGI 涉及硬件、认知、社会、实时系统和学习理论。论文提供框架,但没有解决所有实现问题。

第四,类人认知和自我意识等概念很复杂,工程定义和哲学定义可能不同。读这篇时要把它当研究路线图,而不是已验证理论。

所以这一节是想说:这篇论文适合定方向,不适合当最终评测标准。

和其他论文的关系

和 mobile-service-robot-foundation-survey 相比,本文更宏观。移动服务机器人综述关注家庭、医院和服务场景的具体约束,本文关注 Embodied AGI 的等级和长期路线。

和 RoboNeuron 相比,本文是目标地图,RoboNeuron 是工程中间层。L3+ 需要机器人系统能稳定接入工具、ROS、VLA 和低层执行,这正是 RoboNeuron 试图解决的一类问题。

和 NavFoM 相比,本文提出“通用具身智能应具备什么”,NavFoM 是 navigation 方向尝试跨任务、跨本体泛化的具体模型。

和 VLA manipulation survey 相比,本文更关注 AGI 层级和类人能力,而不只是现有 VLA 结构和训练分类。

所以这一节是想说:Batch 5 用这篇论文给后续服务机器人、导航和系统中间层提供上位目标。

和本导读的关系

本站的很多论文可以映射到 L1-L5:单任务控制、复合任务、跨本体泛化、服务机器人、导航基础模型、系统中间层。读这篇后,读者可以开始问每篇论文到底推进了哪个层级。

它也帮助本站避免夸大。一个模型“更强”不等于 Embodied AGI。只有当它在多模态、类人认知、实时响应和开放泛化上都有进展,才更接近 L3+。

所以这一节是想说:它给本站的论文库加了一把分级尺。

思考题

  1. 为什么作者认为当前 Embodied AI 大致在 L1-L2?
  2. L3 和 L2 的关键差别是什么?
  3. 为什么 omnimodal 不等于简单多加几个传感器?
  4. 实时响应为什么是 Embodied AGI 的核心能力,而不是工程优化项?

FAQ

Q:L5 是不是已经有明确实现路线? A:没有。L5 是终极目标,论文主要提供路线图和概念框架。

Q:这篇论文是不是说 AGI 必须是机器人? A:它关注 Embodied AGI,并认为 AGI 常被视为 inherently embodied;但这是论文立场,不是所有 AGI 讨论的唯一观点。

Q:为什么用自动驾驶分级类比? A:因为自动驾驶 L1-L5 提供了可沟通的阶段框架,能帮助具身智能避免泛泛谈“通用”。

Q:本站有没有验证 L1-L5? A:没有。这里只记录论文路线图,不把它写成本站评测体系。

进一步读什么

  • mobile-service-robot-foundation-survey:看 Embodied AGI 在服务机器人场景中的约束。
  • embodied-navigation-foundation-model:看跨任务/跨本体导航模型如何接近 L3 的一部分。
  • roboneuron:看工程中间层如何支撑 agentic embodied systems。
  • vla-manipulation-survey:看 VLA 在具身操作中的现有结构。

精读补充:L2 到 L3 为什么是关键门槛

这篇论文最值得反复看的不是 L5 这个远大目标,而是 L2 到 L3 的分界。L2 的机器人已经能做组合任务,比如把“拿杯子再放到桌上”拆成抓取、移动、放置。听起来已经很强,但它仍然可能只是在有限任务集合内组合技能。L3 则要求 conditional general-purpose task completion,也就是在一定条件下跨任务类别、跨环境、跨人类指令工作,并且具备 robust real-time responsiveness。

这一步之所以难,是因为组合已知技能和面对新任务不是一回事。组合技能像搭积木,积木形状都知道;新任务则可能需要判断新物体、新社交情境、新失败模式和新动作约束。很多现有机器人在演示中像 L2,但一旦任务类别变化,就会退回到需要人工重新定义动作、重新收集数据或重新调参。

四个能力维度也可以用 L2 到 L3 来理解。Omnimodal capabilities 要求机器人不仅看图听话,还要融合声音、触觉、力、热、位置、人体状态等更丰富信号。Humanoid cognitive abilities 要求它不只是执行命令,还要理解社会关系、程序性记忆和长期经验。Real-time responsiveness 要求它不能“想半分钟再动一下”。Generalization 要求它不只换光照和背景,还能跨任务类别泛化。

论文对当前状态的判断比较克制:大致在 L1-L2。这个判断对学习很有帮助,因为它能防止我们被单篇论文的强标题带偏。一个模型如果只在某类抓取或导航 benchmark 上好,不应该直接被称为 Embodied AGI。它可能只是 L1 或 L2 上的优秀方法。

L3+ robotic brain 的概念框架也不是具体产品设计图,而是需求列表。它要求 omnimodal streaming input、跨模态对齐、物理导向训练、类人认知机制和实时交互。换句话说,未来机器人“大脑”不只是更大的 LLM,而是一个持续感知、持续行动、持续学习、持续和世界校准的系统。

对本站来说,这篇论文可以当成“防夸大工具”。之后读任何新论文,都可以问:它推进的是哪一级?它只提升单任务,还是提升组合任务?它有没有跨任务类别?有没有实时响应?有没有类人认知或长期记忆?这些问题能帮助我们把 hype 拆成可审查的能力维度。

另一个容易忽略的点是,L3 不是“把 L2 做大一点”。L2 可以通过更多技能库、更好的任务分解、更强的提示词工程继续提升,但这些改进仍然可能依赖封闭任务空间。L3 需要机器人在条件约束内识别新任务、判断可行性、主动澄清、失败恢复,并把感知变化及时反馈到行动中。这里的“条件通用”不是无限能力,而是明确边界下的可迁移能力。

这也解释了为什么论文把 real-time responsiveness 单独列出来。桌面聊天系统可以慢几秒,但机器人慢几秒可能已经错过抓取时机,或者在人靠近时没有及时停下。实时性不是把模型部署得更快这么简单,它要求感知、规划、控制和人机交互形成闭环。模型如果只会离线推理路线,仍然很难进入 L3。

因此,L1-L5 分级适合作为阅读索引,而不是打分榜。我们可以用它标注论文贡献:某篇论文可能提升多模态感知,某篇提升跨本体数据,某篇提升中间层 orchestration,某篇提升导航泛化。它们未必单独把系统推到 L3,但可以分别补 L3 所需的拼图。

所以这一节是想说:L3 才是从技能组合走向条件通用的分水岭,当前大多数工作还没真正跨过去。

后续核验清单

如果之后要把本文从 UNVERIFIED 提升到人工核验状态,应逐项核对:L1-L5 定义是否按原文表述;四个能力维度是否为 omnimodal capabilities、humanoid cognitive abilities、real-time responsiveness、generalization;当前状态 L1-L2 的判断是否保留综述边界;L3+ robotic brain 是否写成 conceptual framework 而非已实现系统。

原文信息

@article{wang2025embodiedagiroadahead,
  title = {Toward Embodied AGI: A Review of Embodied AI and the Road Ahead},
  author = {Wang, Yequan and Sun, Aixin},
  journal = {arXiv preprint arXiv:2505.14235},
  year = {2025}
}

引用本笔记 / Cite this note
BibTeX
@online{eai_embodied_agi_road_ahead_2026,
  title       = {(readable note) Toward Embodied AGI: A Review of Embodied AI and the Road Ahead},
  author      = {Xun, Jason},
  year        = {2026},
  note        = {Note on a 2025 paper},
  howpublished = {\url{https://estelledc.github.io/embodied-ai-reading-station/papers/embodied-agi-road-ahead/}},
  organization = {Embodied AI: Zero to One}
}

All 202 papers (full index)
  1. 1. LLaVA: Visual Instruction Tuning
  2. 2. 3DShape2VecSet: 3D Shape Representation for Diffusion Models
  3. 3. SayCan: Do As I Can, Not As I Say
  4. 4. OpenVLA: An Open-Source Vision-Language-Action Model
  5. 5. VLAS: VLA Model With Speech Instructions
  6. 6. MLA: Multisensory Language-Action Model
  7. 7. Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control
  8. 8. CartoRadar: RF-Based 3D SLAM Rivaling Vision Approaches
  9. 9. mmCLIP: Boosting mmWave-based Zero-shot HAR via Signal-Text Alignment
  10. 10. mmNorm: Non-Line-of-Sight 3D Object Reconstruction via mmWave Surface Normal Estimation
  11. 11. Proactive Hearing Assistants that Isolate Egocentric Conversations
  12. 12. NeuralAids: Wireless Hearables With Programmable Speech AI Accelerators
  13. 13. Creating speech zones with self-distributing acoustic swarms
  14. 14. Conv-TasNet: Surpassing Ideal Time-Frequency Magnitude Masking for Speech Separation
  15. 15. SoundStream: An End-to-End Neural Audio Codec
  16. 16. AudioLM
  17. 17. Conformer
  18. 18. Dual-path RNN
  19. 19. EnCodec
  20. 20. Meta-StyleSpeech
  21. 21. MusicLM
  22. 22. Robust Speech Recognition via Large-Scale Weak Supervision
  23. 23. SeamlessM4T
  24. 24. Stable Audio
  25. 25. Universal Source Separation with Weakly Labelled Data
  26. 26. Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning
  27. 27. RLBench: The Robot Learning Benchmark & Learning Environment
  28. 28. robosuite: A Modular Simulation Framework and Benchmark for Robot Learning
  29. 30. CALVIN
  30. 31. LIBERO
  31. 32. RH20T
  32. 33. What Matters in Learning from Offline Human Demonstrations for Robot Manipulation
  33. 34. DROID
  34. 35. Open X-Embodiment
  35. 36. RoboCasa
  36. 37. SimplerEnv
  37. 38. Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
  38. 39. 3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations
  39. 40. Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation
  40. 41. EquiBot: SIM(3)-Equivariant Diffusion Policy
  41. 42. DiT-Policy
  42. 43. Diffusion Policy Policy Optimization (DPPO)
  43. 44. Affordance-based Robot Manipulation with Flow Matching
  44. 45. FlowPolicy: 3D Flow-based Policy via Consistency Flow Matching
  45. 46. FAST: Efficient Action Tokenization for VLA
  46. 47. π₀: A Vision-Language-Action Flow Model for General Robot Control
  47. 48. pi_0.5: VLA with Open-World Generalization
  48. 49. A Reduction of Imitation Learning and Structured Prediction to No-Regret Online Learning
  49. 50. Generative Adversarial Imitation Learning
  50. 51. Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware (ACT/ALOHA)
  51. 52. AnyTeleop
  52. 53. Behavior Transformers: Cloning k Modes with One Stone
  53. 54. Implicit Behavioral Cloning
  54. 55. RoboCat
  55. 56. ALOHA 2
  56. 58. HumanPlus
  57. 59. Generalizable Humanoid Manipulation with 3D Diffusion Policies (iDP3)
  58. 60. Mobile ALOHA
  59. 61. SmolVLA
  60. 62. Universal Manipulation Interface
  61. 63. Behavior Generation with Latent Actions (VQ-BeT)
  62. 64. ImageBind: One Embedding Space To Bind Them All
  63. 65. Connecting Touch and Vision via Cross-Modal Prediction
  64. 66. AnyMAL: An Efficient and Scalable Any-Modality Augmented Language Model
  65. 67. AudioPaLM
  66. 68. FROMAGe: Grounding LLMs to Images
  67. 69. OneLLM
  68. 70. X-VLM: Multi-Grained Vision Language Pre-Training
  69. 71. Tactile Beyond Pixels (Sparsh-X)
  70. 72. Sparsh: Self-supervised Touch Representations
  71. 73. Tactile-VLA
  72. 74. TLA: Tactile-Language-Action
  73. 75. Code as Policies: Language Model Programs for Embodied Control
  74. 76. Inner Monologue: Embodied Reasoning through Planning with Language Models
  75. 77. LLM+P: Empowering LLMs with Optimal Planning
  76. 78. PaLM-E: An Embodied Multimodal Language Model
  77. 79. ProgPrompt
  78. 80. ChatGPT for Robotics
  79. 81. GenSim
  80. 82. RoboFlamingo
  81. 83. Tree-Planner
  82. 84. VoxPoser
  83. 85. See Through Smoke: Robust Indoor Mapping with Low-cost mmWave Radar
  84. 86. Can WiFi Estimate Person Pose?
  85. 87. 3DRIMR: 3D Reconstruction and Imaging via mmWave Radar based on Deep Learning
  86. 88. milliEgo: Single-chip mmWave Radar Aided Egomotion Estimation via Deep Sensor Fusion
  87. 89. High Resolution Point Clouds from mmWave Radar
  88. 90. RadarSLAM: Radar based Large-Scale SLAM in All Weathers
  89. 91. Through-Wall Pose Imaging in Real-Time with a Many-to-Many Encoder/Decoder Paradigm
  90. 92. RFMask: A Simple Baseline for Human Silhouette Segmentation with Radio Signals
  91. 93. RFPose-OT: RF-Based 3D Human Pose Estimation via Optimal Transport Theory
  92. 94. Argus: Multi-View Egocentric Human Mesh Reconstruction Based on Stripped-Down Wearable mmWave Add-on
  93. 95. Diffusion Model is a Good Pose Estimator from 3D RF-Vision
  94. 96. Enabling Visual Recognition at Radio Frequency (PanoRadar)
  95. 97. Wave-Former: Through-Occlusion 3D Reconstruction via Wireless Shape Completion
  96. 98. Habitat: A Platform for Embodied AI Research
  97. 99. Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning
  98. 101. Habitat 2.0
  99. 102. ManiSkill
  100. 103. ProcTHOR
  101. 104. SAPIEN: A SimulAted Part-based Interactive ENvironment
  102. 105. BEHAVIOR-1K
  103. 106. BridgeData V2
  104. 106. Habitat 3.0
  105. 107. Isaac Lab
  106. 108. DexMV
  107. 108. MuJoCo Playground
  108. 109. DexCap
  109. 109. RT-1: Robotics Transformer for Real-World Control at Scale
  110. 110. 3D Diffusion Policy (DP3)
  111. 111. Octo: An Open-Source Generalist Robot Policy
  112. 112. RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
  113. 113. RT-Trajectory: Robotic Task Generalization via Hindsight Trajectory Sketches
  114. 114. 3D-VLA
  115. 116. GR-2: Generative Video-Language-Action Model
  116. 117. DexVLA
  117. 117. OpenHelix
  118. 118. Cosmos World Foundation Model
  119. 118. OpenVLA-OFT
  120. 119. RDT-1B: Diffusion Foundation Model for Bimanual Manipulation
  121. 120. RoboMamba
  122. 121. SpatialVLA
  123. 122. TinyVLA
  124. 123. TraceVLA: Visual Trace Prompting
  125. 124. Learning Transferable Visual Models From Natural Language Supervision
  126. 125. Flamingo: a Visual Language Model for Few-Shot Learning
  127. 126. BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
  128. 127. BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
  129. 128. DeepSeek-VL: Towards Real-World Vision-Language Understanding
  130. 129. EVA-CLIP: Improved Training Techniques for CLIP at Scale
  131. 130. FILIP: Fine-grained Interactive Language-Image Pre-Training
  132. 131. Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
  133. 132. InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks
  134. 133. Improved Baselines with Visual Instruction Tuning
  135. 134. OBELICS
  136. 135. Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond
  137. 136. Sigmoid Loss for Language Image Pre-Training
  138. 137. What matters when building vision-language models?
  139. 138. Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
  140. 139. The Llama 3 Herd of Models
  141. 140. LLaVA-NeXT-Interleave
  142. 141. LLaVA-OneVision: Easy Visual Task Transfer
  143. 142. Long-CLIP: Unlocking the Long-Text Capability of CLIP
  144. 143. Pixtral 12B
  145. 144. Dream to Control: Learning Behaviors by Latent Imagination
  146. 145. World Models
  147. 146. DayDreamer
  148. 147. Mastering Atari with Discrete World Models
  149. 148. Dreamer V3: Mastering Diverse Domains through World Models
  150. 149. Transformers are Sample-Efficient World Models
  151. 150. TWM: Transformer-based World Models
  152. 151. 1X World Model Challenge
  153. 153. GAIA-1
  154. 154. Genie: Generative Interactive Environments
  155. 155. Navigation World Models
  156. 156. UniSim
  157. 157. LeRobot: An Open-Source Library for End-to-End Robot Learning
  158. 158. CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulation
  159. 159. Universal Actions for Enhanced Embodied Foundation Models
  160. 160. LoHoVLA: A Unified Vision-Language-Action Model for Long-Horizon Embodied Tasks
  161. 161. AutoRT: Embodied Foundation Models for Large Scale Orchestration of Robotic Agents
  162. 162. EO-1: Interleaved Vision-Text-Action Pretraining for General Robot Control
  163. 163. Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments
  164. 164. RealMirror: A Comprehensive, Open-Source Vision-Language-Action Platform for Embodied AI
  165. 165. LLaDA-VLA: Vision Language Diffusion Action Models
  166. 166. Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies
  167. 167. Vlaser: Vision-Language-Action Model with Synergistic Embodied Reasoning
  168. 168. X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model
  169. 169. Embodiment Transfer Learning for Vision-Language-Action Models
  170. 170. HiMoE-VLA: Hierarchical Mixture-of-Experts for Generalist Vision-Language-Action Policies
  171. 171. Green-VLA: Staged Vision-Language-Action Model for Generalist Robots
  172. 172. AC^2-VLA: Action-Context-Aware Adaptive Computation in Vision-Language-Action Models for Efficient Robotic Manipulation
  173. 173. MoS-VLA: A Vision-Language-Action Model with One-Shot Skill Adaptation
  174. 174. Asynchronous Fast-Slow Vision-Language-Action Policies for Whole-Body Robotic Manipulation
  175. 175. VLA-Forget: Vision-Language-Action Unlearning for Embodied Foundation Models
  176. 176. Membership Inference Attacks on Vision-Language-Action Models
  177. 177. A Survey on Efficient Vision-Language-Action Models
  178. 178. Survey of Vision-Language-Action Models for Embodied Manipulation
  179. 179. Embodied AI with Foundation Models for Mobile Service Robots: A Systematic Review
  180. 180. Toward Embodied AGI: A Review of Embodied AI and the Road Ahead
  181. 181. RoboNeuron: A Middle-Layer Infrastructure for Agent-Driven Orchestration in Embodied AI
  182. 182. Embodied Navigation Foundation Model
  183. 183. MiMo-Embodied: X-Embodied Foundation Model Technical Report
  184. 184. Open-H-Embodiment: A Large-Scale Dataset for Enabling Foundation Models in Medical Robotics
  185. 185. AlanaVLM: A Multimodal Embodied AI Foundation Model for Egocentric Video Understanding
  186. 186. 3D Generation for Embodied AI and Robotic Simulation: A Survey
  187. 187. DISCO: Language-Guided Manipulation with Diffusion Policies and Constrained Inpainting
  188. 188. Time-Unified Diffusion Policy with Action Discrimination for Robotic Manipulation
  189. 189. Learning Diffusion Policy from Primitive Skills for Robot Manipulation
  190. 190. Trace-Focused Diffusion Policy for Multi-Modal Action Disambiguation in Long-Horizon Robotic Manipulation
  191. 191. Gaze2Act: Gaze-Conditioned Vision-Language-Action Policies for Interactive Robot Manipulation
  192. 192. LACY: A Vision-Language Model-based Language-Action Cycle for Self-Improving Robotic Manipulation
  193. 193. villa-X: Enhancing Latent Action Modeling in Vision-Language-Action Models
  194. 194. InstructVLA: Vision-Language-Action Instruction Tuning from Understanding to Manipulation
  195. 195. Discrete Policy: Learning Disentangled Action Space for Multi-Task Robotic Manipulation
  196. 196. Towards Generalizable Vision-Language Robotic Manipulation: A Benchmark and LLM-guided 3D Policy
  197. 197. A Survey of Language-Conditioned Robot Manipulation
  198. 198. SafeEmbodAI: a Safety Framework for Mobile Robots in Embodied AI Systems
  199. 199. The Essential Role of Causality in Foundation World Models for Embodied AI
  200. 200. A call for embodied AI
  201. 201. Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis
  202. 202. What Breaks Embodied AI Security: LLM Vulnerabilities, CPS Flaws, or Something Else?