Rapier/WASM

Follow-up: VSArena now has a proper VLA track (camera + language, no privileged state) — repo and docs are public

VSArena, an open-source robotics benchmark platform, has launched a new Vision-Language-Action (VLA) track that restricts policy inputs to only camera images and language instructions—removing privileged state information like cube poses—to better reflect real-world embodied AI constraints.