Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
4
Retrieval-constrained LLM adjudication boosts clinical value set completion accuracy, achieving a macro F1 score increase from 0.287 to 0.549.
GPT-4o hallucinates nearly half of its clinical codes when asked to define clinical concepts, but a retrieval-augmented classifier cuts the error rate by 75% while also outperforming simpler models.