How to evaluate LLM generated responses

[manisnesan]

https://radekosmulski.com/how-to-evaluate-an-llm-on-your-data/

• use semantic similarity, where you generate an answer from the LLM and compare it to the ground truth
• evaluating the LLM's outputs using perplexity, which measures how well the model predicts the next word in a sequence
• using both semantic similarity and perplexity to evaluate the LLM's outputs