VIMA-Bench: leaderboard
17 tabletop robot-manipulation task templates in PyBullet with prompts that mix text and images, tested at four generalization levels up to unseen objects and tasks; Stanford and NVIDIA, ICML 2023.
Source: github.com.
Interactive version: theaggregate.ai/benchmark?slug=vima-bench · How It Works · Data refreshed daily, snapshot 2026-09-05.