Zero-Shot RL & Generalization Reading List

Curated by Mouhssine Rifaki | Stanford Electrical Engineering | Last updated August 2026

Zero-shot generalization tests whether a policy trained on one distribution can succeed on unseen levels, dynamics, or goals without fine-tuning. These papers define the benchmarks and methods that move RL from memorization to generalization.

Zero-Shot RL & Generalization: 10 key papers

1. Fast Task Inference with Variational Intrinsic Successor Features
   Hansen et al. arXiv 2019.
2. Successor Features for Transfer in Reinforcement Learning
   Barreto et al. arXiv 2016.
3. Does Zero-Shot Reinforcement Learning Exist?
   Touati et al. arXiv 2022.
4. Learning One Representation to Optimize All Rewards
   Touati and Ollivier. arXiv 2021.
5. Universal Successor Features Approximators
   Borsa et al. arXiv 2018.
6. Transfer in Deep Reinforcement Learning Using Successor Features and Generalised Policy Improvement
   Barreto et al. arXiv 2019.
7. Zero-Shot Task Generalization with Multi-Task Deep Reinforcement Learning
   Oh et al. arXiv 2017.
8. Zero-Shot Reinforcement Learning from Low Quality Data
   Jeen et al. arXiv 2023.
9. Reward-Free Exploration for Reinforcement Learning
   Jin et al. arXiv 2020.
10. Fast Adaptation via Policy-Dynamics Value Functions
    Raileanu et al. arXiv 2020.

Related reading lists

• Goal-Conditioned Reinforcement Learning
• Multi-Task & Transfer Reinforcement Learning
• Quality-Diversity & Novelty Search
• Domain Adaptation & Domain Generalization
• Generalisation in RL
• Bayesian RL & Thompson Sampling
• Information Theory & RL
• Kernel Methods & RKHS