EN /

Trace Studio

Trace Studio is an open-source research group building game AI systems. We develop high-performance game engines, train AI solvers using heuristic search, reinforcement learning, and large language models, and publish technique reports documenting our methods and findings.

Engine, Solvers, and Research

High-Performance Engine

Fast, correct game engines built for AI research and human play.

AI Solvers

Heuristic, reinforcement learning, and LLM-based game-playing agents.

Technique Reports

Detailed write-ups on methods, benchmarks, and findings.

From Heuristics to LLMs

[ 01 ]

Heuristic

Hand-crafted rules and search algorithms as strong baselines.

[ 02 ]

Reinforcement Learning

Self-play and policy optimization to surpass heuristic performance.

[ 03 ]

LLM Agents

Large language models as game-playing agents with strategic reasoning.

Get Involved

Trace Studio is open to contributors. Visit our GitHub organization to explore projects, open issues, or submit pull requests.