VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning
VIGOR: Zero-Shot Visual Generalization via Latent-Space Consistency in Model-Based Reinforcement Learning
This research presents VIGOR, a technique enabling model-based reinforcement learning agents to generalize zero-shot to unseen visual distractions such as changes in background, lighting, or camera settings.
The paper introduces VIGOR, a method for zero-shot visual generalization in model-based reinforcement learning by enforcing latent-space consistency against distractions like lighting or background changes; it reports quantitative improvements on DeepMind Control Suite and Robosuite benchmarks using TD-MPC2 and DreamerV3 backbones.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- code: Reported pending Open source S6
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
model-based baselines in zero-shot generalization, surpassing the second-best baseline by 3.4% on DMC and 43.6% on Robosuite, with sample efficiency comparable to its backbone and fewer environment interactions than the other baselines. Ablations show that VIGOR is augmentation-agnostic: augmentations from distinct perturbation families preserve strong generalization, confirming that latent-space consistency, not th
Open source S5
-only and weak-to-strong latent views within a batch; (ii) dynamics-level consistency, which enforces augmentation-invariant transition predictions via direct latent regression; and (iii) encoder-level stabilization, which prevents encoder drift under the cross-augmentation supervision of (ii). On the DeepMind Control Suite (DMC) and Robosuite, VIGOR instantiated on TD-MPC2 outperforms state-of-the-art model-free and
Open source S4
e augmentation choice, drives robustness. A second instantiation on DreamerV3, a structurally distinct recurrent state-space model, also improves zero-shot generalization on both benchmarks, indicating backbone-agnostic benefits. Code and videos: https://vigor-gen.github.io/.
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org