Research map / Reinforcement learning / Visual and representation RL
Redeeming intrinsic rewards via constrained optimization
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 EIPO extrinsic-optimality constraint to RND | RND with PPO · 61 ATARI games | P(method > EO): 0.49 -> 0.65 | Helped |
| use KL-regularized Decoupled-RND (similarity assumption without constraint) (ablation) | Ext-norm-RND · 61 ATARI games vs EO | P(method > EO) | No clear effect |
| learn alpha dynamically (EIPO) instead of fixed linear decay of intrinsic reward (ablation) | EIPO-RND · 61 ATARI games | P(EIPO-RND > Decay-RND): 0.59 | Helped |
| use EIPO instead of Ext-norm-RND on hard exploration games | RND · ATARI games where RND beats EO | PPO-normalized score mean: 427.08 -> 435.56 | No clear effect |
| tune per-game intrinsic coefficient lambda for RND | RND · selected ATARI games | return | Mixed |
| normalize extrinsic rewards in RND (Ext-norm-RND) (ablation) | RND · 61 ATARI games vs EO | P(method > EO) | Helped |
Closest papers by meaning
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.89)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.88)
- MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization (ICLR 2025 · similarity 0.88)
- Rethinking Exploration in Reinforcement Learning with Effective Metric-Based Exploration Bonus (NeurIPS 2024 · similarity 0.88)
- Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 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.