Research map / Reinforcement learning / Policy gradient and actor-critic
Guarded Policy Optimization with Imperfect Online Demonstrations
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 value-based rather than action-similarity-based intervention | TS2C · MetaDrive | training reward | Helped |
| Use value-based rather than action-similarity-based intervention | TS2C · MetaDrive | training cost | Helped |
| Add a medium-performance teacher through online shared control | TS2C · MetaDrive | test reward | Helped |
| Add a medium-performance teacher through online shared control | TS2C · MetaDrive | training cost | Helped |
| Use value-based rather than action-based intervention | TS2C · MuJoCo Pendulum | return | Mixed |
| Use value-based rather than action-based intervention | TS2C · MuJoCo Hopper and Walker | return | Helped |
| Add intervention-minimization cost to the student update (ablation) | TS2C · MetaDrive | performance | Helped |
| Use ensembled teacher value networks (ablation) | TS2C · MetaDrive | performance | Helped |
| Vary the variance threshold in the intervention function (ablation) | TS2C · MetaDrive | performance | No clear effect |
| Replace value-based intervention with action-based intervention (ablation) | TS2C · MetaDrive | performance | Hurt |
| Vary the value-based intervention threshold (ablation) | TS2C · MetaDrive | performance | No clear effect |
Closest papers by meaning
- TGRL: An Algorithm for Teacher Guided Reinforcement Learning (ICML 2023 · similarity 0.90)
- Learner-aware Teaching: Inverse Reinforcement Learning with Preferences and Constraints (NeurIPS 2019 · similarity 0.86)
- Hybrid Policy Optimization from Imperfect Demonstrations (NeurIPS 2023 · similarity 0.85)
- Student-Informed Teacher Training (ICLR 2025 · similarity 0.85)
- Unsupervised Training Sequence Design: Efficient and Generalizable Agent Training (AAAI 2024 · similarity 0.85)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.85)
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.