Research map / Reinforcement learning / Visual and representation RL
Reinforcement Learning with Latent Flow
Wenling Shang, Xiaofei Wang, Aravind Srinivas, Aravind Rajeswaran, Yang Gao, Pieter Abbeel, Michael Laskin
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 |
|---|---|---|---|
| Replace RAD frame stacking with Flare latent flow | RAD-SAC · DMControl Quadruped Walk 500K | return: 206 -> 296 | Improved |
| Replace RAD frame stacking with Flare latent flow | RAD-SAC · DMControl Pendulum Swingup 500K | return: 79 -> 242 | Improved |
| Replace RAD frame stacking with Flare latent flow | RAD-SAC · DMControl Finger Turn hard 1M | return: 249 -> 661 | Improved |
| Replace RAD frame stacking with Flare latent flow | RAD-SAC · DMControl Walker Run 1M | return: 628 -> 556 | Hurt |
| Replace RAD frame stacking with Flare latent flow | RAD-SAC · DMControl Hopper Hop 1M | return: 211 -> 217 | No clear effect |
| Add Flare latent flow to Rainbow DQN | Rainbow DQN · Atari Phoenix | score: 16992 -> 60974 | Improved |
| Add Flare latent flow to Rainbow DQN | Rainbow DQN · Atari Defender | score: 44694 -> 86982 | Improved |
| Add Flare latent flow to Rainbow DQN | Rainbow DQN · Atari Seaquest | score: 24090 -> 13901 | Hurt |
| Add Flare latent flow to Rainbow DQN | Rainbow DQN · Atari Tutankham | score: 247 -> 248 | No clear effect |
| Add Flare latent flow to Rainbow DQN | Rainbow DQN · Atari Montezuma | score: 900 -> 1668 | Improved |
| Average over DMControl tasks, Flare vs RAD | Flare · DMControl 500K and 1M | return ratio: 1.9x and 1.5x | Improved |
| Use pixel flow (pixel differences) instead of latent flow (ablation) | Flare · DMControl Quadruped Walk | return | Hurt |
Closest papers by meaning
- Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Model (NeurIPS 2020 · similarity 0.88)
- Latent Space Policies for Hierarchical Reinforcement Learning (ICML 2018 · similarity 0.87)
- Semantic Temporal Abstraction via Vision-Language Model Guidance for Efficient Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Towards Principled Representation Learning from Videos for Reinforcement Learning (ICLR 2024 · similarity 0.87)
- Improving Computational Efficiency in Visual Reinforcement Learning via Stored Embeddings (NeurIPS 2021 · similarity 0.86)
- Task-Agnostic Dynamics Priors for Deep Reinforcement Learning (ICML 2019 · 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.
