Table 5 Test accuracy of FLAN-5 models in scenario 2 (three combination of train sets).

From: Verbal lie detection using Large Language Models

Train set

Test set

Model size

Test accuracy

Opinion + Memory

Intention

FLAN-T5 small

55.37

FLAN-T5 base

55.67

Opinion + Intention

Memory

FLAN-T5 small

55.37

FLAN-T5 base

54.23

Memory + Intention

Opinion

FLAN-T5 small

53.12

FLAN-T5 base

49.40

  1. The performance comparison is among the small and base version of the FLAN-T5 model in the three combination of train set: opinion + memory, opinion + intention, memory + intention.
Back to article page