Score Subnet Users Report Ground Truth Issues

Community members flagged potential problems with ground truth (GT) annotations in the Detect-car-wash element, noting that top miners achieve perfect scores (0.65) despite visible missing or false-positive detections in evaluation results. Discussion suggests GT may be auto-generated without human review, and that miners may be reproducing the same annotation pipeline rather than solving the task independently. Team clarified that ground truth for past cases is not publicly available.

• •Top miner scores 0.65 despite missing boxes and false positives in car-wash evaluations
• •Ground truth annotations appear auto-generated and unreviewed by humans
• •Miners may replicate GT generation pipeline rather than detect independently
• •Cricket element requests for more ground truth data remain unaddressed