Research map / Reinforcement learning / Visual and representation RL
Autonomous Reinforcement Learning: Formalism and Benchmarking
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 |
|---|---|---|---|
| increase the interval between environment resets from 1000 to 200,000 steps (ablation) | SAC on cheetah · DeepMind Control cheetah | policy return | Hurt |
| add a biased TD update that resets the target to reward every 1000 steps (ablation) | SAC · DeepMind Control cheetah | policy return | Helped |
| apply the biased TD update at longer training horizons (ablation) | SAC · DeepMind Control fish | policy return | Hurt |
| use autonomous rather than oracle-reset training | existing RL methods · EARL, final deployed return | return | Hurt |
| use VaPRL instead of oracle RL | Tabletop-Organization · Tabletop-Organization | final deployed return: 0.80 (0.11) -> 0.98 (0.02) | Helped |
| use R3L instead of oracle RL | Minitaur-Pen · Minitaur-Pen | final deployed return: -41.50 (3.40) -> -186.30 (34.79) | Hurt |
| optimize for continuing lifetime reward rather than deployed policy performance | autonomous RL evaluation · EARL | performance | Mixed |
| train policies autonomously rather than with oracle resets | FBRL and VaPRL · Tabletop-Organization with uniform initial states | performance depreciation: 2% and 14.3% vs 37.4% | Helped |
Closest papers by meaning
- A State-Distribution Matching Approach to Non-Episodic Reinforcement Learning (ICML 2022 · similarity 0.92)
- Demonstration-free Autonomous Reinforcement Learning via Implicit and Bidirectional Curriculum (ICML 2023 · similarity 0.91)
- DynaMITE-RL: A Dynamic Model for Improved Temporal Meta-Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Dual RL: Unification and New Methods for Reinforcement and Imitation Learning (ICLR 2024 · similarity 0.89)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.89)
- When Do Skills Help Reinforcement Learning? A Theoretical Analysis of Temporal Abstractions (ICML 2024 · similarity 0.88)
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.