Research map / Reinforcement learning / Visual and representation RL
VIME: Variational Information Maximizing Exploration
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 |
|---|---|---|---|
| add VIME intrinsic reward to TRPO on sparse-reward tasks | TRPO · MountainCar, CartPoleSwingup, HalfCheetah (sparse) | average return | Helped |
| use l2 BNN prediction error as intrinsic reward instead of VIME (ablation) | TRPO · sparse-reward tasks | average return | Hurt |
| naive Gaussian control noise exploration | TRPO · sparse-reward tasks | average return | Hurt |
| add VIME to TRPO | TRPO · SwimmerGather | average return | Helped |
| add VIME to TRPO, ERWR, REINFORCE on shaped-reward tasks | TRPO, ERWR, REINFORCE · CartPole, CartPoleSwingup, DoublePendulum, MountainCar | average return | Helped |
| add VIME to REINFORCE and ERWR | REINFORCE, ERWR · DoublePendulum, MountainCar | convergence to optimum | Helped |
| add VIME to policy gradient methods | TRPO · CartPole | average return | No clear effect |
| add VIME to TRPO | TRPO · Walker2D | average return | Helped |
| add Gaussian noise to the rewards as a baseline (ablation) | RL algorithms · continuous control tasks | average return | No clear effect |
| set eta too high (ablation) | VIME · MountainCar | average return | Hurt |
| set eta too low (ablation) | VIME · MountainCar | average return | Hurt |
| vary eta in a wide middle range (ablation) | VIME · MountainCar | average return | No clear effect |
Closest papers by meaning
- MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization (ICLR 2025 · similarity 0.88)
- Bayesian Exploration Networks (ICML 2024 · similarity 0.87)
- Parameter Exploration for RLVR via Variational Learning (NeurIPS 2026 · similarity 0.87)
- Effective Exploration Based on the Structural Information Principles (NeurIPS 2024 · similarity 0.86)
- Exploration from a Primal-Dual Lens: Value-Incentivized Actor-Critic Methods for Sample-Efficient Online RL (NeurIPS 2025 · similarity 0.86)
- Random Latent Exploration for Deep Reinforcement Learning (ICML 2024 · similarity 0.86)
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.