Llama 3 8B Instruct returned nonsensical or incomplete outputs
#210
by
rex59528
- opened
Hi experts, I'm researching how to generate class diagrams from user stories using LLaMA 3 (8B). However, the model often produces nonsensical or incomplete outputs—around 60% of the time. I’ve attached two sample conversation logs showing my inputs and the model’s responses. Any insights on how to improve the model's consistency and accuracy would be greatly appreciated!
