Search papers, labs, and topics across Lattice.
University of Sheffield
1
0
3
Logic pre-pretraining accelerates language model skill acquisition by 36B tokens while enhancing compressibility, revealing a new path for efficient model training.
