Research map / Probabilistic and scientific modeling / Molecules and proteins
MIMEx: Intrinsic Rewards from Masked Input Modeling
Toru Lin, Allan Jabri
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 MIMEx intrinsic reward to PPO base agent | MVP with PPO · PixMC-Sparse (8 tasks) | task success rate | Improved |
| add MIMEx intrinsic reward to DrQv2 | DrQv2 with DDPG · DeepMind Control (6 tasks) | average return | Mixed |
| MIMEx versus ICM and RND baselines on harder Pick and Move tasks | MIMEx · PixMC-Sparse Pick and Move | task success rate | Improved |
| increase exploration sequence length beyond 2 (ablation) | MIMEx · KukaReach and KukaPick | task success rate | Improved |
| average prediction error over 5 random maskings (nm5) instead of once (ablation) | MIMEx · KukaReach and KukaPick | task success rate | Improved |
| double decoder depth, heads and embedding size (ablation) | MIMEx · KukaPick | task success rate | Improved |
| vary mask ratio among 20%, 50%, 70% (ablation) | MIMEx · KukaReach and KukaPick | task success rate | Mixed |
| mask on feature dimension with 80-95% ratio instead of time-only 70% (ablation) | MIMEx · KukaReach and KukaPick | task success rate | No clear effect |
| remove dense reward term from base agent with random action noise | base RL agent · KukaPick | success rate | Hurt |
| add MIMEx to sparse-reward base agent | base RL agent · KukaPick sparse | success rate | Improved |
Closest papers by meaning
- CIG: Exploration via Conditional Information Gain (NeurIPS 2026 · similarity 0.87)
- MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization (ICLR 2025 · similarity 0.85)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.85)
- Portable Reward Tuning: Towards Reusable Fine-Tuning across Different Pretrained Models (ICML 2025 · similarity 0.85)
- Improving Intrinsic Exploration by Creating Stationary Objectives (ICLR 2024 · similarity 0.85)
- Can In-Context Learning Support Intrinsic Curiosity? (NeurIPS 2026 · 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.
