⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

From Curiosity to Competence: How World Models Interact with the Dynamics of Exploration

arXiv 25.7 2025 45.7 method

TLDR

This paper investigates how world models balance curiosity and competence in exploration, comparing Tabular and Dreamer agents.

Reasoning

Strengths: bridges cognitive theories with RL, formalizes exploration trade-off. Weaknesses: limited to handcrafted abstractions and Dreamer; may not generalize to complex environments.

Read-first score

Read-first score 45.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 31.

Recency 8%
86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%
50

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=analysis

Topical relevance 42%
44.3

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%
30

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 404.

Keyword Scores

world model
9
model-based reinforcement learning world model
8
world dynamics prediction
6
generative world model
4
world simulator
2
interactive world model
1
video world model
1

Deep Analysis

Innovations

  • Bridging cognitive theories of intrinsic motivation with reinforcement learning to study the trade-off between curiosity (novelty/information gain) and competence (empowerment)
  • Comparing two model-based agents (Tabular with handcrafted state abstractions and Dreamer with learned world model) to reveal distinct exploration patterns and interactions
  • Formalizing adaptive exploration as a balance between pursuing the unknown and the controllable, offering insights for cognitive theories and efficient RL

Methodology

The study compares two model-based agents: a Tabular agent using handcrafted state abstractions and a Dreamer agent that learns an internal world model. Both agents are evaluated in an environment where they must balance curiosity (novelty or information gain) and competence (empowerment), with analysis of exploration patterns and representation learning dynamics.

Key Results

The Tabular agent shows that curiosity and competence guide exploration in distinct patterns, and prioritizing both improves exploration. The Dreamer agent reveals a two-way interaction between exploration and representation learning, mirroring the developmental co-evolution of curiosity and competence.

Resources

Paper

Tags