[Marked for Deprecation. please visit https://github.com/brain-score/language for the migrated project] Benchmarking of Language Models using Human Neural and Behavioral experiment data
We rely on 'tokenized length' to detect boundaries between previous and current stimulus. However, this assumption works when we sequentially add a stimulus to the context group as in the unidirectional case. In the bidirectional case, the entire context group is used, so the tokenized length is always the total length.
https://github.com/language-brainscore/langbrainscore/blob/a2ad6bd81ac08350aaf2e21488290522ede57dd4/langbrainscore/encoder/ann.py#L193-L194
We rely on 'tokenized length' to detect boundaries between previous and current stimulus. However, this assumption works when we sequentially add a stimulus to the context group as in the
unidirectionalcase. In the bidirectional case, the entire context group is used, so the tokenized length is always the total length.