Zero-shot Reinforcement Learning From Low Quality Data

Abstract

Zero-shot reinforcement learning (RL) promises to provide agents that can perform any task in an environment after an offline, reward-free pre-training phase. Methods leveraging successor measures and successor features have shown strong performance in this setting, but require access to large heterogenous datasets for pre-training which cannot be expected for most real problems. Here, we explore how the performance of zero-shot RL methods degrades when trained on small homogeneous datasets, and propose fixes inspired by conservatism, a well-established feature of performant single-task offline RL algorithms. We evaluate our proposals across various datasets, domains and tasks, and show that conservative zero-shot RL algorithms outperform their non-conservative counterparts on low quality datasets, and perform no worse on high quality datasets. Somewhat surprisingly, our proposals also outperform baselines that get to see the task during training. Our code is available via https://enje

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyjeen2023zero

Related papers

• On Zero-shot Reinforcement Learning (2025)0.00
• Does Zero-shot Reinforcement Learning Exist? (2022)0.00
• A Unified Framework For Zero-shot Reinforcement Learning (2025)0.00
• Long-horizon Model-based Offline Reinforcement Learning Without Conservatism (2025)0.00
• Fewer May Be Better: Enhancing Offline Reinforcement Learning With Reduced Dataset (2025)0.00
• Model-based Offline Reinforcement Learning With Lower Expectile Q-learning (2024)0.00
• Learning From Sparse Offline Datasets Via Conservative Density Estimation (2024)0.00
• An Optimistic Perspective On Offline Reinforcement Learning (2019)0.00