A multimodal offline RL framework that fuses LLM-encoded clinical notes with structured vitals and labs modestly improves sepsis policy scores on some datasets, with evaluation caveats.
Deep reinforcement learning extracts the optimal sepsis treatment policy from treatment records
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
MORE-CLEAR: Multimodal Offline Reinforcement learning for Clinical notes Leveraged Enhanced State Representation
A multimodal offline RL framework that fuses LLM-encoded clinical notes with structured vitals and labs modestly improves sepsis policy scores on some datasets, with evaluation caveats.