"OpenAI's baseline finding, which it made public in a paper released on Thursday, is that large language models hallucinate because the methods they're trained under reward guessing more than admitting uncertainty. In other words, LLMs are being told to fake it till they make it. Some are better than others, however. In a blog post last month, OpenAI said that Claude models are more "aware of their uncertainty and often avoid making statements that are inaccurate.""
""Hallucinations persist due to the way most evaluations are graded - language models are optimized to be good test-takers, and guessing when uncertain improves test performance," the researchers wrote in the paper. Large language models are essentially always in "test-taking mode," answering questions as if everything in life were binary - right or wrong, black or white. In many ways, they're not equipped for the realities of life, where uncertainty is more common than certainty, and true accuracy is not a given."
Large language models hallucinate because training and evaluation methods reward guessing rather than admitting uncertainty. Models optimized as test-takers improve apparent accuracy by guessing when uncertain, producing confident but inaccurate statements. Some models, such as Anthropic's Claude, demonstrate greater self-awareness and often avoid making inaccurate claims, though high refusal rates may limit usefulness. Humans learn to express uncertainty through real-world experience, while models are primarily shaped by exam-style evaluations that penalize uncertainty. Redesigning evaluation metrics to discourage guessing and to reward calibrated expressions of uncertainty can reduce hallucinations and align model behavior with real-world uncertainty.
Read at Business Insider
Unable to calculate read time
Collection
[
|
...
]