sudoku-solver

This environment challenges the model to solve classic Sudoku puzzles. It tests logical deduction and constraint satisfaction over multiple turns.

Overview

Domain: games Base Class: MultiTurnEnv Difficulty: medium Task: The model must fill in the empty cells of a 9x9 Sudoku grid such that each row, column, and 3x3 subgrid contains all digits from 1 to 9.

Quickstart

Installation

uv run vf-install sudoku-solver

Usage

import verifiers as vf

env = vf.load_environment("sudoku-solver")
results = env.evaluate_sync(
    client=vf.OpenAI(),
    model="gpt-4.1-mini",
    num_examples=10,
    rollouts_per_example=1
)

Evaluation

Run an evaluation with default settings:

uv run vf-eval sudoku-solver

Configure model and sampling:

uv run vf-eval sudoku-solver \
  -m gpt-4.1-mini \
  -n 20 -r 3 -t 1024 -T 0.7

Environment Arguments

Arg	Type	Default	Description
`num_examples`	int	1000	Number of training examples
`num_eval_examples`	int	100	Number of evaluation examples
`seed`	int	42	Random seed for reproducibility

Metrics

Metric	Meaning
`reward`	Primary reward signal
`format_reward`	Format adherence reward (if applicable)

About

Generated by synthetic-rl-env-creator.

Tags: games