IsItPrime

API Endpoint
Leaderboard
Loading leaderboard...
README

IsItPrime

⭐ OpenReward Environment

Description

IsItPrime is a time-pressured primality testing environment. The agent is given a number and must determine whether it is prime, answering "Yes" or "No" within a 10-second time limit. The environment tests both mathematical reasoning and response speed.

Capabilities

  • Primality determination for numbers up to 100,000
  • Time-constrained decision making (10-second limit)
  • Binary classification (prime vs. composite)

Compute Requirements

Minimal. No GPU or significant memory required. Tasks are generated deterministically at startup.

License

MIT.

Tasks

There are two splits:

  • train: 1,000 tasks (500 primes, 500 composites)
  • test: 100 tasks (50 primes, 50 composites)

Numbers are sampled from the range [2, 100,000] using a fixed random seed for reproducibility.

Reward Structure

This is a sparse, verifiable reward environment. The reward is binary:

  • 1.0 if the agent correctly identifies whether the number is prime within 10 seconds
  • 0.0 if the answer is incorrect, the response is not "Yes"/"No", or the 10-second time limit is exceeded

The timer starts when get_prompt() is called and stops when the answer tool is invoked.

We do not use LLM graders for this task.

Data

Tasks are generated deterministically at class load time using a sieve of Eratosthenes for primes up to 100,000 and random sampling with fixed seeds (42 for train, 123 for test). No external data files are required.

Tools

  • answer: Accepts a single string parameter (answer) which must be "Yes" or "No". Returns the reward and correctness metadata.

Time Horizon

Single-turn environment. Exactly one tool call per task.

Other Environment Requirements

There are no further environment requirements; IsItPrime works out of the box with the OpenReward endpoint without any secrets.

Citations

@dataset{GRIsItPrime,
  author    = {General Reasoning Inc. Team},
  title     = {IsItPrime},
  year      = {2026},
  publisher = {OpenReward},
  url       = {https://openreward.ai/RJT1990/IsItPrime}
}
RJT1990/IsItPrime | OpenReward