Video-Adverb Retrieval (Unseen Compositions)

The task aims to recognize adverbs beyond seen adverb-action compositions, i.e. compositions that were not seen during training.