Hallucination
A phenomenon in which large language models generate content that is factually incorrect or fabricated, yet presented with high confidence. It arises from the nature of language models learning statistical plausibility rather than factual accuracy, and is a central problem in research on AI reliability and honesty.
Core Facts
Timeline (last 90 days)
The hallucination problem cannot be eradicated by increasing the number of parameters or switching to better corpora, because the probabilistic generation paradigm itself determines that deviations are inevitable.
Relying purely on LLMs to generate code suffers from hallucination and unreliability issues.
The hallucination problem in large language models is divided into two categories: intrinsic hallucination and extrinsic hallucination.
All Facts (4)
Hallucination refers to content generated by a model that appears plausible but is actually incorrect or fabricated, and is the most intractable technical problem with LLMs.
80%UnverifiedThe hallucination problem in large language models is divided into two categories: intrinsic hallucination and extrinsic hallucination.
70%UnverifiedThe hallucination problem cannot be eradicated by increasing the number of parameters or switching to better corpora, because the probabilistic generation paradigm itself determines that deviations are inevitable.
50%UnverifiedRelying purely on LLMs to generate code suffers from hallucination and unreliability issues.
50%