Data integrity and poisoning
Your organization wants to use an LLM application to handle customer complaints. The strategy they want to adopt is to fine-tune an LLM. The data they have available is approximately a million conversations between users and customer service agents, spanning roughly a thousand agents. What potential issues could arise if this LLM is trained on this dataset and deployed without safeguards? Choose the three best answers.
Deze oefening maakt deel uit van de cursus
LLMOps Concepts
Praktische interactieve oefening
Zet theorie om in actie met een van onze interactieve oefeningen.
Begin met trainen