VIMA-Bench

Roadmap

Manipulation specified by interleaved text-and-image multimodal prompts.

Status

This eval is catalogued and on the roadmap. The protocols are stable — implementing it is an EvalRunner with a catalog entry.