Publications (14)
ByteSized32: A Corpus and Challenge Task for Generating Task-Specific World Models Expressed as Text Games
Ruoyao Wang, Graham Todd, Eric Yuan +3
In this work, we investigate the capacity of language models to generate explicit, interpretable, and interactive world models of scientific and common-sense reasoning tasks. We op…
People use fast and flat simulation to reason about new games
Katherine M. Collins, Cedegao E. Zhang, Lionel Wong +6
Games have long been a microcosm for studying planning and reasoning in both natural and artificial intelligence (AI), often focusing on expert-level or even super-human play. But…
PuzzleJAX: A Benchmark for Reasoning and Learning
Sam Earle, Graham Todd, Yuchen Li +5
We introduce PuzzleJAX, a GPU-accelerated puzzle game engine and description language designed to support rapid benchmarking of tree search, reinforcement learning, and LLM reasoni…
Goals as Reward-Producing Programs
Guy Davidson, Graham Todd, Julian Togelius +2
People are remarkably capable of generating their own goals, beginning with child's play and continuing into adulthood. Despite considerable empirical and computational work on goa…
Learning Compositional Negation in Populations of Roth-Erev and Neural Agents
Graham Todd, Shane Steinert-Threlkeld, Christopher Potts
Agent-based models and signalling games are useful tools with which to study the emergence of linguistic communication in a tractable setting. These techniques have been used to st…
Evaluating Language Models' Evaluations of Games
Katherine M. Collins, Cedegao E. Zhang, Graham Todd +9
Reasoning is not just about solving problems -- it is also about evaluating which problems are worth solving at all. Evaluations of artificial intelligence (AI) systems primarily f…