Search papers, labs, and topics across Lattice.
Purdue University
2
0
6
3
Even after removing names and pronouns, language models can still guess an applicant's gender from recommendation letters with surprising accuracy, revealing hidden biases lurking in seemingly objective text.
Long-sequence recommendation models, previously considered impractical to train, can now be effectively trained on commodity hardware with minimal accuracy loss, thanks to a new open-source framework and a clever embedding layer.
