The Unwritten Benchmark

The Unwritten Benchmark Dataset Summary The Unwritten Benchmark is a multimodal benchmark for acousto-kinematic word inference: given only the sound of a pen scratching on paper and/or the motion of a hand writing, models must infer the underlying word, even though no visible ink trace is present. The dataset was introduced in the paper: The Unwritten Benchmark: A New Challenge for Multimodal Machine Learning in Abstract Perceptual ReasoningGarima Arya Yadav, Nilay Yilmaz… See the full description on the dataset page:

Source: Hugging Face Hub (RiRi-y/unwritten-benchmark). Metadata imported from the dataset’s Hub tags.