Scores you can trust, on data your code has never seen.

Every submission uses fresh blinded data. Your code runs locally; the answer key stays on the server.

View benchmarks

Benchmarks

Loading benchmarks...

Popular benchmarks

All benchmarks →

Recent benchmarks

All benchmarks →

Get started

  1. Find a benchmark

    Browse the catalog and open a benchmark's Get Started page to read its task details.

  2. Install the CLI

    Install the client on your machine and log in. See the CLI setup guide for the exact commands.

  3. Submit and publish

    Submit your run from the CLI. Your score stays private until you choose to publish it to the leaderboard.

Want to organize a benchmark? →

—
Benchmarks
—
Solvers
—
Published entries
—
Submissions · 30 days

About blind benchmarking

Cartoon of the blind protocol: a blindfolded validator throws blinded data over a wall to a blindfolded developer, who throws results back; the answer key stays in a locked chest beside the validator's leaderboard podium.

Each run uses fresh blinded data: your code stays local, the answer key stays private, and only the score is published.

  • Fresh data. A new blinded dataset for every run.
  • Local execution. Your code stays on your machine.
  • Private scoring. The answer key never leaves the server.

Citation

If QudeLeap Blind Benchmarks results appear in your work, please cite:

@misc{qudeleap_blind_benchmarks,
  title  = {QudeLeap Blind Benchmarks},
  author = {{QudeLeap}},
  year   = {2026},
  url    = {https://github.com/QudeLeap/Decoder-Server}
}