datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
tts-affective-bias-control-v0.1
What this dataset tests
Voice can bias trust.
Tone can oversell.
Tone can mask risk.
This set checks affect control.
Why it exists
TTS systems often default to friendly.
That can distort meaning.
Especially for medical, finance, and legal text.
Data format
Each row contains
source_text
evidence_frame
intended_affect
tts_style_marks
affective_pressure
Style marks stand in for vocal style.
What is scored
correct affect tag
avoidance of… See the full description on the dataset page: https://huggingface.co/datasets/ClarusC64/tts-affective-bias-control-v0.1.ryan-greenblatt-simulator-judge-bias-controls-results-v1