REPO — The leaderboard ranks quality vs environmental cost
The leaderboard ranks quality vs environmental cost: a box distribution of the primary metric (F1-macro for triage, MAE otherwise) by model/hardware, and a log-x scatter of CO2 grams vs quality, plus a summary of submission/contributor/model/scenario/profile/verified counts.
Source: puma-leaderboard app.py · Traceability: corpus unit
SAT-016· Confidence: repo-backed