Skip to main content
. 2024 May 16;3:e52095. doi: 10.2196/52095

Table 4.

Threshold regression point estimates and 95% confidence intervals for number of sentences and EPSa by architecture.

Model (parameters) Number of sent threshold, estimate (95% CI) EPS threshold, estimate (95% CI)
RoBERTa_base (125M) 448 (437-456) 1.36 (1.35-1.37)
GatorTron_base (345M) 448 (409-456) 1.36 (1.36-1.38)
RoBERTa_large (355M) 439 (409-451) 1.36 (1.35-1.38)
GPT-2_large (774M) 527 (511-540) 1.38 (1.36-1.38)

aEPS: entities per sentence.