Search papers, labs, and topics across Lattice.
Department of Computer Science
3
0
5
15
Optimal granularity in RAG benchmarks varies by dimension, with question complexity thriving on fine distinctions while other factors favor medium granularity.
The hardest AI tasks remain largely unsolved, with current models achieving only a 2.6% success rate on economically valuable workflows.
You can accurately predict the NDCG of a 1B-parameter reranking model by only training models up to 400M parameters, unlocking massive compute savings.
