Search papers, labs, and topics across Lattice.
1
0
3
22
LLMs can now scale depth more effectively: a new attention mechanism recovers diluted features in deeper layers, boosting performance with negligible overhead.
