WinAsMuchAsYouCan

Description

WinAsMuchAsYouCan is an environment for evaluating agents on strategic decision-making and cooperation in a multi-player coordination game. This environment wraps the WinAsMuchAsYouCan implementation from TextArena, a framework for text-based game environments.

Capabilities

• Strategic coordination in multi-player settings
• Communication and persuasion during bonus rounds
• Risk assessment with varying payoff multipliers
• Adapting to opponent behavior patterns

Compute Requirements

WinAsMuchAsYouCan does not require a sandbox. It has minimal compute requirements.

License

MIT.

Tasks

There are two splits: train (150 tasks) and test (150 tasks). Each split contains 50 tasks across each of 3 variants:

• WinAsMuchAsYouCan-v0
• WinAsMuchAsYouCan-v0-train
• WinAsMuchAsYouCan-v0-raw

Each task is seeded for reproducibility.

Reward Structure

This is a sparse reward environment. Rewards are mapped from TextArena's native range of {-1, 0, 1} to {0.0, 0.5, 1.0} via (raw + 1) / 2.

We do not use LLM graders for this environment; reward is determined programmatically.

Data

Game state is generated procedurally by the TextArena engine using seeded randomness. No external data files are required.

Tools

Agents are given three tools:

• send_message(message): Send a chat message during talk phases.
• choose_x(): Choose X for this round.
• choose_y(): Choose Y for this round.

Time Horizon

WinAsMuchAsYouCan is a multi-turn environment.

Environment Difficulty

Medium-High. Agents must navigate complex payoff structures across 10 rounds with 4 players, where outcomes depend on the collective choices of all players. Bonus rounds with talk phases and higher multipliers require strategic communication and coordination.

Other Environment Requirements

This environment requires an OpenAI API key (passed via secrets) to power the LLM opponents.

Safety

Agents in WinAsMuchAsYouCan interact only with a coordination game and have no access to external systems, the internet, or sensitive data. The environment does not present safety risks.

Citations

@software{textarena2024,
  author    = {Guertler, Leon and Banting, Wilfried and Pignatelli, Eduardo},
  title     = {TextArena},
  year      = {2024},
  publisher = {GitHub},
  url       = {https://github.com/LeonGuertler/TextArena}
}