Skip to content
Evaluation sources

Roboflow Vision Evals

Roboflow · Image reading, counting, OCR and image information extraction

Official evaluation
In NewsObserving
Evidence budgetNot scored
Upstream data as ofTo be confirmed
Last syncedNot collected yet

What it measures, and how

Six tasks share a budget; strict accuracy and model-judge scores are distinguished, and the half-range of three runs is not mixed with the 95% interval.

How this evidence is used

Under observation: the new leaderboard's data interface returns rate limiting; the old open-source repository does not contain this round's new models, and old data is not passed off as the new leaderboard.

Limits and data attribution

Under observation: the new leaderboard's data interface returns rate limiting; the old open-source repository does not contain this round's new models, and old data is not passed off as the new leaderboard.

Data licence: Leaderboard data usage boundaries to be confirmed

Scores published by Roboflow. Raw scores and the News consensus score use different scales and cannot be added directly.