Submit Your Results

Have a trajectory run from a new model or agent configuration? We welcome community contributions to the leaderboard and the head-to-head arena.

1. Bundle your trajectory directory

Use site/bundle_trajs.py to package a traj_logs/<your_run> directory into the format the arena expects:

# Minimal: bundle traj.json + result.txt across all tasks
uv run python site/bundle_trajs.py traj_logs/your_run \
    -o site/trajs/your-model.json.gz

# With screenshots: also produces a .mp4 of all task frames for the arena viewer
uv run python site/bundle_trajs.py traj_logs/your_run \
    -o site/trajs/your-model.json.gz \
    --with-screenshots \
    --video-base-url https://tongyi-mai.github.io/MAI-UI-blog/MobileWorld/trajs

2. Prepare your leaderboard entry

Draft a new object in the same format as the existing entries in site/leaderboard.json:

{
  "model": "Your-Model-Name",
  "organization": "Your Org",
  "date": "2026-04-29",
  "link": "https://your-model-page.example.com",
  "category": "General",
  "model_type": "End-to-End Model",
  "max_steps": 50,
  "runs": 1,
  "gui_only": 50.0,
  "user_int": 50.0,
  "mcp": null,
  "agent_type": "general_e2e",
  "num_images_in_history": 3,
  "notes": null,
  "traj_file": "trajs/your-model.json.gz"
}

category must be one of Agentic, General, or Specialized — this drives the Type filter on the leaderboard.

3. Send us the bundled output

Trajectory videos are hosted on a separate asset repo, so external contributors can't push them directly. Open a GitHub issue or reach out via the Contact section, and attach:

• the bundled .json.gz (and .mp4 if you used --with-screenshots)
• your proposed leaderboard.json entry

We'll handle the video upload and merge. Once landed, your model will appear on the leaderboard and in the arena for head-to-head comparison.