About

Julian Togelius is an associate professor of computer science and engineering at NYU Tandon School of Engineering and the Director of the Game Innovation Lab. His research focuses on artificial intelligence and games, aiming to develop methods that make games more engaging, easier to design and develop, more adaptive, and capable of interactive experiences that are not yet possible. Togelius's interests encompass a wide range of game types, including video games, board games, card games, and mind games. His methodological roots are in evolutionary computation and neural networks, and he actively engages with various AI techniques such as player modeling, procedural content generation, automatic game design, believable bot behavior, coevolution, neuroevolution, cybersecurity, emerging media, genetic programming, and Monte Carlo tree search. He holds a Ph.D. in computer science from the University of Essex, obtained in 2007, a Master of Science in Evolutionary and Adaptive Systems from the University of Sussex in 2003, and a Bachelor of Arts in Philosophy from Lund University in 2002. Togelius has contributed to the understanding of AI's capabilities and limitations in gaming, notably highlighting the brittleness of current AI systems in adapting to new, unseen games and environments. His work includes developing AI systems capable of generating engaging word puzzles, such as the Connections game, demonstrating AI's potential in creative domains. Togelius emphasizes the importance of games as a central evaluation tool for AI, advocating for their role in testing adaptability and creativity, which are essential components of artificial general intelligence.

Research topics

• Computer Science
• Sociology
• Political Science
• Artificial Intelligence
• Epistemology
• Multimedia
• Algorithm
• Engineering
• Human–computer interaction
• Biology
• Ecology
• Engineering drawing
• Psychology

Selected publications

• Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes

  ArXiv.org · 2026-04-22

  articleOpen access

  We introduce Dream-Cubed, a large-scale dataset of Minecraft worlds at voxel resolution, and a family of models using cubes as powerful compositional units for efficient generation of interactive 3D environments. Dream-Cubed comprises tens of billions of tokens from a carefully curated mixture of procedural biome terrain and high-quality human-authored maps. We use this dataset to conduct the first large-scale study of 3D diffusion models for voxel generation, analyzing discrete and continuous diffusion formulations, data compositions, and architectural design choices. Our models operate directly in the space of blocks, enabling efficient and semantically grounded generation while supporting interactive user workflows such as inpainting and outpainting from user-authored blocks. To quantitatively evaluate our models, we adapt the FID metric to assess semantic differences between real and generated world renderings, and validate generation quality through a human preference study. We release the full dataset, code, and all our pretrained models, which we hope will provide a foundation for future research in efficient generative modeling for structured, interactive 3D environments.

  PublisherOA PDF

• Agentic PCG: Procedural Content Generation via Tool-using LLMs

  Zenodo (CERN European Organization for Nuclear Research) · 2026-03-31

  preprintOpen accessSenior author
  PublisherDOI

• Dream-Cubed: Controllable Generative Modeling in Minecraft by Training on Billions of Cubes

  arXiv (Cornell University) · 2026-04-22

  preprintOpen access

  We introduce Dream-Cubed, a large-scale dataset of Minecraft worlds at voxel resolution, and a family of models using cubes as powerful compositional units for efficient generation of interactive 3D environments. Dream-Cubed comprises tens of billions of tokens from a carefully curated mixture of procedural biome terrain and high-quality human-authored maps. We use this dataset to conduct the first large-scale study of 3D diffusion models for voxel generation, analyzing discrete and continuous diffusion formulations, data compositions, and architectural design choices. Our models operate directly in the space of blocks, enabling efficient and semantically grounded generation while supporting interactive user workflows such as inpainting and outpainting from user-authored blocks. To quantitatively evaluate our models, we adapt the FID metric to assess semantic differences between real and generated world renderings, and validate generation quality through a human preference study. We release the full dataset, code, and all our pretrained models, which we hope will provide a foundation for future research in efficient generative modeling for structured, interactive 3D environments.

  PublisherDOI

• Agentic PCG: Procedural Content Generation via Tool-using LLMs

  Zenodo (CERN European Organization for Nuclear Research) · 2026-05-10

  preprintOpen accessSenior author
  PublisherDOI

• Agentic PCG: Procedural Content Generation via Tool-using LLMs

  Zenodo (CERN European Organization for Nuclear Research) · 2026-05-10

  preprintOpen accessSenior author
  PublisherDOI

• The Garden of Forking Paths: Narrative Arc-Conditioned Gameplay Planning

  2026-04-13

  articleOpen access

  Narrative archetypes (e.g., Hero’s Journey, Three-act structure) provide universal story structures that resonate across cultures and media and are important for video game storytelling, yet existing LLM-based methods lack explicit use of these archetypes in procedurally generated games. We propose Forking Garden, a framework for narrative arc-conditioned gameplay planning that generates branching games from user-provided storylines. Our approach first generates a diverse pool of independent nodes, then assembles them into a dungeon graph via arc-guided constraint algorithms, where each node achieves multimodal alignment of gameplay elements. We develop an end-to-end interactive system that instantiates the framework.

  PublisherDOI

• PCGRLLM: Large Language Model-Driven Reward Design for Procedural Content Generation Reinforcement Learning

  IEEE Transactions on Games · 2026-01-01

  preprintOpen access

  Reward design plays a pivotal role in the training of game AIs, requiring substantial domain-specific knowledge and human effort. In recent years, several studies have explored reward generation for training game agents and controlling robots using large language models (LLMs). In the content generation literature, there has been early work on generating reward functions for reinforcement learning agent generators. This work introduces <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">PCGRLLM</i>, an extended architecture based on earlier work, which employs a feedback mechanism and several reasoning-based prompt engineering techniques. We evaluate the proposed method on a story-to-reward generation task in a two-dimensional environment using two state-of-the-art LLMs across various reasoning-based prompting methods. Our experiments provide insightful evaluations that demonstrate the capabilities of LLMs essential for content generation tasks. The results demonstrate a substantial performance improvement over the previous structure, achieving performance comparable to that of humans. Our work demonstrates the potential to reduce human dependency in game AI development, while supporting and enhancing creative processes.

  PublisherOA PDFDOI

• Video Game Level Design as a Multi-Agent Reinforcement Learning Problem

  Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment · 2025-11-07

  articleOpen accessSenior author

  Procedural Content Generation via Reinforcement Learning (PCGRL) offers a method for training controllable level designer agents without the need for human datasets, using metrics that serve as proxies for level quality as rewards. Existing PCGRL research focuses on single generator agents, but are bottlenecked by the need to frequently recalculate heuristics of level quality and the agent’s need to navigate around potentially large maps. By framing level generation as a multi-agent problem, we mitigate the efficiency bottleneck of single-agent PCGRL by reducing the number of reward calculations relative to the number of agent actions. We also find that multi-agent level generators are better able to generalize to out-of-distribution map shapes, which we argue is due to the generators’ learning more local, modular design policies. We conclude that treating content generation as a distributed, multi-agent task is beneficial for generating functional artifacts at scale.

  PublisherOA PDFDOI

• Evolutionary Level Repair

  Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2025-07-14

  articleOpen access

  We address the problem of game level repair, which consists of taking a designed but non-functional game level and making it functional. This might consist of ensuring the completeness of the level, reachability of objects, or other performance characteristics. The repair problem may also be constrained in that it can only make a small number of changes to the level. We investigate search-based solutions to the level repair problem, particularly using evolutionary and quality-diversity algorithms, with good results. This level repair method is applied to levels generated using a machine learning-based procedural content generation (PCGML) method that generates stylistically appropriate but frequently broken levels. This combination of PCGML for generation and search-based methods for repair shows great promise as a hybrid procedural content generation (PCG) method.

  PublisherOA PDFDOI

• Moonshine: Distilling Game Content Generators into Steerable Generative Models

  Proceedings of the AAAI Conference on Artificial Intelligence · 2025-04-11

  articleOpen accessSenior author

  Procedural Content Generation via Machine Learning (PCGML) has enhanced game content creation, yet challenges in controllability and limited training data persist. This study addresses these issues by distilling a constructive PCG algorithm into a controllable PCGML model. We first generate a large amount of content with a constructive algorithm and label it using a Large Language Model (LLM). We use these synthetic labels to condition two PCGML models for content-specific generation, a diffusion model and the five-dollar model. This neural network distillation process ensures that the generation aligns with the original algorithm while introducing controllability through plain text. We define this text-conditioned PCGML as a Text-to-game-Map (T2M) task, offering an alternative to prevalent text-to-image multi-modal tasks. We compare our distilled models with the baseline constructive algorithm. Our analysis of the variety, accuracy, and quality of our generation demonstrates the efficacy of distilling constructive methods into controllable text-conditioned PCGML models.

  PublisherOA PDFDOI

Recent grants

• RI: Small: General Intelligence through Algorithm Invention and Selection

  NSF · $427k · 2017–2020

• SaTC: CORE: Small: Dictionary Attacks on Biometrics

  NSF · $483k · 2020–2024

Frequent coauthors

• Ahmed Khalifa

  101 shared

• Georgios N. Yannakakis

  University of Malta

  91 shared

• Michael Cerny Green

  New York University

  65 shared

• Andy Nealen

  University of Southern California

  50 shared

• Antonios Liapis

  University of Malta

  45 shared

• Sebastian Risi

  IT University of Copenhagen

  41 shared

• Simon M. Lucas

  Queen Mary University of London

  39 shared

• Sam Earle

  36 shared

Labs

• Center for Urban Science and Progress (CUSP)PI