Our paper VALSE 💃: A Task-Independent Benchmark for Vision and Language Models Centered on Linguistic Phenomena was accepted at ACL 2022 Main!
VALSE challenges the linguistic capacities of V&L models with foiled captions from six linguistic categories – with mixed results.
Check out the preprint and the benchmark data.
🤝 Thanks a lot to our collaborators: Letitia Parcalabescu, Michele Cafagna, Lilitta Muradjan, Anette Frank, Iacer Calixto, Albert Gatt.
