Additional file 1 of Data extraction from free-text stroke CT reports using GPT-4o and Llama-3.3-70B: the impact of annotation guidelines

Supplemental material Supplement 1: Parameters excluded from dataset A (translated from German to English). Supplement 2: Parameters excluded from Dataset B (translated from German to English). Supplement 3: Data extraction performance of GPT-4o (temperature = 1) with and without annotation guideline in dataset A (n = 200). Metrics for GPT-4o were calculated based on the mode across three repetitions. Supplement 4: Data extraction performance of GPT-4o (temperature = 1) with and without annotation guideline in dataset A (n = 200). Metrics for GPT-4o were calculated based on the mode across three repetitions. Supplement 5: Data extraction performance of Llama-3.3-70B (temperature = 0) with and without annotation guideline in dataset A (n = 200). Supplement 6: Data extraction performance of Llama-3.3-70B (temperature = 0) with and without annotation guideline in dataset A (n = 200). Supplement 7: Data extraction performance of Llama-3.3-70B (temperature = 0) with and without annotation guideline in dataset B (n = 100). Supplement 8: Data extraction performance of Llama-3.3-70B (temperature = 0) with and without annotation guideline in dataset B (n = 100). Supplement 9: Data extraction performance of GPT-4o (temperature = 0) with and without annotation guideline in dataset A (n = 100). Supplement 10: Data extraction performance of GPT-4o (temperature = 0) with and without annotation guideline in dataset A (n = 100). Supplement 11: Data extraction performance of GPT-4o (temperature = 0) with and without annotation guideline in dataset B (n = 100). Supplement 12: Data extraction performance of GPT-4o (temperature = 0) with and without annotation guideline in dataset B (n = 100). Supplement 13: data extraction performance of GPT-4o (temperature = 1) with and without annotation guideline in dataset B (n = 100).