04:00
2026-08-24
arxiv.org
large-language-models
Inhibitory Attention for Clinical Long-Context Reasoning: Characterizing and Mitigating Lost-in-the-Middle Effects in EHR Processing
A new study from arXiv (2608.20348v1) characterizes the clinical lost-in-the-middle (CLitM) effect in electronic health record (EHR) processing, where large language models (LLMs) retrieve information…