VTS Data (VideoTreeSearch)

VTS Data Training and evaluation data for VideoTreeSearch (VTS): Searching Videos as Trees — Self-Correcting Agents for Grounded Long Video QA. Code & full documentation: Model checkpoint: This dataset contains the synthesized tree-search trajectories, the SFT and RL training sets, pre-built scene-tree caches, and the inference / evaluation annotations used in the paper. Videos are not… See the full description on the dataset page: data.

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