Persian Audio Reasoning and Speech Assessment Benchmark
16 tasks · 8,000+ samples · 3 evaluation dimensions
Persian poses unique audio understanding challenges through its classical poetry, traditional music, and pervasive code-switching — none captured by existing benchmarks. We introduce PARSA-Bench (Persian Audio Reasoning and Speech Assessment Benchmark), the first benchmark for… See the full description on the dataset page:
https://huggingface.co/datasets/MohammadJRanjbar/PARSA-Bench.
