ReverseTicTacToe is an environment for evaluating agents on the inverse of classic tic-tac-toe where getting three in a row loses. This environment wraps the ReverseTicTacToe implementation from TextArena, a framework for text-based game environments.
- Inverted strategic reasoning
- Avoidance tactics and defensive positioning
- Forcing opponent mistakes through constraint creation
- Two-player competitive gameplay against an LLM opponent
ReverseTicTacToe does not require a sandbox. It has minimal compute requirements.
MIT.
There are two splits: train (50 tasks) and test (50 tasks). Each split contains 50 tasks across each of 1 variants:
- ReverseTicTacToe-v0
Each task is seeded for reproducibility.
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.
Game state is generated procedurally by the TextArena engine using seeded randomness. No external data files are required.
Agents are given a single tool:
place_mark(position): Place your mark on the board at the given position (0-8). 0=top-left, 4=center, 8=bottom-right.
ReverseTicTacToe is a multi-turn environment.
Moderate. Reverse TicTacToe requires inverting standard tic-tac-toe intuitions and thinking defensively to avoid creating three in a row while forcing the opponent into such positions.
This environment requires an OpenAI API key (passed via secrets) to power the LLM opponent.
Agents in ReverseTicTacToe interact only with a board game and have no access to external systems, the internet, or sensitive data. The environment does not present safety risks.
@software{textarena2024,
author = {Guertler, Leon and Banting, Wilfried and Pignatelli, Eduardo},
title = {TextArena},
year = {2024},
publisher = {GitHub},
url = {https://github.com/LeonGuertler/TextArena}
}