Play2Perfect: What Matters in Dexterous Play Pretraining for Precise Assembly?

Authors: Tyler Ga Wei Lum, Kushal Kedia, C. Karen Liu, Jeannette Bohg · CoRL 2026

Licences

• Open Access
• Paper: arXiv.org perpetual, non-exclusive license 1.0

Under arXiv's licence, everything below is written from the abstract only.

Why it matters

Precise assembly with multi-fingered hands has stayed out of reach because it is contact-rich, which makes demonstrations hard to collect, and sparse-reward, which makes direct exploration with reinforcement learning intractable. The authors report a play-pretrained prior that is 33x more sample-efficient than training from scratch and reaches 60% success on 0.5 mm-clearance insertions in zero-shot sim-to-real transfer.

Proposed solution

Play2Perfect pretrains a task-agnostic policy through play on diverse objects and goals, learning reusable priors such as grasping, in-hand reorientation and pose reaching, then fine-tunes it on precise assembly so exploration concentrates on the final high-precision contacts. The authors study object diversity, training objective, trajectory diversity and goal precision as design choices.

Abstract

Multi-fingered robots promise the speed and dexterity of human hands, yet challenging problems such as precise assembly have remained out of reach. These tasks are contact-rich, making data collection for imitation learning difficult, and sparse-reward, making direct exploration with reinforcement learning (RL) intractable. Consequently, prior work has made progress by structuring the problem with specialized grippers, tool attachments, and environment fixtures. In this work, we argue that before a robot can perfect precise assembly, it must first learn to play. We further ask the question: what factors in the process of learning to play matter for precise assembly? We propose Play2Perfect, an RL framework for task-agnostic pretraining through play on diverse objects and goals, which is then perfected on precise assembly. The goal of play is to acquire reusable manipulation priors, such as grasping, in-hand reorientation and pose reaching. Finetuning then adapts this general prior to assembly, focusing exploration on the final contact-rich, high-precision interactions needed for success. We systematically study key design choices in play pretraining, including object diversity, training objective, trajectory diversity, and goal precision. We show that our prior is 33x more sample-efficient than RL training from scratch, even when provided with dense, multi-stage rewards. We demonstrate zero-shot sim-to-real transfer, achieving 60% success on tight insertions with only 0.5 mm contact clearance, and over 50% success on long-horizon multi-part assembly and screwing.

Links

• arXiv 2606.26428: https://arxiv.org/abs/2606.26428
• PDF: https://arxiv.org/pdf/2606.26428
• Code: https://github.com/kushal2000/play2perfect
• Project page: https://play2perfect.github.io/