Research map / Reinforcement learning / Policy gradient and actor-critic
Task-Agnostic Exploration via Policy Gradient of a Non-Parametric State Entropy Estimate
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 MEPOL instead of MaxEnt | Task-agnostic exploration · MountainCar | state entropy index | Helped |
| Use MEPOL instead of MaxEnt | Task-agnostic exploration · Ant, Humanoid, and HandReach | state entropy index | Helped |
| Use MEPOL instead of a random policy | Task-agnostic exploration · 200D GridWorld | state entropy index | Helped |
| Initialize TRPO with a MEPOL exploration policy | TRPO · Sparse-reward GridWorld, Ant, and Humanoid tasks | average return | Helped |
| Train the exploration policy with a shorter horizon, such as 200 steps (ablation) | MEPOL · GridWorld | average state entropy over the testing horizon | Hurt |
| Match the training horizon to the downstream-task horizon (ablation) | MEPOL exploration policy · GridWorld | state entropy | Helped |
| Train with a shorter horizon (ablation) | MEPOL exploration policy · Humanoid | state entropy at longer testing horizons | No clear effect |
| Vary the k-NN estimator parameter k (ablation) | MEPOL · Task-agnostic exploration domains | state entropy index | No clear effect |
| Include velocity features in the entropy computation (ablation) | MEPOL · Humanoid and HandReach | state entropy index | Hurt |
| Include external-force features in the entropy computation (ablation) | MEPOL · Humanoid and HandReach | state entropy index | No clear effect |
| Remove the preliminary steps that train the policy to output a zero mean (ablation) | MEPOL · Task-agnostic exploration | algorithm behavior | No clear effect |
| Use SMM instead of a random baseline | Task-agnostic exploration · Continuous domains | state entropy index | No clear effect |
Closest papers by meaning
- Fast Rates for Maximum Entropy Exploration (ICML 2023 · similarity 0.92)
- Behavioral Entropy-Guided Dataset Generation for Offline Reinforcement Learning (ICLR 2025 · similarity 0.90)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.89)
- Towards Principled Unsupervised Multi-Agent Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Global Optimality for Constrained Exploration via Penalty Regularization (NeurIPS 2026 · similarity 0.89)
- DIME: Diffusion-Based Maximum Entropy Reinforcement Learning (ICML 2025 · 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.