Research map / Multi-agent systems and games / Human-AI and goal-conditioned
A Consciousness-Inspired Planning Agent for Model-Based Reinforcement Learning
Harry Zhao, Zhen Liu, Sitao Luan, Shuyuan Zhang, Doina Precup, Yoshua Bengio
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 |
|---|---|---|---|
| add consciousness-inspired bottleneck (CP(8)) to set-based MPC planner (ablation) | UP (set-based MBRL agent) · OOD MiniGrid lava navigation 8x8 | success rate | Improved |
| use MPC planning instead of Dyna with true model | Dyna* baseline · OOD MiniGrid lava navigation 8x8 | success rate | No clear effect |
| two-stage world-model pretraining instead of end-to-end training (ablation) | CP · in-distribution MiniGrid | success rate | Hurt |
| add action regularization (inverse-model loss) in bottleneck (ablation) | CP(4) · OOD MiniGrid difficulty 0.35 | success rate | Improved |
| semi-hard top-k attention instead of soft attention for selection (ablation) | CP · in-distribution dynamics accuracy | relative L1 error | Mixed |
| vectorized representation instead of set representation (NOSET) | UP · in-distribution MiniGrid | success rate | Hurt |
| add noise input to dynamics model (CP(8)+) (ablation) | CP(8) · color-distraction Turn-and-Forward OOD | success rate | No clear effect |
Closest papers by meaning
- On the role of planning in model-based deep reinforcement learning (ICLR 2021 · similarity 0.88)
- Consciousness-Inspired Spatio-Temporal Abstractions for Better Generalization in Reinforcement Learning (ICLR 2024 · similarity 0.86)
- Interpreting Emergent Planning in Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.85)
- Representation Learning Enables Scalable Multitask Deep Reinforcement Learning (NeurIPS 2026 · similarity 0.85)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.84)
- Reward-Respecting Subtasks for Model-Based Reinforcement Learning (Abstract Reprint) (AAAI 2024 · 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.
