zaleni/wsa1-robot-control

WSA-Base — World-Spatial-Action Robot Control

Interactive demo of **WSA**, a 3D-centric World-Spatial-Action embodied foundation model for generalizable robot control (code).

Provide a head camera view and a wrist camera view of a tabletop robot scene plus a natural-language instruction; the WSA-Base policy predicts a short chunk of 7-DoF end-effector actions and can decode its action-conditioned world-model prediction of the next head-camera frame.

• —Model: `zaleni/WSA-Base-LIBERO` (3B, Qwen3-VL-2B backbone, LIBERO franka adapter).
• —Example scenes are real LIBERO-Goal frames.