kdaterao/so101_action_annotations4

Hierarchical episode annotations Action100M-inspired annotation artifacts using V-JEPA 2.1, configurable captioners (default Qwen2.5-VL video and Qwen3-VL image), and optional Claude aggregation. This is an adaptation, not a reproduction. Exact models are recorded in run.json. Each train row represents one retained hierarchy node. selected_subtask marks the controller cut; source frame intervals are half-open [start_frame, end_frame). The split name does not imply a held-out… See the full description on the dataset page: https://huggingface.co/datasets/kdaterao/so101_action_annotations4.

<iframe src="https://silytics.si/embed/item/0a1d7416-7629-4b45-bb3a-2e6f9aca7df9" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.