Research map / Reinforcement learning / Policy gradient and actor-critic
Ready Policy One: World Building Through Active Learning
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 adaptive reward-plus-uncertainty exploration instead of fixed V+R | RP1 · HalfCheetah at 10^4 and 10^5 timesteps | return: -1.1 -> 100.51; 283.27 -> 390.49 | Helped |
| Use adaptive reward-plus-uncertainty exploration instead of fixed V+R | RP1 · Ant at 10^4 and 10^5 timesteps | return: 95.07 -> 113.63; 217.08 -> 238.4 | Helped |
| Use adaptive reward-plus-uncertainty exploration instead of fixed V+R | RP1 · Swimmer at 10^4 and 10^5 timesteps | return: 1.07 -> 3.24; 62.89 -> 64.19 | Mixed |
| Use adaptive reward-plus-uncertainty exploration instead of fixed V+R | RP1 · Hopper at 10^4 and 10^5 timesteps | return: 139.03 -> 322.22; 570.09 -> 619.73 | Helped |
| Use reward variance rather than next-state variance | RP1 · HalfCheetah, Ant, Swimmer, and Hopper | return: 319.41 -> 390.49; 214.33 -> 238.4; 63.47 -> 64.19; 549.19 -> 619.73 | Helped |
| Add adaptive exploration rather than fixing λ=0 (ablation) | RP1 · HalfCheetah, Ant, Swimmer, and Hopper | return: 319.14 -> 390.49; 223.21 -> 238.4; 41.14 -> 64.19; 603.52 -> 619.73 | Helped |
| Add early stopping rather than using a fixed batch size (ablation) | RP1 · HalfCheetah, Ant, Swimmer, and Hopper | return: 247.82 -> 390.49; 197.95 -> 238.4; 41.04 -> 64.19; 595.77 -> 619.73 | Helped |
Closest papers by meaning
- Jump-Start Reinforcement Learning (ICML 2023 · similarity 0.90)
- Model-Free Active Exploration in Reinforcement Learning (NeurIPS 2023 · similarity 0.89)
- PWM: Policy Learning with Multi-Task World Models (ICLR 2025 · similarity 0.89)
- PC-MLP: Model-based Reinforcement Learning with Policy Cover Guided Exploration (ICML 2021 · similarity 0.89)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.89)
- MOPO: Model-based Offline Policy Optimization (NeurIPS 2020 · similarity 0.89)
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.