Research map / Reinforcement learning / Visual and representation RL
Risk-Aware Transfer in Reinforcement Learning using Successor Features
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| risk-averse (beta=-0.1) instead of risk-neutral policy evaluation in GPI | GPI with successor features · Grid with traps, target costs (20,0) | optimality of risk-adjusted policy | Helped |
| RaSFQL versus SFQL (risk-averse versus risk-neutral) | Successor feature Q-learning · Four-Room | cumulative reward | Hurt |
| RaSFQL versus SFQL | Successor feature Q-learning · Four-Room | number of failures | Helped |
| RaSFQL versus RaPRQL and PRQL (GPI/GPE vs policy reuse) | RaSFQL · Four-Room | cumulative reward | Helped |
| RaSFQL versus RaPRQL | RaSFQL · Four-Room | number of failures | Helped |
| RaSFC51 versus RaC51 and C51 on test tasks | RaSFC51 · Reacher test tasks | normalized return | Helped |
| RaSFC51 versus SFC51 | RaSFC51 · Reacher test tasks | number of failures | Helped |
| universal value function (RaC51) instead of successor features (ablation) | RaSFC51 · Reacher training tasks | return | Hurt |
| increase magnitude of beta (ablation) | RaSFC51 and RaC51 · Reacher | normalized test return | Hurt |
| RaSFC51 versus RaC51 across beta values (ablation) | RaSFC51 · Reacher | normalized test return | Helped |
| replace DQN by C51 for learning successor features (ablation) | SFDQN · Reacher | normalized test return | Helped |
Closest papers by meaning
- Successor Features for Transfer in Reinforcement Learning (NeurIPS 2017 · similarity 0.93)
- A New Representation of Successor Features for Transfer across Dissimilar Environments (ICML 2021 · similarity 0.90)
- Optimistic Linear Support and Successor Features as a Basis for Optimal Policy Transfer (ICML 2022 · similarity 0.90)
- Risk-Conditioned Reinforcement Learning: A Generalized Approach for Adapting to Varying Risk Measures (AAAI 2024 · similarity 0.90)
- Distributional Successor Features Enable Zero-Shot Policy Optimization (NeurIPS 2024 · similarity 0.88)
- Reward-Aware Proto-Representations in Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.