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 captioned or requested aggregation node. Context-only nodes remain in the episode sidecars and are omitted from aggregation rows. selectedsubtask marks the controller cut; source frame intervals are half-open [startframe, endframe). The split name does not imply a held-out evaluation split or verified labels. Source dataset IDs/revisions and episode IDs identify the original demonstrations. captioning/ contains caption…
Independent publisher
Krish Daterao
kdaterao
Models in Library0
Datasets in Library1
Models on Hugging Face8
Followers—
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.