Research map / Reinforcement learning / Visual and representation RL
Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Model
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 |
|---|---|---|---|
| Learn a sequential stochastic latent representation and train an off-policy actor-critic in latent space | Image-based continuous control · DeepMind Control Suite | sample efficiency and final performance | Helped |
| Learn a sequential stochastic latent representation and train an off-policy actor-critic in latent space | Image-based continuous control · OpenAI Gym | final performance | Helped |
| Use SLAC instead of PlaNet | Image-based continuous control · DeepMind Control Suite | final performance | Helped |
| Use a fully stochastic model instead of a partially stochastic model (ablation) | SLAC · Image-based continuous control | performance | Helped |
| Add temporal dependencies to the latent model (ablation) | SLAC · Image-based continuous control | performance | Helped |
| Use autoregressive two-variable latent factorization (ablation) | SLAC · Image-based continuous control | performance | Helped |
| Use a fully stochastic model instead of a deterministic model (ablation) | SLAC · Image-based continuous control | performance | Helped |
| Give the critic latent samples instead of observation-action history (ablation) | SLAC · Image-based continuous control | performance | Helped |
| Change the actor input between observation-action history and latent samples (ablation) | SLAC · Image-based continuous control | performance | No clear effect |
| Pretrain the latent model before task learning (ablation) | SLAC · Image-based continuous control | learning speed | Helped |
| Pretrain the latent model before task learning (ablation) | SLAC · Image-based continuous control | asymptotic performance | Helped |
| Increase model pretraining from 50000 to 100000 iterations (ablation) | SLAC · DeepMind Control Suite | performance: 50000 ... 100000 | No clear effect |
Closest papers by meaning
- Mask-based Latent Reconstruction for Reinforcement Learning (NeurIPS 2022 · similarity 0.89)
- Representation Learning Enables Scalable Multitask Deep Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Safe Reinforcement Learning From Pixels Using a Stochastic Latent Representation (ICLR 2023 · similarity 0.88)
- Latent State Marginalization as a Low-cost Approach for Improving Exploration (ICLR 2023 · similarity 0.88)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.87)
- Probabilistic Subgoal Representations for Hierarchical Reinforcement Learning (ICML 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.