Research map / Reinforcement learning / Visual and representation RL
Towards Generalization and Simplicity in Continuous Control
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 |
|---|---|---|---|
| Use an RBF policy instead of a neural-network policy | Continuous-control policies · OpenAI Gym-v1 tasks | return | Helped |
| Use a linear policy instead of a neural-network policy | Continuous-control policies · OpenAI Gym-v1 tasks | return | Mixed |
| Use diverse initial states during training | Policies for modified control tasks · Modified tasks | perturbation resistance | Helped |
| Train with a narrow initial-state distribution | Policies for control tasks · Standard benchmark tasks | perturbation resistance | Hurt |
| Increase the number of random Fourier features (ablation) | RBF policy · Modified Walker | score | Helped |
| Use an RBF policy rather than a linear policy (ablation) | Policies trained on diverse initial states · Modified Walker | walking efficiency | Helped |
| Use diverse initialization rather than peaked initialization (ablation) | Policies evaluated under external perturbations · Modified swimmer, walker, and hopper | distance traveled | Helped |
| Precondition policy-gradient updates with natural-gradient iterations (ablation) | Natural policy gradient · Modified Walker | learning progress | Helped |
| Use an intermediate number of conjugate-gradient steps, such as 20 (ablation) | Natural policy gradient · Large-scale control problems | wall-clock performance | Helped |
| Use normalized rather than un-normalized step size (ablation) | Natural policy gradient · Diverse swimmer, hopper, and walker | learning stability | Helped |
| Use GAE with λ=0.97 rather than a low-bias estimate (ablation) | Policy-gradient advantage estimation · Continuous-control tasks | initial learning speed | Helped |
| Use a higher-λ, lower-bias advantage estimate rather than λ=0.97 (ablation) | Policy-gradient advantage estimation · Continuous-control tasks | asymptotic performance | Helped |
Closest papers by meaning
- LipsNet++: Unifying Filter and Controller into a Policy Network (ICML 2025 · similarity 0.88)
- Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (AAAI 2025 · similarity 0.88)
- Enhancing Robustness in Deep Reinforcement Learning: A Lyapunov Exponent Approach (NeurIPS 2024 · similarity 0.87)
- Active Fine-Tuning of Multi-Task Policies (ICML 2025 · similarity 0.87)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.87)
- 1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities (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.