ManiSkill‑ViTac Challenge 2026 is a real‑robot challenge on bimanual manipulation with fused visual
and tactile sensing in contact‑rich tasks. It builds on the
ViTaMIn‑B
visuo‑tactile interface, standardizing its hardware and data format into a common benchmark for learning
and evaluating multi‑modal policies.
The 2026 edition focuses on three scenarios: environment‑driven visuo‑tactile manipulation, language‑guided
visuo‑tactile‑language manipulation, and robot‑free visuo‑tactile data collection with unified real‑robot
evaluation. Together, they target core capabilities in visuo‑tactile perception, bimanual coordination and
language‑grounded policy learning toward robust, generalizable embodied agents for real‑world applications.