Table 6 Processing and training time.

From: Detecting stigmatizing language in clinical notes with large language models for addiction care

Approach

Average Inference Time per Input (s)

Train Time (hrs)

Supervised-Fine Tuning (SFT)

4.80

5.90

In-Context Prompt

3.37

n/a

Retrieval Augmented Prompt (RAG)

46.46

n/a

Zero-Shot

1.57

n/a

Keyword

<0.001

n/a

  1. Average inference time per input was averaged across all evaluation data (excluding external validation). Training time was only relevant to SFT.
Back to article page