Search papers, labs, and topics across Lattice.
University of Turku
3
0
4
LLMs trained on a synthetic corpus can outperform native data benchmarks while using significantly fewer tokens, challenging the assumption that more data always leads to better performance.
Forget training objectives – the secret to a great embedding model might just be how consistently it organizes its embedding space.
LLMs can categorize 350K historical mentions of leisure activities and organizational memberships with near-expert accuracy, unlocking large-scale quantitative analysis of social integration from digitized archives.
