TycoonLE Released
TycoonLE, a Jax reinforcement learning environment, is released for long-horizon planning, allowing agents to operate in a simulated logistics economy.
Key points
- TycoonLE is designed for economically grounded, long-horizon planning with agents allocating capital and managing debt.
- The environment uses a fixed-shape interface, making rollouts compatible with JAX transformations.
- TycoonBench provides a companion benchmark report for comparing agent and model performance on TycoonLE planning tasks.
- The environment can be installed using Python 3.11 or 3.12 and requires the installation of specific packages.
- A quickstart guide is provided, including example code and a browser UI for loading replays.
TycoonLE is a reinforcement learning environment designed for long-horizon planning. It allows agents to operate in a simulated logistics economy, allocating capital, building transport routes, and managing debt. The environment is built using Jax and provides a fixed-shape interface, making it compatible with JAX transformations such as jit, vmap, and scan. The TycoonLE environment is designed to study various aspects of decision-making, including action legality, candidate-frontier decision interfaces, financing timing, delayed rewards, procedural variation, and replayable audit traces. A companion benchmark report, TycoonBench, is provided to compare agent and model performance on TycoonLE planning tasks. The environment can be installed using Python 3.11 or 3.12, and a quickstart guide is provided, including example code and a browser UI for loading replays.
Sources
The WireByte editorial team synthesises technology news from multiple primary sources, verifies the facts, and links every source. Articles are produced with AI assistance and reviewed under our editorial policy.