Hugging Face's logo

Datasets:
utkarsh4430
/
SteerBench

Tasks:
Audio-to-Audio
Modalities:
Audio
Text
Formats:
parquet
Languages:
English
Size:
< 1K
ArXiv:
Tags:
speech
spoken-dialogue
steerability
benchmark
evaluation
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
SteerBench / data

Commit History

Yesterday

  1. SteerBench: a benchmark for spoken steerability in full-duplex speech assistants. Self-contained parquet with 390 items, 1067 binary rubrics, and the user + reference audio embedded. CC BY-NC-SA 4.0.

    utkarsh4430
    8769094