datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
vivos_noisysbpn-mms-final-vote-5episodes-review
SBPN + MMS final-vote five-episode review set
This dataset contains 474 final selected clips (0.59 hours)
from five source episodes. It is intentionally marked pending manual review.
Selection policy
One winner is retained per transcript chunk using:
frozen_vote
+ 0.10 * mean_rank(MMS_forward, SBPN_forward)
+ 0.10 * mean_rank(-SBPN_previous_context_gain, -SBPN_next_context_gain)
Ranks are calculated only among candidate cuts for the same chunk. This policy was… See the full description on the dataset page: https://huggingface.co/datasets/NancyT/sbpn-mms-final-vote-5episodes-review.