LIBERO-Long

The 'LIBERO-Long' dataset/benchmark contains diverse human videos used to evaluate the effectiveness of learning latent actions for vision-language-action model pretraining.

Papers using LIBERO-Long (3)

Towards Long-horizon Embodied Agents with Tool-Aligned Vision-Language-Action Models2026MVP-LAM: Learning Action-Centric Latent Action via Cross-Viewpoint Reconstruction2026Atomic Action Slicing: Planner-aligned Options For Generalist VLA Agents2025