TY  - RPRT
TI  - RL$^2$-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models
AU  - Derek Ming Siang Tan
AU  - Shailesh Shailesh
AU  - Srikrishna Iyer
AU  - William Wei Jie Teo
AU  - Yuanliang Ju
AU  - Qiao Gu
AU  - Guillaume Sartoretti
PY  - 2026
UR  - https://arxiv.org/abs/2607.26991
ID  - 2607.26991
ER  -