Research map / Reinforcement learning / Policy gradient and actor-critic
Utilizing Prior Solutions for Reward Shaping and Composition in Entropy-Regularized Reinforcement 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 potential-based reward shaping from a previously solved task | Entropy-regularized RL · Modified Frozen Lake mazes | Bellman backup convergence | Helped |
| Learn the corrective value function K* instead of learning the composed task value function directly | Entropy-regularized RL task composition · AND composition in a spiral-shaped maze | Bellman backup convergence | Helped |
| Use the zero-shot approximation from transformed primitive value functions | Entropy-regularized RL task composition · OR composition in a spiral-shaped maze | Approximation to optimal solution | Helped |
| Learn the corrective value function K* rather than the composed task value function directly | Entropy-regularized RL task composition · OR composition in a spiral-shaped maze | Bellman backup convergence | Helped |
| Apply reward shaping using a solved task as the maze size increases | Entropy-regularized RL · Modified Frozen Lake mazes of varying sizes | Bellman backup convergence | Helped |
| Apply reward shaping using a solved task while varying wall height | Entropy-regularized RL · Modified Frozen Lake mazes with varying wall heights | Bellman backup convergence | Helped |
| Vary wall height, used as a proxy for overlap between the solved and target task value functions | Reward shaping · Modified Frozen Lake mazes with varying wall heights | Convergence rate | No clear effect |
Closest papers by meaning
- Composing Value Functions in Reinforcement Learning (ICML 2019 · similarity 0.91)
- Convergence Theorems for Entropy-Regularized and Distributional Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- State Entropy Regularization for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Intrinsic Benefits of Categorical Distributional Loss: Uncertainty-aware Regularized Exploration in Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- ORSO: Accelerating Reward Design via Online Reward Selection and Policy Optimization (ICLR 2025 · similarity 0.88)
- Reward Shaping for Reinforcement Learning with An Assistant Reward Agent (ICML 2024 · 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.