Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
2
A strong LLM can review grounded drafts from a weak model, retaining most task performance while halving latency and memory traffic.
