mlsanityData to 7 Oct 2026

An Empirical Study on What Matters for Viewpoint-Generalizable Policies in Visual Imitation Learning

arXiv 2609.32762 ↗First posted 26 Sep 2026cs.RO cs.CV cs.LGcited by 0

On the map: Offline RL and imitation learning from data → Robot manipulation imitation learning

Abstract (arXiv)

Visual imitation learning is a promising approach to training robot manipulation policies capable of completing a wide variety of tasks. However, policies today remain brittle to viewpoint perturbations, making deployment in diverse environments a challenge. We present a controlled empirical study of which design choices allow visuomotor policies to generalize across viewpoints. We find that viewpoint generalization improves when dense visual tokens are retained and the action head participates in geometric reasoning. On a suite of simulated tasks that span a wide range of camera poses, we show that these design choices yield a policy that remains performant across viewpoints. As a practical consequence, a policy trained with these design choices also transfers zero-shot from simulation to the real world under random camera configurations.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Robot manipulation imitation learning, one of 13 directions in Offline RL and imitation learning from data. Builds on 34 of its 38 references that are in the map. First posted on arXiv 26 Sep 2026.

Builds on

34 of 38 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Do You Know Where Your Camera Is? View-Invariant Policy Learning with Camera Conditioning2 Oct
  2. 2025VGGT-DP: Generalizable Robot Control via Vision Foundation Models23 Sep
  3. 20253D FlowMatch Actor: Unified 3D Policy for Single- and Dual-Arm Manipulation14 Aug
  4. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware23 Apr

Other references · 30

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  3. 2022Scalable Diffusion Models with Transformerscited by 5,931
  4. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  5. 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,337
  6. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  7. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  8. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  9. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  10. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 801
  11. 2025Depth Anything 3: Recovering the Visual Space from Any Viewscited by 626
  12. 2020SAPIEN: A SimulAted Part-based Interactive ENvironmentcited by 584
18 more
  1. 2022Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulationcited by 545
  2. 20243D Diffuser Actor: Policy Diffusion with 3D Scene Representationscited by 278
  3. 2023RVT: Robotic View Transformer for 3D Object Manipulationcited by 224
  4. 2024RVT-2: Learning Precise Manipulation from Few Demonstrationscited by 135
  5. 2024From Imitation to Refinement -- Residual RL for Precise Assemblycited by 80
  6. 2023Multi-View Masked World Models for Visual Robotic Manipulationcited by 78
  7. 2024DextrAH-RGB: Visuomotor Policies to Grasp Anything with Dexterous Handscited by 37
  8. 2024Learning to Manipulate Anywhere: A Visual Generalizable Framework For Reinforcement Learningcited by 36
  9. 2024View-Invariant Policy Learning via Zero-Shot Novel View Synthesiscited by 35
  10. 2025Fast-FoundationStereo: Real-Time Zero-Shot Stereo Matchingcited by 23
  11. 2024Local Policies Enable Zero-shot Long-horizon Manipulationcited by 21
  12. 2025Opening the Sim-to-Real Door for Humanoid Pixel-to-Action Policy Transfercited by 13
  13. 2026MolmoB0T: Large-Scale Simulation Enables Zero-Shot Manipulationcited by 11
  14. 2025ArticuBot: Learning Universal Articulated Object Manipulation Policy via Large Scale Simulationcited by 10
  15. 2025CLASS: Contrastive Learning via Action Sequence Supervision for Robot Manipulationcited by 10
  16. 2026Emergent Dexterity via Diverse Resets and Large-Scale Reinforcement Learningcited by 9
  17. 2026ReMAP-DP: Reprojected Multi-view Aligned PointMaps for Diffusion Policycited by 2
  18. 2026GHOST: Hierarchical Sub-Goal Policies for Generalizing Robot Manipulationcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2021Seeing All the Angles: Learning Multiview Manipulation Policies for Contact-Rich Tasks from Demonstrations
  2. 2026Learning to See While Learning to Act: Diffusion Models for Active Perception in Robot Imitation
  3. 2025Viewpoint-Agnostic Manipulation Policies with Strategic Vantage Selection
  4. 2026Multi-Camera View Scaling for Data-Efficient Robot Imitation Learning
  5. 2025Zero-Shot Visual Generalization in Robot Manipulation
  6. 2023Decomposing the Generalization Gap in Imitation Learning for Visual Robotic Manipulation
  7. 20243D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations
  8. 2025Imagination at Inference: Synthesizing In-Hand Views for Robust Visuomotor Policy Inference