Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
2
A strong LLM can act as a reviewer of grounded drafts from a weak model, retaining most performance while drastically cutting latency and memory.