223
edits
Hallucination: Difference between revisions
no edit summary
No edit summary |
No edit summary |
||
| (One intermediate revision by the same user not shown) | |||
| Line 7: | Line 7: | ||
=== Types of Hallucinations === | === Types of Hallucinations === | ||
There are various forms of hallucinations that can manifest in LLM outputs: | There are various forms of hallucinations that can manifest in LLM outputs: | ||
* | *'''Sentence Contradiction:''' When a generated sentence contradicts a previous one within the same context. | ||
* | *'''Prompt Contradiction:''' Occurs when the response directly opposes the initial prompt's intent. | ||
* | *'''Factual Errors:''' These are outright inaccuracies or misrepresentations of verifiable information. | ||
* | *'''Nonsensical Outputs:''' Responses that, while possibly grammatically correct, are irrelevant or absurd in the given context. | ||
== Causes of Hallucinations == | == Causes of Hallucinations == | ||
| Line 35: | Line 35: | ||
=== Multi-Shot Prompting === | === Multi-Shot Prompting === | ||
Providing multiple examples of the desired output or context can help the model better understand and adhere to the user's expectations. This approach is particularly effective for tasks requiring specific formats or styles. | Providing multiple examples of the desired output or context can help the model better understand and adhere to the user's expectations. This approach is particularly effective for tasks requiring specific formats or styles. | ||
==Hallucination in Machine Learning== | ==Hallucination in Machine Learning== | ||