Fig 1.
One-shot prompt for the Large Language model Meta AI (LLaMA) model with 7 billion parameters.
Fig 2.
The fine-tuning flowchart for the qLLaMA_LoRA-7B model by updating the pre-trained LLaMA-7B model’s parameters using the Low-Rank Adaptation (LoRA), a supervised learning algorithm, from 100,000 Quora question pairs.
LLaMA-7B: Large Language model Meta AI with 7 billion parameters.
Table 1.
Model comparison for question pairs detection using 10-fold cross validation.
Table 2.
External validation results between two fine-tuned models (qLLaMA_LoRA-7B and qLLaMA3.1_LoRA-70B) based on GLUE benchmark.