1 paper
Kaushal Bhogale, Manas Dhir, Amritansh Walecha +11
Existing Indic ASR benchmarks often use scripted, clean speech and leaderboard driven evaluation that encourages dataset specific overfitting. In addition, strict single reference…