Reasoning Gym

We are building a python library of procedural dataset generators and algorithmically verifiable reasoning environments for training Reasoning Models with reinforcement learning (RL).

The goal is to generate virtually infinite data with adjustable complexity.

Quick Start

from reasoning_gym.arithmetic import ChainSum, ChainSumConfig

# Configure a simple arithmetic task generator
config = ChainSumConfig(
    min_terms=2,          # At least 2 numbers per expression
    max_terms=4,          # At most 4 numbers per expression
    min_digits=1,         # Single digit numbers
    max_digits=2,         # Up to 2-digit numbers
    allow_negation=False, # Only positive numbers
    size=5,              # Generate 5 examples
    seed=42              # For reproducibility
)

# Create the dataset
dataset = ChainSum(config)

# Generate some examples
for i in range(len(dataset)):
    item = dataset[i]
    print(f"Question: {item['question']}")
    print(f"Answer: {item['answer']}\n")

Example output:

Question: 7 + 42 - 15 =
Answer: 34

Question: 91 - 8 =
Answer: 83

Question: 4 + 67 - 12 =
Answer: 59

Question: 28 + 35 =
Answer: 63

Question: 51 - 24 + 7 =
Answer: 34

Generator / Environment Ideas

• math tasks
• algorithmic tasks (counting, sorting, re-ordering, ..)
• logic riddles
• logic inductive programming tasks
• ARC-AGI synthetic riddles

Call for Contributions

If you have ideas for additional procedural dataset generators or please create an issue here.

Or contact us in the #arc-agi-2 channel of the GPU-Mode discord server.