ManiSkill-3

ManiSkill3 is a benchmark dataset used to evaluate robotic control tasks in deep reinforcement learning, focusing on the performance of agents in manipulating objects with varying complexity.

Papers using ManiSkill-3 (4)

Squint: Fast Visual Reinforcement Learning for Sim-to-Real Robotics2026Refined Policy Distillation: From VLA Generalists to RL Experts2025 · 1 citesGaze on the Prize: Shaping Visual Attention with Return-Guided Contrastive Learning2025Decoupling Task and Behavior: A Two-Stage Reward Curriculum in Reinforcement Learning for Robotics2026