Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
1
0
1
2
Removing the least important attention heads from transformers can preserve performance in tabular data tasks, challenging conventional wisdom about model architecture optimization.