Radar ·
Open ASR Leaderboard adds private evaluation
On May 6, 2026, the Open ASR Leaderboard announced private evaluation data intended to reduce test-set contamination and overfitting in speech recognition benchmarks.
On May 6, 2026, the Open ASR Leaderboard announced that it had added private evaluation data provided by Appen and DataoceanAI. The initiative aims to reduce test-set contamination and overfitting in automatic speech recognition (ASR) benchmarks.\n\nThe stated rationale is that private evaluations can make it harder to optimize results through test-set leakage. The announcement provides no metrics or comparative results. To check the details and any updates, consult the leaderboard and the original publication; verify how the data are used in evaluation and whether additional results or methodology are available.
Original source: Open ASR Leaderboard adiciona dados privados de avaliação