mlsanityData to 14 Oct 2026

SmolWAM: A Latent World-Action Model for Robotic Manipulation on Edge Devices

arXiv 2610.13740 ↗First posted 9 Oct 2026cs.RO cs.LGcited by 0

On the map: World models → Embodied and robot world models

Abstract (arXiv)

World-action models (WAMs) predict a robot's actions together with the observations those actions will produce. However, the strongest models rely on video diffusion transformers that are far too computationally expensive for resource-constrained edge devices. We present SmolWAM, a latent WAM that we derive from the 2B-parameter Cosmos Policy by keeping its pretrained trunk and rebuilding its inference interface: single-frame visual latents, cached text features that preserve the trunk's attention sinks, 12 of 28 transformer blocks, and 1,008 tokens from which actions and future visual latents are read out in one rectified-flow step. Trained in 3.6 GPU-hours, SmolWAM succeeds in 95.5% of 400 LIBERO episodes at 57 ms and 2.3 TFLOP per 16-action chunk with 2.4 GiB of GPU memory, 11x faster than its teacher with 18.5x fewer FLOPs and 3.2x less memory, and its world head still forecasts scene motion.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Embodied and robot world models, one of 11 directions in World models. Builds on 43 of its 45 references that are in the map. First posted on arXiv 9 Oct 2026.

Builds on

43 of 45 references.

Closest ancestors · 11

Same direction, newest first

  1. 2026World-Action Models for Robot Learning and Control: A Survey13 Sep
  2. 2026Spatially Aware World Action Model via Geometric Latent Diffusion2 Sep
  3. 2026ABot-M0.5: Unified Mobility-and-Manipulation World Action Model1 Jul
  4. 2026Light-WAM: Efficient World Action Models with State-Fusion Action Decoding6 Jun
  5. 2026OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulation7 May
  6. 2026Fast-WAM: Do World Action Models Need Test-time Future Imagination?17 Mar
  7. 2026World Action Models are Zero-shot Policies17 Feb
  8. 2025WorldVLA: Towards Autoregressive Action World Model26 Jun
  9. 2025V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning11 Jun
  10. 2025FLARE: Robot Learning with Implicit World Modeling21 May
  11. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusion7 Mar

Other references · 32

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  2. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,599
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  4. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  5. 2022Scalable Diffusion Models with Transformerscited by 6,054
  6. 2022Flow Matching for Generative Modelingcited by 5,702
  7. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  8. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 3,080
  9. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  10. 2025Wan: Open and Advanced Large-Scale Video Generative Modelscited by 2,708
  11. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  12. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
20 more
  1. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,186
  2. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
  3. 2024Octo: An Open-Source Generalist Robot Policycited by 1,484
  4. 2024Depth Anything V2cited by 1,422
  5. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  6. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 883
  7. 2025Cosmos World Foundation Model Platform for Physical AIcited by 864
  8. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 665
  9. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 530
  10. 2024DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planningcited by 386
  11. 2024ShortGPT: Layers in Large Language Models are More Redundant Than You Expectcited by 383
  12. 2023Unleashing Large-Scale Video Generative Pre-training for Visual Robot Manipulationcited by 364
  13. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 339
  14. 2026Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planningcited by 300
  15. 2026LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixelscited by 273
  16. 2025Real-Time Execution of Action Chunking Flow Policiescited by 216
  17. 2024Compact Language Models via Pruning and Knowledge Distillationcited by 148
  18. 2024Prediction with Action: Visual Policy Learning via Joint Denoising Processcited by 70
  19. 2026Fast LeWorldModelcited by 15
  20. 2026TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAMcited by 10

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026MemoryWAM: Efficient World Action Modeling with Persistent Memory
  2. 2026AcrossWAM1.0:A Modular Latent World-Action Stack for Compact Robot Policies
  3. 2026DELE-w0.5: Inferring Action from Future Latent State for Robotic Manipulation
  4. 2026LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies
  5. 2026DIM-WAM: World-Action Modeling with Diverse Historical Event Memory
  6. 2026MARS: Systematic Multi-Level Acceleration for World Action Models
  7. 2026Efficient-WAM: A 1B-Parameter World-Action Model with Low-Cost Future Imagination
  8. 2026τ0-WM: A Unified Video-Action World Model for Robotic Manipulation