Tag
Introduces a 'Complexity Score' algorithm to determine when detailed prompts improve LLM performance for extracting suicide circumstances from NVDRS narratives, finding that LLMs outperform fine-tuned models on rare circumstances and proposing a hybrid approach.