Closed loops, not prompts
LSS YAML → LoopGym SimEnv → Success@k and observed LES.
Leaderboard
Generalist rank — mean of four suite scores.
View
Task
Chart
| # | Loop | Submitter | LES ↕ | Success@k ↕ | Cost ↕ | Harness | Spec | Repro |
|---|
Tasks
Metrics
What LoopBench measures
Reproducible
Five seeds, SimEnv v0.2, auditable specs.
Community scoreboard
External rows credited; human review on merge.
Run your first score
pip install "le-loopforge>=0.2.0" "le-loopctl>=0.1.0" loopbench loopgym
loopbench run --suite suite-repair --spec your-loop.yaml --seeds 0,1,2,3,4 -o results.json
loopbench validate results.json