A two-stage pipeline of LLM text encoding and PageRank over transition clusters yields 5-element state vectors that modestly predict MMSE and ADAS-Cog scores in 50 dementia patients.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Two-Stage Representation Learning for Analyzing Movement Behavior Dynamics in People Living with Dementia
A two-stage pipeline of LLM text encoding and PageRank over transition clusters yields 5-element state vectors that modestly predict MMSE and ADAS-Cog scores in 50 dementia patients.