Minesweeper AI research
Sweeper
A Minesweeper AI system benchmarking symbolic reasoning, exact inference, neural networks, and hybrid agents.
The project
Sweeper studies how an agent should make decisions in Minesweeper. It starts with symbolic proofs, uses exact mine probabilities when deduction runs out, and evaluates neural and hybrid strategies against the same board environments. The browser interface makes the work inspectable through a playable board, assisted modes, and benchmark results.
What it does
Reasoning first
Turns visible board clues into constraints and identifies provably safe moves before guessing.
Exact probabilities
Counts valid frontier assignments to rank candidate moves when the board has no certain answer.
Comparable strategies
Evaluates symbolic, neural, and hybrid agents within the same reproducible board environment.
