mlsanityData to 7 Oct 2026

Human-video and latent-action pretraining

103 papers2 this weekHigh confidence in this grouping: the papers form one clear topicHigh confidence in this grouping: the papers form one clear topic · About

Keywords: action, human, robot, latent, manipulation

Key facts. Human-video and latent-action pretraining: 103 papers on arXiv since 2024 (data through 7 Oct 2026). 43 of its 103 mapped papers (42%) cite “Latent Action Pretraining from Videos” (2024, arXiv 2410.11758). 36 of its 103 mapped papers (35%) cite “UniVLA: Learning to Act Anywhere with Task-centric Latent Actions” (2025, arXiv 2505.06111). 11 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

11 new papers, 17 Sep – 7 Oct; 2 this week.

  1. This week · 2
  2. YUBI-STAG: Contact and Semantic-Rich Alignment for VLAs via Automated Video-Language Grounding also fits: Tactile, force-aware and humanoid VLA
  3. ESP: Energy-Score Policy for One-Step Multimodal Action Generation also fits: Reinforcement learning and post-training
  4. Earlier · 9
  5. NarrativeFlow: Flow-Based Vision-Language-Action Model Using Robot Velocity Fields also fits: Generalist VLA models and benchmarks
  6. FineART: Fine-Grained Annotated Robotic Trajectory Dataset and Vision-Language-Action Model for Bimanual Manipulation
  7. CrossSafe: Towards Cross-Embodiment Latent Safety Filters also fits: Safety and adversarial attacks
  8. ActGaze: Learning Action-Grounded Gaze through Counterfactual Visual Interventions for High-Precision Manipulation also fits: Failure recovery and robustness
  9. Direction-Scale Decomposition in Action Representation: Rethinking What to Tokenize for Vision-Language-Action Models also fits: 3D and spatial-aware VLA
  10. SCULPT-VLA: Learning Structured Control through Staged Action Grounding also fits: Reasoning and chain-of-thought in VLA
  11. GALA: Geometry-Aware Latent Action Modeling for Vision-Language-Action Model Pretraining across Embodiments
  12. A Sim-to-Real Integration Pipeline for Training and Deployment of Chunk-Based VLA Manipulation Policies also fits: Generalist VLA models and benchmarks
All 11
  1. HIL-UMI: Bringing Human-in-the-Loop Post-Training of Vision-Language-Action Models to Universal Manipulation Interface also fits: Reinforcement learning and post-training

Foundations

Most cited by papers here

  1. Latent Action Pretraining from Videos2024cited by 43 here
  2. UniVLA: Learning to Act Anywhere with Task-centric Latent Actions2025cited by 36 here
  3. EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos2025cited by 17 here
  4. Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videos2025cited by 17 here
  5. Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations2024cited by 13 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 78 here
  2. 2024Octo: An Open-Source Generalist Robot Policycited by 55 here
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 87 here
  4. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 11 here
  5. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 71 here
  6. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 32 here
  7. 2025DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Controlcited by 18 here
  8. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 59 here
  9. 2025GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Datacited by 9 here
  10. 2025Being-H0: Vision-Language-Action Pretraining from Large-Scale Human Videoscited by 17 here
14 more it builds on, not in this chain
  1. 2024Robotic Control via Embodied Chain-of-Thought Reasoningcited by 13 here
  2. 2024Latent Action Pretraining from Videoscited by 43 here
  3. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 20 here
  4. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 9 here
  5. 2024Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representationscited by 13 here
  6. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 18 here
  7. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 32 here
  8. 2025DexGraspVLA: A Vision-Language-Action Framework Towards General Dexterous Graspingcited by 9 here
  9. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 18 here
  10. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 36 here
  11. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 14 here
  12. 2025EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videoscited by 17 here
  13. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 20 here
  14. 2026StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developingcited by 10 here

The 24 most cited of 253.

Browse

30 of 103, ranked 6–35

  1. 6DexGraspVLA: A Vision-Language-Action Framework Towards General Dexterous Grasping20259
  2. 7GraspVLA: a Grasping Foundation Model Pre-trained on Billion-scale Synthetic Action Data20259
  3. 8Scalable Vision-Language-Action Model Pretraining for Robotic Manipulation with Real-Life Human Activity Videos20258
  4. 9Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization20267
  5. 10mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAs20255
  6. 11Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments20264
  7. 12CLAP: Contrastive Latent Action Pretraining for Learning Vision-Language-Action Models from Human Videos20263
  8. 13A Systematic Study of Data Modalities and Strategies for Co-training Large Behavior Models for Robot Manipulation20263
  9. 14Latent Action Pretraining Through World Modeling20253
  10. 15End-to-End Dexterous Arm-Hand VLA Policies via Shared Autonomy: VR Teleoperation Augmented by Autonomous Hand VLA Policy for Efficient Data Collection20253
All 30
  1. 16LatBot: Distilling Universal Latent Actions for Vision-Language-Action Models20253
  2. 17Emergence of Human to Robot Transfer in Vision-Language-Action Models20253
  3. 18Joint-Aligned Latent Action: Towards Scalable VLA Pretraining in the Wild20262
  4. 19LARY: A Latent Action Representation Yielding Benchmark for Generalizable Vision-to-Action Alignment20262
  5. 20RotVLA: Rotational Latent Action for Vision-Language-Action Model20262
  6. 21FrameSkip: Learning from Fewer but More Informative Frames in VLA Training20262
  7. 22DexVLG: Dexterous Vision-Language-Grasp Model at Scale20252
  8. 23TwinVLA: Data-Efficient Bimanual Manipulation with Twin Single-Arm Vision-Language-Action Models20252
  9. 24See Once, Then Act: Vision-Language-Action Model with Task Learning from One-Shot Video Demonstrations20252
  10. 25Bi-VLA: Vision-Language-Action Model-Based System for Bimanual Robotic Dexterous Manipulations20242
  11. 26Vision-Language Models Unlock Task-Centric Latent Actions20261
  12. 27ConLA: Contrastive Latent Action Learning from Human Videos for Robotic Manipulation20261
  13. 28MVP-LAM: Learning Action-Centric Latent Action via Cross-Viewpoint Reconstruction20261
  14. 29LAP: Language-Action Pre-Training Enables Zero-shot Cross-Embodiment Transfer20261
  15. 30MOTIF: Learning Action Motifs for Few-shot Cross-Embodiment Transfer20261
  16. 31UniLACT: Depth-Aware RGB Latent Action Learning for Vision-Language-Action Models20261
  17. 32SkillVLA: Tackling Combinatorial Diversity in Dual-Arm Manipulation via Skill Reuse20261
  18. 33Cross-Hand Latent Representation for Vision-Language-Action Models20261
  19. 34DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control20261
  20. 35Disentangled Robot Learning via Separate Forward and Inverse Dynamics Pretraining20261