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

Design and Optimization of Reinforcement Learning-Based Agents in Text-Based Games

arXiv 25.9 2025 40.3 method, application

TLDR

Proposes deep reinforcement learning agent with world model for text-based games, achieving improved game completion and win rates.

Reasoning

The paper presents a clear combination of deep learning for world modeling and policy gradient for agent learning in text-based games, with empirical results showing improvement. However, the abstract lacks details on the specific architecture, comparison baselines, and generalizability beyond text-based games.

Read-first score

Read-first score 40.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 22.

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=experiment

Topical relevance 42%
31.4

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: 304.

Keyword Scores

world model
10
model-based reinforcement learning world model
5
world dynamics prediction
3
world simulator
2
generative world model
1
interactive world model
1
video world model
0

Deep Analysis

Innovations

  • Novel approach combining deep learning-based world model with policy gradient reinforcement learning for text-based games
  • Application of deep learning to process game text and build a world model
  • Policy gradient-based deep reinforcement learning method to convert state value to optimal policy

Methodology

A deep learning model is first applied to process game text and build a world model. Then, the agent is learned through a policy gradient-based deep reinforcement learning method to facilitate conversion from state value to optimal policy.

Key Results

The enhanced agent works better in several text-based game experiments and significantly surpasses previous agents on game completion ratio and win rate.

Tags