Generative AI foundations · Question 43 of 60
When analysing a very long log, an LLM no longer takes the early entries into account and contradicts them. What is the most likely cause?
1 answer is correct. Hard Created with AI
Show answer
Answer
- C The input exceeds the model's context window
Explanation
Whatever goes beyond the context window is no longer visible to the model, so early parts drop out. A low temperature makes answers more uniform but does not explain the forgetting.