Skip to content
Open access

Map Generation for Roguelike Games from Textual Descriptions Using Large Language Models

Sep 2026 · Brazilian Symposium on Games and Digital Entertainment · 0 citations · 25 references

TL;DR

This work proposes an automated system that generates playable and structurally coherent roguelike maps from textual descriptions from textual descriptions using an LLM-driven Assets Generator and a deterministic PCG algorithm for spatial layout generation.

Abstract

Introduction: Procedural Content Generation (PCG) enhances replayability in roguelike games but remains difficult to control via algorithmic parameters. Large Language Models (LLMs) introduce a “text-to-map” paradigm but still struggle with spatial reasoning. Objective: We propose an automated system that generates playable and structurally coherent roguelike maps from textual descriptions. Methodology: The architecture combines an LLM-driven Assets Generator for semantic content with a deterministic PCG algorithm for spatial layout generation and uses Retrieval-Augmented Generation (RAG) to map textual intent into visual assets, based on a dataset of 489 pixel-art images. Results: Evaluation indicates high semantic fidelity, with coherence scores of up to 93.6%, yielding visually cohesive and playable maps.

Read PDF

Similar papers

Open access Sep 2026

Towards Player-Oriented Procedural Narratives with Stochastic Grammar-Generated Quests

This work investigates whether a player-adaptive quest generation method can increase content variability and align generated quests with distinct player behavior profiles, including partial evidence of alignment between generated content and player profiles, although this effect depends on parameter tuning.

Luiza Torello Vieira, L. T. Pereira, C. Toledo · 0 citations
Preprint Oct 2026

Form and Void: Entangled Composition through an Autonomous AI Agent

Positive and negative space is a fundamental principle in visual composition, supporting visually coherent forms and layered semantic relationships. Generating such compositions is challenging because it requires coordinated control over two semantic concepts that share a common boundary. Although recent text-to-image...

Shi-Wen Wang, Jian Yang, Xu Wang et al. · 0 citations
Sep 2026

Canvas Beyond Words: Teaching Machines to Sketch Empowers Their Spatial Intelligence.

SIG is introduced, a structured grid-based schema that explicitly represents object layouts, inter-object relations, and physically grounded priors that demonstrates the generalizability of SIG across diverse scenarios and multiple MLLMs, and systematically examines how data sample selection influences SIG-based ICL.

Guan-Lin Wu, Bo-Yan Su, Yang Zhao et al. · 0 citations
Open access Nov 2026

The explain-on-blackboard paradigm for interleaved visual-text reasoning

A pedagogical distillation pipeline building on an Explain-on-Blackboard (EoB) formulation, inspired by dual-coding and multimedia-learning theories, that generates temporally and semantically aligned teaching trajectories that comprise a discourse channel for logical explanation and a canvas channel for incremental HT...

Lin-Zhuang Sun, Si-Zhe Qiu, Jian-Wei Cao et al. · 0 citations
Sep 2026

Reasoning with Image Generation

Across six diverse visual reasoning tasks including multi-view spatial reasoning and collision prediction, ReImaGin consistently outperforms both text-only reasoning and specialist vision-tool baselines, with gains of up to 25\%, demonstrating the advantage of flexible, generative visual reasoning.

Nishad Singhi, Hector P. Rodriguez, Aditya Arora et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.