Firehose

30 JUL 2026 · Paper

RL^2-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models

This paper develops a method to improve the performance of Vision-Language-Action models by adapting their steering strategy at test time, allowing them to generalize better to new tasks and domains. Practitioners can benefit from this approach by improving the robustness of their VLA models in real-world applications.

reinforcement learning vision-language-action models adaptive steering continual learning test-time scaling

Browse by tag

RL^2-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models