Research map / Reinforcement learning / Visual and representation RL
STORM: Efficient Stochastic Transformer based World Models for Reinforcement Learning
Weipu Zhang, Gang Wang, Jian Sun, Yetian Yuan, Gao Huang
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 |
|---|---|---|---|
| apply reconstruction loss to encoder output instead of sequence model output (decoder at rear) (ablation) | STORM world model · Atari 100k | score | Hurt |
| feed reward/continuation predictors with h_t instead of z_t (ablation) | STORM world model · Ms Pacman | score | Improved |
| feed reward/continuation predictors with h_t instead of z_t (ablation) | STORM world model · Pong | score | No clear effect |
| increase Transformer layers beyond 2 (ablation) | STORM world model · Atari 100k | score | No clear effect |
| include h_t in agent state alongside z_t (ablation) | STORM agent · Ms Pacman | score | Improved |
| include h_t in agent state alongside z_t (ablation) | STORM agent · Pong / Kung Fu Master | score | No clear effect |
| use only h_t as agent state (ablation) | STORM agent · Pong | score | Hurt |
| add one demonstration trajectory to replay buffer (ablation) | STORM · Freeway | score: 0 -> 34 | Improved |
| add one demonstration trajectory to replay buffer (ablation) | STORM · Pong | score | Improved |
| add one demonstration trajectory to replay buffer (ablation) | STORM · Ms Pacman | score | Hurt |
| include Freeway demonstration in overall results (ablation) | STORM · Atari 100k | mean human normalized score (%): 122.3% -> 126.7% | Improved |
Later papers that compared with this method
2 later accepted papers report STORM as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Drama: Mamba-Enabled Model-Based Reinforcement Learning Is Sample and Parameter Efficient (ICLR 2025 · better on 0 of 2 shared results)
- Efficient World Models with Context-Aware Tokenization (ICML 2024 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining (ICLR 2025 · similarity 0.88)
- EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling (NeurIPS 2025 · similarity 0.87)
- Diffusion for World Modeling: Visual Details Matter in Atari (NeurIPS 2024 · similarity 0.87)
- Learning Transformer-based World Models with Contrastive Predictive Coding (ICLR 2025 · similarity 0.87)
- Efficient World Models with Context-Aware Tokenization (ICML 2024 · similarity 0.87)
- PWM: Policy Learning with Multi-Task World Models (ICLR 2025 · 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.
