ParaVT-Source

ParaVT-Source Source media archives for the ParaVT training corpus. Pair this repository with the annotations in ParaVT/ParaVT-Parquet. Overview ParaVT is a multi-agent agentic framework for long-video understanding, post-trained with PARA-GRPO (Parseability-Anchored and Ratio-gAted GRPO). This dataset bundles the raw video files referenced by every row in ParaVT-Parquet, packaged as per-source zip archives. Layout Files are grouped by… See the full description on the dataset page:

Source: Hugging Face Hub (ParaVT/ParaVT-Source). Metadata imported from the dataset’s Hub tags.