Research map / Reinforcement learning / Visual and representation RL
Imagination-Augmented Agents for Deep Reinforcement Learning
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 |
|---|---|---|---|
| augment a model-free policy with learned encodings of imagined environment-model rollouts | I2A · Sokoban | levels solved: 85% vs. a maximum of under 60% | Helped |
| increase rollout length from 1 to 3 steps (ablation) | I2A · Sokoban | levels solved | Helped |
| increase rollout length from 3 to 5 steps (ablation) | I2A · Sokoban | levels solved | Helped |
| increase rollout length from 5 to 15 steps (ablation) | I2A · Sokoban | levels solved: above 90% | Helped |
| use an inaccurate, smaller environment model instead of the accurate model (ablation) | I2A · Sokoban | levels solved | No clear effect |
| replace the learned rollout encoder with fixed Monte-Carlo return estimates (ablation) | I2A · Sokoban | levels solved | Hurt |
| replace the environment model with a copy model (ablation) | I2A · Sokoban | levels solved | Hurt |
| remove reward prediction while retaining imagined observation predictions (ablation) | I2A · Sokoban | levels solved: close to 80% after 3e9 steps | Hurt |
| add outer-loop Monte-Carlo search using the trained I2A for rollout policy (ablation) | I2A · Sokoban | levels solved: 95% (up from 87%) | Helped |
Closest papers by meaning
- Open-World Reinforcement Learning over Long Short-Term Imagination (ICLR 2025 · similarity 0.86)
- RIG: Synergizing Reasoning and Imagination in End-to-End Generalist Policy (ICLR 2026 · similarity 0.85)
- Acting Beyond Learning: Imagination-Assisted Decision-Making in the Visual-based Multi-Agent Cooperative Scenarios (AAAI 2025 · similarity 0.85)
- Imagine, Initialize, and Explore: An Effective Exploration Method in Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.85)
- Dynamics-Aligned Latent Imagination in Contextual World Models for Zero-Shot Generalization (NeurIPS 2025 · similarity 0.84)
- What if Agents Could Imagine? Reinforcing Open-Vocabulary HOI Comprehension through Generation (NeurIPS 2026 · similarity 0.84)
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.