What causes AI models to hallucinate?
Not measured — no engine answer has been captured on this question yet. That is different from the engines answering and naming nobody.
1 of 5 names confirmed · named in 2 or more of 5 answers · 0 engines
AI models hallucinate because they are probabilistic systems optimized to predict the most statistically likely sequence of tokens, rather than to verify factual truth. This can lead to the generation of confident but incorrect outputs, especially when the model is faced with incomplete or ambiguous context. The primary causes of hallucination include data limitations, statistical nature, evaluation and incentives, context and memory limits, and ambiguity and task mismatch.
- 1OpenAInamed in 4 of 5 answers
- 2IBMnamed in 1 of 5 answersone answer
- 3ACMnamed in 1 of 5 answersone answer
- 4Arxivnamed in 1 of 5 answersone answer
- 5Googlenamed in 1 of 5 answersone answer
LOCKED
The full measurement
- The position each of the 5 engines gave all 5 names.
- How many of the 5 answers named each of them.
- Where the engines disagree, and by how much.
- Fan-out — the query each engine actually searched.
- Every citation, and the sources nobody cited.