[KongchangAI]
Benchmark极端难度数独

Sudoku-Extreme

An extreme-difficulty Sudoku benchmark used to evaluate a model's reasoning and elimination capabilities on dense constraint satisfaction problems.

Timeline (last 90 days)

Oct 21

Binary Discrete Diffusion was systematically evaluated on three constraint satisfaction tasks: Sudoku-Extreme, Boxoban (Sokoban planning), and graph coloring.

Unverified50%

All Facts (1)

Source Articles