Search arXiv⌕ Search

arXiv subjects

Julian Togelius

Publications and source records attributed to Julian Togelius.

At least 19 recordsLinked to original sources

Language as Trait, Politeness as Strategy: Controlled Tests of Linguistic Effects in LLM Social-Deduction Agents

Large language model (LLM) agents coordinate through conversation, and the same message can be phrased in simpler or more complex language, bluntly or politely. We test whether such differences change how other agents judge and act, using the LURK Werewolf framework, which separates each utterance's speech-act predicate from its natural-language realization. We rewrite only the realization, check that the game content is preserved, and replay each game day from saved checkpoints against a re-sampled control. We vary linguistic range consistently across a speaker's messages and politeness strategy per utterance. Across 30 games and 4,335 replays, neither manipulation changes the trust other agents place in the speaker, and no effect on votes or night actions survives correction for multiple comparisons; the largest uncorrected effect is that shifting range away from a neutral paraphrase changes about 4 percentage points more of the other players' votes. In natural play, 88% of face-threatening speech is bald on-record regardless of role.

cs.CL↗

Rethinking Negative Prompting for Concept Suppression: Unconditional Anchoring and Localized Guidance

Text-to-image diffusion models have rapidly transitioned from research prototypes to widely deployed tools. Along with their impressive capabilities, these models can also generate harmful or undesirable content (e.g., NSFW imagery), motivating practical mechanisms for removing such concepts. Negative prompting, which uses the negative-prompt prediction as the guidance anchor, is among the most widely used mechanisms for steering sampling trajectories away from unwanted concepts. However, for jailbreak prompts, it can fail in a regime we refer to as \emph{guidance collapse}, where the model's denoising predictions for the prompt and negative prompt become similar, so that the guidance difference driving negative prompting vanishes and the guided prediction degenerates to the negative-prompt prediction, undermining suppression. To fix this issue, we propose TraSCE (Trajectory Steering for ConcEpt Suppression), a training-free inference-time method that makes negative prompting reliable under guidance collapse. TraSCE combines two complementary components: (i) unconditional anchoring, which makes the fallback under guidance collapse a neutral unconditional prediction rather than the negative-prompt prediction, and (ii) an alignment-localized steering step, which deliberately induces this collapse for adversarial prompts and steers the sample to the unconditional fallback, while remaining minimally intrusive on benign prompts. We evaluate TraSCE on adversarial prompt benchmarks targeting nudity and violence, and further demonstrate its applicability to suppressing artistic styles and objects. Across these settings, TraSCE substantially improves the robustness of negative prompting to jailbreak prompts without any training, weight updates, or pre-collected concept-specific data (prompts or images), while keeping the impact on general image quality small.

cs.CV↗

PuzzleJAX: A Benchmark for Reasoning and Learning

We introduce PuzzleJAX, a GPU-accelerated puzzle game engine and description language designed to support rapid benchmarking of tree search, reinforcement learning, and LLM reasoning abilities. Unlike existing GPU-accelerated learning environments that provide hard-coded implementations of fixed sets of games, PuzzleJAX allows dynamic compilation of any game expressible in its domain-specific language (DSL). This DSL follows PuzzleScript, which is a popular and accessible online game engine for designing puzzle games. In this paper, we validate in PuzzleJAX several hundred of the thousands of games designed in PuzzleScript by both professional designers and casual creators since its release in 2013, thereby demonstrating PuzzleJAX's coverage of an expansive, expressive, and human-relevant space of tasks. By analyzing the performance of search, learning, and language models on these games, we show that PuzzleJAX can naturally express tasks that are both simple and intuitive to understand, yet often deeply challenging to master, requiring a combination of control, planning, and high-level insight.

cs.AI↗

Learning Local Constraints for Reinforcement-Learned Content Generators

Constraint-based game content generators that learn local constraints from existing content, such as Wave Function Collapse (WFC), can generate visually satisfying game levels but face challenges in optimizing global properties, such as playability. On the other hand, reinforcement-learning-trained generators can optimize global properties---because such properties can easily be included in reward functions---but the results can be visually dissatisfying. In this paper, we explore ways to combine these methods. Specifically, we constrain the action space of a PCGRL generator with constraints learned by WFC, effectively allowing the PCGRL generator to achieve global properties while being forced to adhere to local constraints. To better analyze how this hybrid content generation method operates, we vary the number and type of inputs, and we test whether to randomly collapse the starting state and exclude rare patterns. While the method is sensitive to hyperparameter tuning, the best of our trained generators produce visually satisfying and playable puzzle-platform game levels---such as Lode Runner levels---with desired global properties.

cs.AI↗

State Trace Rationale As Auxiliary Task in Reinforcement Learning

We propose STRAT, an auxiliary task that trains deep reinforcement learning (RL) agents to predict a short textual trace of their own state. Inspired by human spatial navigation, the description combines landmark, route, and survey knowledge, tracking the agent's position, inventory, goals, and immediate progress. Environment rules generate this text online without human labelling. Our method adds a single auxiliary head to a standard policy. Across 60 sparse-reward XLand-MiniGrid tasks, STRAT solves complex environments where standard RL fails outright, while compacting state representations and preventing rank collapse. Beyond performance gains, the predicted trace provides a readable account of agent beliefs at every step for no extra cost.

cs.AI↗

The Garden of Forking Paths: Threading Narrative Archetype as a Semantic Signal Through Gameplay Planning

Generative models can produce individual game facets, but whole-game generation remains an orchestration problem: narrative, level structure, encounters, objectives, rewards, and visuals must express shared intent. We present Forking Garden, a branching game generation system that uses narrative archetype as a persistent semantic signal across the generation pipeline. Narrative progression is represented as soft Rise/Fall states; candidate plot nodes are generated before structural constraints assemble archetype-conforming graphs. The same state then conditions encounter composition, objectives, rewards, and runtime difficulty adaptation, while a shared symbolic schema preserves narrative entities and gameplay configurations through content instantiation. Across 10 storylines, generated paths exhibit distinguishable archetypal trajectories, narrative threat predicts damage taken by a threat-blind combat agent, and generate-first-constrain-later yields 2.6 times the entity diversity of a hierarchical baseline. A 16-participant study further suggests that propagated Rise/Fall distinctions can remain meaningful during play, while also supporting narrative understanding and creator-oriented interpretation.

cs.HC↗

Programmable Cellular Automata

Cellular automata is a local computation paradigm where complex behavior can arise from local interactions between simple functions. This paradigm has been used to explain many systems such as biological processes, traffic simulation, computer networks, etc. In games, cellular automata have been used in games such as SimCity and for the generation of spatial content such as caves or dungeons. However, creating effective local rules is hard and unintuitive. Cellular automata can be effectively evolved, but may still be hard to interpret. In this work, we introduce the concept of programmable cellular automata, where we represent the system as Python code. We also modularize the cellular automata into local functions and a decision function. Local functions take a local neighborhood and return a value, while the decision function takes the output of the local functions and decides the value of the next state. Separating the cellular automata into modules written in Python helps with understanding how these systems are working. We also explore adding global functions where they take the whole state and compute a function from it. We tested generating levels for three different games from the PCG Benchmark. The results showed that global functions decrease the number of iterations that cellular automata need to solve a problem, and that we cannot find solutions for some problems with purely local functions. Looking into the generated functions, we can see common functions that have been used in different experiments, which not only helps us understand the generator but also helps us understand these games better and what is important for them.

cs.NE↗

Procedural Content Metageneration via Program Search and Continual Abstraction Discovery

Large language models can generate executable programs, which makes it possible to search directly over procedural content generators rather than individual levels. We study this approach in Sokoban, Zelda, Dangerous Dave, and Lode Runner. Each run evolves complete Python generators through language-model mutation and crossover. We introduce Continual Abstraction Discovery, or CAD, which extracts reusable primitives from high-fitness programs into a run-specific helper module. A 2x2 experiment crosses CAD with access to a fixed hand-written domain API. The completed data set contains 160 complete runs, with at least ten 50-generation runs in every cell. CAD raises mean final best fitness in all eight domain and API comparisons. Across all CAD runs, learned libraries are adopted by most later programs and repeatedly rediscover validation, reachability, and structural utilities. These results support that discovering reusable primitives improves evolutionary program search for content generators.

cs.AI↗

The Time Value of Evolution

In evolutionary search, a weak child can be a valuable ancestor that makes high-fitness regions reachable. Immediate-return control is blind to this delayed utility, penalizing mutations through their immediate offspring even when they open productive future lineages. We formalize this hidden dynamic as the time value of evolution within a finite-horizon Markov decision process. To exploit it, we introduce Lineage-Value Policy Gradients (LVPG), a long-horizon actor-critic framework for automated trading policy discovery. Our architecture decouples search control into specialized policy heads over a shared generative backbone: a bootstrapped critic head estimates the value of finite-horizon lineage potential from multi-step mutation trees, while an actor head dynamically modulates mutation intensity over the remaining search budget. We isolate the impact of long-horizon credit assignment against immediate-return optimization across 90 paired runs under matched operators, lineage supervision, folds, seeds, and budgets. Path-based credit assignment substantially accelerates finite-budget search, increasing validation best-so-far AUC by 0.394 Sharpe units. LVPG also produces fewer temporary regressions than immediate-return optimization and recovers from them more often. Finite-horizon lineage value yields more selective non-monotonic search and stronger policies within identical resource constraints.

cs.LG↗

ELMER: Evolutionary Language Model that Explores and Refines

Program evolution can measure whether a mutation helped, but it rarely controls how far the mutation moves in behavior space. Syntactic edit size is an unreliable proxy: a small code change can alter nearly every action, while a larger rewrite can preserve the same execution trace. We introduce an Evolutionary Language Model that searches over natural-language policy descriptions and compiles typed programs for execution. A fully fine-tuned Qwen3-8B model learns three task-conditioned operations: conditional semantic mutation, natural language to domain-specific language (GPTL) compilation, and GPTL to natural language translation. The model is fine-tuned with conditional input on the mutation strength (low, medium, high) using Direct Preference Optimization (oDPO). Across 252 fixed-budget evolutionary searches, oDPO improves both behavioral calibration and finite-budget search efficiency. Natural-language attains the highest observed held-out fitness. Our analysis shows that the condition input (mutation strength) systematically changes semantic edit composition and that language mutations preserve more parent fitness at matched small-to-moderate behavioral displacement. These results show that language can serve as a steerable, execution-grounded search representation over executable program space.

cs.LG↗

Generation and Evaluation in the Human Invention Process through the Lens of Game Design

The human ability to learn rules and solve problems has been a central concern of cognitive science research since the field's earliest days. But we do not just follow rules and solve problems given to us by others: we modify those rules, create new problems, and set new goals and tasks for ourselves and others. Arguably, even more than rule following and problem solving, human intelligence is about creatively breaking and stretching the rules, changing the game, and inventing new problems worth thinking about. Creating a good rule or a good problem depends not just on the ideas one can think up but on how one evaluates such proposals. Here, we study invention through the lens of game design. We focus particularly on the early stages of novice, "everyday" game creation, where the stakes are low. We draw on a dataset of over 450 human created games, created by participants who saw an initial seed set of two-player grid-based strategy games. We consider two different cognitive mechanisms that may be at work during the early processes of intuitive game invention: an associative proposal based on previous games one has seen and compute-bounded model-based evaluation that an everyday game creator may use to refine their initial draft proposals. In our preliminary work, we conduct a model-based analysis of how people invented new games based on prior experience and find that generated games are best described by a model which incorporates model-based estimates of game quality at a population level. Our work points to how human invention is based not only on what people propose, but how they evaluate and offers a computational toolkit to scale empirical studies of model-based simulation in open-ended human innovation.

cs.HC↗

Evolutionary Wave Function Collapse

Wave Function Collapse (WFC) is a widely used procedural content generation method that learns local adjacency constraints from example inputs to generate larger outputs. In this paper, we explore combining WFC with evolutionary search by evolving the small input examples used by WFC rather than directly evolving complete levels. In this approach, WFC acts as a genotype-to-phenotype mapping. The generated levels are then evaluated through domain-specific fitness functions. We evaluate the method in two domains with different relationships between local and global structure: Maze connectivity maps and Zelda-style dungeon layouts. Our results show that evolutionary optimization over WFC inputs improves generation quality in domains where properties emerge from local relationships, while domains requiring global constraints remain challenging. These findings suggest that evolutionary search can effectively guide WFC generation when target objectives align with local structure.

cs.NE↗

Nonslop: A Gamified Experiment in Human-AI Collaborative Writing

The rapid proliferation of large language models (LLMs) raises critical questions about human creativity and individual expression in an era of AI-assisted creation. When do humans adopt AI suggestions, and what are the implications for individual voice? This study examines these questions through a gamified writing exercise where 74 participants (214 responses) replied to prompts while AI-generated word suggestions were available as they wrote. The game simulates a dystopian future in which an AI is attempting to learn from what remains of human individuality, and disincentivizes AI-like writing. In doing so, it attempts to create conditions that reveal authentic user preferences rather than default behaviors, such as accepting a readily available AI-generated suggestion. Note that this is a deliberate inversion of the "helpful assistant" design pattern; the system is explicitly forbidding you from accepting AI suggestions. We analyze user behavior patterns across different task types, user behaviors, and response characteristics to understand the factors influencing human-AI interaction in creative tasks. The study focuses on when users choose to maintain creative autonomy versus violating the rules of the game and accepting AI assistance. It also explores how these choices relate to response patterns, task characteristics, and user behavior. This gamified approach offers both a framework for studying authentic human-AI interaction and a provocative lens for understanding the tension between efficiency and authenticity in AI-augmented creativity.

cs.AI↗

In Search of the Ingredients of Open-Endedness: Replicating Picbreeder with Large Vision-Language Models

We are in the midst of large-scale industrial and academic efforts to automate the processes of scientific, technological and creative production through AI-driven assistants. Historically, a fundamental property of these processes in their human form has been their open-endedness: their capacity for generating a seemingly endless supply of novel and meaningful new forms. Do artificial agents have any capacity for such fruitful unguided discovery? To answer this question, we turn to Picbreeder, the canonical exemplar of human-driven open-ended search, in which users collaboratively generated a diverse library of images through interactive evolution of small neural networks. We replicate Picbreeder, replacing human users with frontier Vision Language Models (VLMs). We observe clear qualitative differences between the output of our system and the historical human baseline, and attempt to characterize them using metrics of phylogenetic complexity and visual and semantic salience and novelty. In an effort to identify some of the causal factors contributing these differences, we study the addition of exploratory noise to the agents' selection process, of behavioral diversity between agents, and of narrative momentum in the form of memory of past actions. We make our code available at https://github.com/smearle/picbreeder-vlm.

cs.AI↗

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

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 PCGRLLM, 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.

cs.AI↗

GVGAI-LLM: Evaluating Large Language Model Agents with Infinite Games

We introduce GVGAI-LLM, a video game benchmark for evaluating the reasoning and problem-solving capabilities of large language models (LLMs). Built on the General Video Game AI framework, it features a diverse collection of arcade-style games designed to test a model's ability to handle tasks that differ from most existing LLM benchmarks. The benchmark leverages a video game description language that enables the rapid creation of new games (including rules and levels), helping to prevent overfitting over time. Each game scene is represented by a compact set of ASCII characters, allowing for efficient processing by language models. GVGAI-LLM defines interpretable metrics, including meaningful step ratio, step efficiency, and overall score, to assess model behavior. Through zero-shot evaluations across 118 games with diverse challenges and skill depth, we reveal persistent limitations of LLMs in spatial reasoning and basic planning. Current models consistently exhibit spatial and logical errors, motivating structured prompting and spatial grounding techniques. Although these interventions lead to partial improvements, the benchmark remains very far from being solved. GVGAI-LLM serves as a reproducible testbed for advancing research on language model capabilities, with a particular emphasis on agentic behavior and spatial reasoning. Furthermore, its ability to generate infinite benchmarks, both manually and procedurally, provides a scalable framework for longitudinal evaluation.

cs.AI↗

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

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.

cs.CV↗

Ludax: A GPU-Accelerated Domain Specific Language for Board Games

Games have long been used as benchmarks and testing environments for research in artificial intelligence. A key step in supporting this research was the development of game description languages: frameworks that compile domain-specific code into playable and simulatable game environments, allowing researchers to generalize their algorithms and approaches across multiple games without having to manually implement each one. More recently, progress in reinforcement learning (RL) has been largely driven by advances in hardware acceleration. Libraries like JAX allow practitioners to take full advantage of cutting-edge computing hardware, often speeding up training and testing by orders of magnitude. Here, we present a synthesis of these strands of research: a domain-specific language for board games which automatically compiles into hardware-accelerated code. Our framework, Ludax, combines the generality of game description languages with the speed of modern parallel processing hardware and is designed to fit neatly into existing deep learning pipelines. We envision Ludax as a tool to help accelerate games research generally, from RL to cognitive science, by enabling rapid simulation and providing a flexible representation scheme. We present a detailed breakdown of Ludax's description language and technical notes on the compilation process, along with speed benchmarking and a demonstration of training RL agents. The Ludax framework, along with implementations of existing board games, is open-source and freely available.

cs.AI↗