Dreamer-CDP: Improving Reconstruction-free World Models Via Continuous Deterministic Representation Prediction

2026-03-07

AI Summary

This work addresses the limitations of existing world models in high-dimensional observation spaces, where reconstruction-based approaches are often distracted by task-irrelevant details, while current reconstruction-free methods struggle to achieve strong performance in complex environments such as Crafter. The authors propose a novel reconstruction-free world model that, for the first time, integrates a JEPA-style prediction mechanism into continuous deterministic representation learning. By directly predicting future latent representations rather than reconstructing observations, the model achieves efficient environment modeling while maintaining compact and task-relevant representations. This approach significantly enhances planning performance and attains results on the Crafter benchmark comparable to those of Dreamer, demonstrating that reconstruction-free world models can be both effective and competitive in complex reinforcement learning tasks.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Model-based reinforcement learning (MBRL) agents operating in high-dimensional observation spaces, such as Dreamer, rely on learning abstract representations for effective planning and control. Existing approaches typically employ reconstruction-based objectives in the observation space, which can render representations sensitive to task-irrelevant details. Recent alternatives trade reconstruction for auxiliary action prediction heads or view augmentation strategies, but perform worse in the Crafter environment than reconstruction-based methods. We close this gap between Dreamer and reconstruction-free models by introducing a JEPA-style predictor defined on continuous, deterministic representations. Our method matches Dreamer's performance on Crafter, demonstrating effective world model learning on this benchmark without reconstruction objectives.

Problem

• reconstruction-free
• world models
• representation learning
• model-based reinforcement learning
• high-dimensional observations

Innovation

• reconstruction-free
• world models
• continuous deterministic representation
• JEPA-style predictor
• model-based reinforcement learning