This directory contains the FLE leaderboard system, which tracks and displays performance metrics for different LLM agents in the Factorio Learning Environment.
/results/
: JSON files with raw results from each model/processed/
: Combined and processed results (auto-generated, do not modify directly)/src/
: React application source code for the leaderboard UITo submit new model results to the leaderboard:
/leaderboard/results/
directorymodel-name.json
(e.g., claude-3-5-sonnet.json
){
"model": "Your Model Name",
"productionScore": 123456,
"milestones": 20,
"labTasksCompleted": 5,
"mostComplexItem": "advanced-circuit",
"timeToElectricDrill": 3200,
"submittedBy": "Your Name",
"submissionDate": "YYYY-MM-DD"
}
Field | Description | Required |
---|---|---|
model |
Name of the model | Yes |
productionScore |
Total production score achieved | Yes |
milestones |
Number of milestones reached | Yes |
labTasksCompleted |
Number of lab tasks successfully completed | Yes |
mostComplexItem |
Most complex item produced | Yes |
timeToElectricDrill |
Steps until first electric drill deployed | No |
submittedBy |
Your name or identifier | Yes |
submissionDate |
Date of submission (YYYY-MM-DD) | Yes |
To run the leaderboard locally:
cd leaderboard
npm install
npm start
http://localhost:3000
Model results go through a two-stage verification process:
Only verified results are considered official for research comparisons.