Research map / Reinforcement learning / Visual and representation RL
Learning to Perceive the World Through Control: Empowerment-Based Representation Learning
Mahsa Bastankhah, Sophie Broderick, Benjamin Eysenbach
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 |
|---|---|---|---|
| enforce symmetric forward and backward representations (identical backward representations for b and c) (ablation) | MISL empowerment · Tabular MDP, state A | empowerment: 0.69 -> 0 | Hurt |
| use frozen MISL representations instead of raw noisy state for reward maximization | SAC/PPO agent · Point-Maze, Ant, Kitchen with noise | downstream return | Improved |
| scale policy, discriminator and skill dimension (policy 5-layer width 1536, discriminator 8-layer width 2048, skill dim 2 to 16) (ablation) | MISL (METRA) · Point-Maze with 5 noise dims | representation variance along noise: skill dimension 2 -> 16 | No clear effect |
| test on unseen temporally and cross-dimensionally correlated noise (trained on i.i.d. noise) | MISL representations · Point-Maze | noise invariance | No clear effect |
| add linear probe on MISL representations to recover (x,y) | MISL representations · Point-Maze with 5 noise dims | R^2: R^2 approaches 1 | Improved |
Closest papers by meaning
- Information Prioritization through Empowerment in Visual Model-based RL (ICLR 2022 · similarity 0.89)
- Towards Empowerment Gain through Causal Structure Learning in Model-Based Reinforcement Learning (ICLR 2025 · similarity 0.89)
- Learning to Assist Humans without Inferring Rewards (NeurIPS 2024 · similarity 0.88)
- What Is Worth Representing? Representational Empowerment for Continual Model Construction (NeurIPS 2026 · similarity 0.88)
- Towards Generalizable Reinforcement Learning via Causality-Guided Self-Adaptive Representations (ICLR 2025 · similarity 0.85)
- Pre-trained Text-to-Image Diffusion Models Are Versatile Representation Learners for Control (NeurIPS 2024 · 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.
