Jason / Works Embodied AIZero to One
Works
没主意?快捷入口
Diffusion Policy · Plate Nº 189

Learning Diffusion Policy from Primitive Skills for Robot Manipulation

11 min read · 4001 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。本文只记录公开论文文本、arXiv 元数据和公开摘要能支持的结论;本站没有复现 CALVIN、LIBERO 或真实机器人实验,因此不会把论文报告的成功率写成本站 E4 结果。

一句话讲什么(TL;DR)

Learning Diffusion Policy from Primitive Skills for Robot Manipulation 提出 Skill-conditioned Diffusion Policy,简称 SDP。它认为很多 diffusion policy 直接从高层语言指令生成短期动作,会出现粒度错位:语言说“把柠檬放进锅里”,但机器人每一刻真正要做的是 move up、move down、open gripper、close gripper 这类短程基础动作。

SDP 抽象出八个 reusable primitive skills,并用 VLM 从视觉观察和语言指令中提取离散表示,再通过 lightweight router 为每个状态选择一个 primitive skill。之后,skill-conditioned diffusion policy 根据选中的 skill 生成 skill-aligned actions。论文报告 SDP 在 CALVIN、LIBERO 和真实机器人任务上超过 SOTA baseline,例如 CALVIN ABC→D 五连任务成功率 76.9%,LIBERO 平均成功率 96.9%。

如果只记一个直觉:SDP 给机器人加了一层“动作短语”。人说的是完整任务,低层 policy 输出的是关节动作,中间用“打开夹爪、向上移动、旋转”这类 primitive skills 对齐。

所以这一节是想说:SDP 用可解释 primitive skills 缓解语言指令和低层动作之间的粒度错位。

这是个什么场景

机器人操作任务常常是长程组合任务。比如 “Pick up the lemon and put it into the yellow pan” 看起来是一句话,但实际包含接近、下移、张开夹爪、闭合夹爪、抬起、移动、放下等多个短动作。

普通 diffusion policy 可以生成动作序列,但如果它只接收全局高层指令,就要自己隐式学会这些短动作阶段。模型可能知道最终目标,却在某一刻不知道该闭夹爪还是移动,导致 action generation misalignment。

SDP 的场景就是把复杂任务拆成更短、更可解释、更可复用的 primitive skills。它不是让人手写完整任务程序,而是让模型动态选择当前状态该用哪个技能。

高层任务和低层动作之间的粒度差

High-level instruction:
  "Pick up the lemon and put it into the yellow pan"
        │
        ▼
Primitive skill sequence:
  move down -> open gripper -> close gripper -> move up -> translate -> move down
        │
        ▼
Low-level actions:
  continuous robot control signals

所以这一节是想说:长程操作需要中间层,否则语言和动作粒度差太大。

Learning Diffusion Policy from Primitive Skills for Robot Manipulation — 场景示意:这论文要解决的现实问题
Plate Nº ILearning Diffusion Policy from Primitive Skills for Robot Manipulation — 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

原始 diffusion policy 从视觉观察和动作示范中学习动作分布,对单任务或短程控制很强。但在多任务语言条件操作中,模型需要把高层语言直接映射到低层动作,这很难。

一些方法加入语言 encoder,把 instruction embedding 和 noisy action sequence 一起输入 diffusion policy。这样能处理更多任务,但语言 embedding 往往太粗,不一定告诉模型当前时刻应该做哪个短动作。

还有一些 VLA 或 multitask policy 学隐式技能表示。隐式表示可能有效,但很难解释和调试。机器人失败时,我们不知道它是“没理解任务”,还是“选错了当前技能”,还是“低层动作生成错了”。

SDP 的判断是:primitive skills 应该显式出现。它们既能被人理解,又能作为 action generation 的条件,让 diffusion policy 更稳定地产生 skill-consistent behavior。

所以这一节是想说:旧方法把技能藏在模型里,SDP 想把技能显式拿出来。

这篇论文的新想法

第一,新想法是八个 reusable primitive skills。论文列出 roll、yaw、open the gripper、move up、translate、close the gripper、move down、rotate。这些是跨任务共享的短程 manipulations。

第二,新想法是 lightweight router。模型从视觉观察和高层语言指令中提取 representation,然后 router 选择当前状态最合适的 primitive skill。

第三,新想法是 skill-conditioned diffusion policy。选中的 skill 不只是标签,而是参与 action generation。论文设计 skill-dependent FFN layer,用 skill embedding 动态影响 diffusion policy。

第四,新想法是可解释性。每个状态分配一个 primitive skill,研究者可以检查机器人当前“以为自己在做什么”。这比完全端到端的隐式控制更容易分析。

SDP 框架

Visual observations + language instruction
          │
          ▼
VLM / representation encoder
          │
          ▼
Lightweight router selects primitive skill
          │
          ▼
Skill-conditioned diffusion policy
          │
          ▼
Skill-aligned continuous actions

所以这一节是想说:SDP 把 skill selection 和 diffusion action generation 组合成一个层次化但可训练的系统。

它分几步做的(方法)

第 1 步:定义 primitive skill set

输入是多种 robot manipulation tasks。作者观察这些任务可以拆成基础短动作,于是定义八个 primitive skills:roll、yaw、open the gripper、close the gripper、move up、move down、translate、rotate。

处理过程是把高层任务的连续控制过程看成 primitive skill 序列。输出是一个共享 skill set。这个 set 不绑定某个具体任务,而是跨任务复用。

这一步的价值是提供中间语言。人能理解这些 skill,模型也可以用它们作为条件,降低直接从 task instruction 到 action 的难度。

第 2 步:用 compositional prompt ensemble 构造 skill embeddings

论文使用统一模板,例如 “the robot arm is going to {skill}”。每个 primitive skill 都放进这个模板,经过 CLIP text encoder 和 MLP,得到 skill prompt embeddings。

输入是八个 skill 文本。处理是 text encoding 和 MLP projection。输出是八个 skill embeddings,供 router 和 diffusion policy 使用。

这一步让 skill 不只是字符串,而是进入模型计算图的向量表示。

第 3 步:从视觉和语言中提取状态表示

输入包括 static camera、wrist camera 等视觉观察,以及高层语言指令。模型使用 vision-language model 提取 discrete representations 或 multimodal representations。

处理过程是编码视觉和语言上下文,得到当前状态的综合表示。输出是 router 可以判断的 state representation。

这一步解决“当前该做什么”的信息来源。仅看语言不够,因为同一句任务在不同阶段要做不同 skill;仅看图像也不够,因为不知道目标。

第 4 步:router 为每个状态选择 primitive skill

Lightweight router network 输入 state representation,输出八个 primitive skills 的 importance scores。最终通过 top-1 selection 选择当前状态的 desired primitive skill。

处理过程是动态 skill assignment。输出是一个具体 skill,比如 close gripper 或 move up。

这一步是 SDP 的决策中间层。它把长程任务拆成一连串状态相关的 skill choices。

第 5 步:skill-dependent FFN 影响 diffusion policy

选中的 skill embedding 会进入 diffusion policy。论文引入 skill-dependent feed-forward network layer,类似 LoRA-like 动态参数化,让 diffusion policy 根据 skill 调整动作生成。

输入是 noisy actions、observation features、skill embedding。处理是 diffusion denoising,同时用 skill-conditioned layer 调整中间表示。输出是 continuous actions。

这一步让 skill 真正影响低层动作,而不只是附在输入里的标签。比如 open gripper 和 move down 对同一视觉状态应该产生完全不同动作。

第 6 步:在 CALVIN、LIBERO 和真实机器人上评估

CALVIN 测长程语言条件操作,尤其 ABC→D zero-shot generalization。LIBERO 包含 Spatial、Object、Goal、Long 四类 task suites。真实机器人实验则检验 sim-to-real 和视觉干扰下的表现。

论文报告 SDP 在 CALVIN ABC→D 五连任务成功率 76.9%,超过 MoDE 14.5%、UniVLA 20.4%;LIBERO 平均成功率 96.9%,超过 MDT 13.4%、UniVLA 4.4%,且是唯一在 LIBERO-Long 超过 90% 的方法。

这些结果说明 primitive skill 作为中间层能提升长程组合和多任务泛化,但所有数字仍是论文报告。

所以这一节是想说:SDP 的方法是“选当前 primitive skill,再让 diffusion policy 按 skill 生成动作”。

Learning Diffusion Policy from Primitive Skills for Robot Manipulation — 方法示意:核心 pipeline
Plate Nº IILearning Diffusion Policy from Primitive Skills for Robot Manipulation — 方法示意:核心 pipeline

关键数字

数字 原文语境 这说明什么
8 reusable primitive skills 技能中间层规模
4 SDP 生成动作使用的 denoising steps 少于一些 diffusion baseline 的 10 steps
76.9% CALVIN ABC→D 五连任务成功率 论文报告 zero-shot generalization 强
14.5% 相比 MoDE 的提升 CALVIN ABC→D 设置
20.4% 相比 UniVLA 的提升 CALVIN ABC→D 设置
96.9% LIBERO 四套件平均成功率 论文报告多任务表现
13.4% 相比 MDT 的 LIBERO 平均提升 baseline 对比
4.4% 相比 UniVLA 的 LIBERO 平均提升 baseline 对比

这些数字全部是论文报告,不是本站复现实验。后续人工核验应回到 Table 1、Table 2 和真实机器人图表逐项确认。

所以这一节是想说:SDP 的证据集中在 CALVIN/LIBERO 长程和多任务成功率。

实验结果说明了什么

实验说明 primitive skills 对长程任务有帮助。CALVIN 要连续完成多个 instruction,如果每一步都从高层语言直接生成动作,模型容易漂移;SDP 用 skill sequence 给动作生成提供更细粒度约束。

LIBERO 结果说明,skill-conditioned policy 不只适合一个 benchmark。Spatial、Object、Goal、Long 四类任务对空间关系、对象变化、目标变化和长程组合都有要求,SDP 的平均成功率高说明中间 skill 有泛化价值。

Ablation 结果也重要。论文报告 skill-dependent FFN、explicit skill abstraction、compositional prompt ensemble 都有贡献。也就是说,不只是“随便加个 skill label”就有效,skill 如何进入 diffusion policy 很关键。

真实机器人实验说明,primitive skill abstraction 对部署有实际意义。可解释技能使得人可以更容易观察机器人当前阶段是否合理。

更重要的是,SDP 把失败原因变得更可定位。端到端 policy 失败时,我们只知道动作错了,却很难知道模型内部阶段是否正确。引入 primitive skill 后,可以先看 router 是否选错技能,再看 diffusion policy 是否在正确技能下生成了错误轨迹。前者是阶段理解问题,后者是低层控制问题。

这种可分解性对长程任务尤其有价值。比如同样是“把物体放进容器”,失败可能发生在接近、抓取、抬起、移动、释放任一阶段。Primitive skill 序列能作为一种轻量任务进度条,让研究者更容易发现任务在哪个阶段偏离,也更容易针对性收集数据或调模型。

所以这一节是想说:实验支持“技能中间层”能缓解语言到动作的粒度错位。

术语表

  • Primitive skill:短程、细粒度、可复用的基础操作,如 move up、close gripper。
  • Skill-conditioned diffusion policy:以 skill 作为条件生成动作的 diffusion policy。
  • Router network:根据当前视觉和语言状态选择 skill 的轻量网络。
  • Skill-dependent FFN:根据 skill embedding 动态影响 diffusion policy 的前馈层。
  • CALVIN:长程语言条件机器人操作 benchmark。
  • LIBERO:测试知识迁移和多任务机器人操作的 benchmark。
  • LoRA-like:类似低秩适配的参数调制思想。

所以这一节是想说:SDP 的关键词是 primitive skill、router 和 skill-conditioned denoising。

局限和边界

第一,八个 primitive skills 是否足够覆盖所有真实任务仍不确定。复杂接触、双臂协作、工具使用可能需要更丰富 skill set。

第二,router 选错 skill 会直接影响动作生成。如果当前应该 close gripper,却选成 translate,policy 会产生错动作。

第三,primitive skill 的可解释性不等于安全性。人能看懂 skill,不代表动作一定安全。

第四,CALVIN/LIBERO 是重要 benchmark,但真实环境中物体、光照、传感器和动力学差异更大。

第五,SDP 仍依赖示范数据和 diffusion policy 能力。没有足够数据的技能,不能靠 skill 名字凭空学会。

第六,skill assignment 也可能受视觉遮挡影响。如果关键物体被挡住,router 可能无法判断当前阶段,低层 policy 即使能力足够也会接到错误条件。

因此真实系统通常还需要失败检测、重新观察和人工接管,而不是只依赖一次 skill 选择。

所以这一节是想说:SDP 给了更好的中间层,但 skill set、router 和真实部署仍要验证。

和其他论文的关系

disco-diffusion-policy 相比,SDP 用 primitive skills 作为中间层,DISCO 用 VLM-generated keyframes 作为中间层。

time-unified-diffusion-policy 相比,SDP 关注 task/action granularity,TUDP 关注 denoising efficiency。

trace-focused-diffusion-policy 相比,SDP 用当前 skill disambiguate action,TF-DP 用历史 trace disambiguate long-horizon stage。

和原始 diffusion-policy 相比,SDP 把高层 instruction 到动作的映射拆成更可解释的 skill-conditioned generation。

所以这一节是想说:SDP 是 Batch 7 里“技能层次化 diffusion policy”的代表。

和本导读的关系

本站学习 diffusion policy 时,容易把 policy 看成一个黑盒:输入图像和语言,输出动作。SDP 提醒我们,中间表示很重要。可解释 primitive skills 能降低黑盒程度,也能帮助 debugging。

它适合和 imitation learning、VLA、skill learning、hierarchical policy 一起读。读者可以思考:未来通用机器人是不是需要一套可学习、可组合、可解释的 skill vocabulary?

所以这一节是想说:SDP 让 diffusion policy 更接近可解释的长程操作系统。

思考题

  1. 为什么高层语言指令直接生成短期动作容易粒度错位?
  2. 八个 primitive skills 为什么能跨任务复用?
  3. Router network 需要同时看视觉和语言,原因是什么?
  4. Skill-dependent FFN 比简单拼接 skill embedding 可能强在哪里?
  5. 如果一个任务需要“擦拭”或“扭瓶盖”,现有 skill set 是否足够?

FAQ

Q:SDP 是不是手写规则系统? A:不是。Primitive skills 是显式中间层,但 router 和 skill-conditioned diffusion policy 都是学习得到的。

Q:八个 skill 是机器人最终动作吗? A:不是。它们是短程语义动作类别,最终仍由 diffusion policy 生成连续控制信号。

Q:96.9% 是本站复现的吗? A:不是。它是论文在 LIBERO 上报告的平均成功率。

Q:可解释 skill 是否保证机器人安全? A:不保证。它只提高可观察性和结构性,安全仍需要约束、监控和验证。

进一步读什么

  • diffusion-policy:理解动作扩散基础。
  • CALVIN / LIBERO:理解长程语言条件操作 benchmark。
  • MoDE / MDT:理解 diffusion policy 的多任务和 transformer baseline。
  • trace-focused-diffusion-policy:比较 skill condition 和 trace condition。

精读补充:primitive skill 为什么既是能力层,也是调试层

SDP 的 primitive skill 不只是为了提高成功率,也是在给机器人行为加一个可检查的中间状态。端到端 policy 失败时,我们通常只能看到动作错了,却不知道模型内部为什么错。引入 skill 后,失败可以拆成两类:router 是否选错 skill,或者 diffusion policy 是否在正确 skill 下生成了错误动作。

这种可分解性对长程任务很重要。比如机器人把柠檬放进锅里失败,可能是“应该 close gripper 时还在 translate”,也可能是“close gripper 已经选对,但夹爪轨迹太偏”。前者是阶段理解问题,后者是低层控制问题。Primitive skill 让这两个问题更容易分开诊断。

不过,primitive skill 也会带来抽象边界。八个技能听起来通用,但真实世界的动作远比八类丰富。擦拭、插入、旋拧、拉链、柔性物体操作都可能需要更细或不同的 skill vocabulary。如果 skill set 太粗,router 即使选对也无法给 policy 足够指导;如果 skill set 太细,router 学习会变难。

因此,SDP 的更大启发不是“八个技能就是最终答案”,而是“语言到动作之间应该有可学习、可解释、可组合的中间层”。未来系统可能自动发现 skill,也可能允许人类定义一部分技能,再让模型学习组合。

所以这一节是想说:primitive skills 同时提升动作条件、长程分解和失败可诊断性。

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:八个 primitive skills 的原文名称;CPE 模板;router top-1 selection;skill-dependent FFN 设计;CALVIN ABC→D 76.9%、MoDE +14.5%、UniVLA +20.4%;LIBERO 96.9%、MDT +13.4%、UniVLA +4.4%;真实机器人任务设置。

原文信息

@article{gu2026primitiveskills,
  title = {Learning Diffusion Policy from Primitive Skills for Robot Manipulation},
  author = {Gu, Zhihao and Yang, Ming and Zou, Difan and Xu, Dong},
  journal = {arXiv preprint arXiv:2601.01948},
  year = {2026}
}

引用本笔记 / Cite this note
BibTeX
@online{eai_primitive_skill_diffusion_policy_2026,
  title       = {(readable note) Learning Diffusion Policy from Primitive Skills for Robot Manipulation},
  author      = {Xun, Jason},
  year        = {2026},
  note        = {Note on a 2026 paper},
  howpublished = {\url{https://estelledc.github.io/embodied-ai-reading-station/papers/primitive-skill-diffusion-policy/}},
  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?