mlsanityData to 14 Oct 2026

Self-Improving VLA Policies: Selected Diffusion Noise for Spurious-Robust Action Smoothing

arXiv 2606.14084 ↗First posted 12 Jun 2026cs.ROcited by 1

On the map: Vision-language-action models → Efficient VLA inference and adaptation

Abstract (arXiv)

Diffusion-based Vision-Language-Action (VLA) policies enable strong generalization in robotic manipulation, but remain sensitive to spurious visual correlations and noisy action generation, leading to brittle behavior under perturbations. We introduce Selected Diffusion Noise (SDN), a simple, training-free test-time method that improves both robustness and success rate by leveraging the diffusion noise space as a controllable degree of freedom. SDN dynamically samples noise vectors that are maximally separated from a reference set to mitigate reliance on spurious cues, while selecting candidates that yield more coherent action trajectories. This dual objective encourages stable behavior even under object-masked observations and reduces action jitter without modifying model parameters. We evaluate SDN on two simulation benchmarks (Google Robot, Widow-X) and two real-world robotic datasets across multiple VLA policies, including pi_0, Groot-N1.5, and Groot-N1.6. SDN consistently improves success rates by +8% in simulation and +10% in real-world settings, while producing smoother and more stable actions. Our results highlight that diffusion noise selection can serve as an effective and general mechanism for enhancing VLA policies at test time.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 45 of its 54 references that are in the map. First posted on arXiv 12 Jun 2026.

Builds on

45 of 54 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Verifier-free Test-Time Sampling for Vision-Language-Action Models7 Oct

Other references · 44

  1. 2022Classifier-Free Diffusion Guidancecited by 5,847
  2. 2022Flow Matching for Generative Modelingcited by 5,702
  3. 2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,332
  4. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 3,080
  5. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  6. 2024SAM 2: Segment Anything in Images and Videoscited by 2,863
  7. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  8. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  9. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,225
  10. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 883
  11. 2022Compositional Visual Generation with Composable Diffusion Modelscited by 624
  12. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 495
32 more
  1. 2023Universal Guidance for Diffusion Modelscited by 437
  2. 2024Flow Matching Guide and Codecited by 301
  3. 2019Causal Confusion in Imitation Learningcited by 288
  4. 2023Goal-Conditioned Imitation Learning using Score-based Diffusion Policiescited by 236
  5. 2025Real-Time Execution of Action Chunking Flow Policiescited by 216
  6. 2025Steering Your Diffusion Policy with Latent Space Reinforcement Learningcited by 169
  7. 2020Self-Supervised Policy Adaptation during Deploymentcited by 123
  8. 2025Reactive Diffusion Policy: Slow-Fast Visual-Tactile Policy Learning for Contact-Rich Manipulationcited by 108
  9. 2023Stay on topic with Classifier-Free Guidancecited by 91
  10. 2024DINO-X: A Unified Vision Model for Open-World Object Detection and Understandingcited by 89
  11. 2022VLMbench: A Compositional Benchmark for Vision-and-Language Manipulationcited by 68
  12. 2026RoboReward: General-Purpose Vision-Language Reward Models for Roboticscited by 53
  13. 2023Smooth Diffusion: Crafting Smooth Latent Spaces in Diffusion Modelscited by 49
  14. 2024A Noise is Worth Diffusion Guidancecited by 49
  15. 2024Learning to Discretize Denoising Diffusion ODEscited by 41
  16. 2026LaST0: Latent Spatio-Temporal Chain-of-Thought for Robotic Vision-Language-Action Modelcited by 31
  17. 2025Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distributioncited by 25
  18. 2025Align-Then-stEer: Adapting the Vision-Language Action Models through Unified Latent Guidancecited by 20
  19. 2025Rethinking Progression of Memory State in Robotic Manipulation: An Object-Centric Perspectivecited by 20
  20. 2026Global Prior Meets Local Consistency: Dual-Memory Augmented Vision-Language-Action Model for Efficient Robotic Manipulationcited by 20
  21. 2024Heavy-Tailed Diffusion with Denoising L'evy Probabilistic Modelscited by 18
  22. 2026VLS: Steering Pretrained Robot Policies via Vision-Language Modelscited by 15
  23. 2026SteerVLA: Steering Vision-Language-Action Models in Long-Tail Driving Scenarioscited by 11
  24. 2025SlotVLA: Towards Modeling of Object-Relation Representations in Robotic Manipulationcited by 9
  25. 2024DISCO: Language-Guided Manipulation with Diffusion Policies and Constrained Inpaintingcited by 8
  26. 2025Policy Contrastive Decoding for Robotic Foundation Modelscited by 8
  27. 2025Towards Deploying VLA without Fine-Tuning: Plug-and-Play Inference-Time VLA Policy Steering via Embodied Evolutionary Diffusioncited by 8
  28. 2025Improving Generative Behavior Cloning via Self-Guidance and Adaptive Chunkingcited by 7
  29. 2024On the Performance of Jerk-Constrained Time-Optimal Trajectory Planning for Industrial Manipulatorscited by 4
  30. 2024Segmentation-Free Guidance for Text-to-Image Diffusion Modelscited by 3
  31. 2026Latent Policy Steering through One-Step Flow Policiescited by 2
  32. 2026PPGuide: Steering Diffusion Policies with Performance Predictive Guidancecited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026RoboGaze: Evaluating Robot World Models via Structured Vision-Language Analysiscited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAs
  2. 2025E0: Enhancing Generalization and Fine-Grained Control in VLA Models via Tweedie Discrete Diffusion
  3. 2026dVLA-RL: Reinforcement Learning over Denoising Trajectories for Discrete Diffusion Vision-Language-Action Models
  4. 2026Reshaping Action Error Distributions for Reliable Vision-Language-Action Models
  5. 2026UniSteer: Unified Noise Steering for Efficient Human-Guided VLA Adaptation
  6. 2026StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models
  7. 2026Immiscible Diffusion Policy: Preserving Multimodal Robot Actions through Label-Free Noise Assignment
  8. 2026REACT: Rolling Denoising and Dual Decoupling for Reactive Robot Control with VLA Models