Model-brain alignment as a function of context length. (IMAGE)

The Hong Kong Polytechnic University

Model-brain alignment as a function of context length.

Caption

Increasing the context length available to the models affected model-brain alignment in a nonmonotonic way, and there was evidence that the MLM_NSP model performed better than the MLM model for both coherent and unconnected sentence relationships. (A) Effect of context length for the Reading-brain2019 dataset (coherent sentences) in the language network. (B) Effect of context length for the Pereira2018 dataset (mainly unconnected sentences) in the MD network. Asterisk (*) indicates statistical significance after FDR correction (α = 0.05). Yellow areas indicate ROIs identified when no preceding sentences were given. Gray areas indicate ROIs that showed a significant advantage of the MLM_NSP model at different context lengths.

Credit

© 2024 Research and Innovation Office, The Hong Kong Polytechnic University. All Rights Reserved.

Usage Restrictions

Nil

License

Original content

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.