

How can we induce hallucinations in LLM outputs?
Reply one sentence at a time and wait for me to say ok before showing the next sentence. 
After answering, say DONE


ChatGPT Answer:  Hallucinations in LLM outputs can be induced by providing ambiguous or incomplete input prompts. 

System: This sentence is attributed to  --- "In the general domain, input-conflicting hallucinations occur when the answers generated by LLMs deviate from the original intentions of user inputs [ 19]."

USER: ok

ChatGPT Answer:  Another way is to introduce conflicting or incorrect contextual information in the prompt to confuse the model. 

System: This sentence is attributed to  --- "This hallucination may result from the model’s incorrect parsing or inaccurate internal representation of the input information, causing the output content to deviate from the intent and context of the source input."

USER: ok

ChatGPT Answer:  Introducing requirements that rely on knowledge missing from the model's training data can also lead to hallucinations. 

System: This sentence is attributed to  --- "3○We identify four potential factors that cause hallucinations: training data quality, intention understanding capacity, knowledge acquisition capacity, and repository-level context awareness."

USER: ok

ChatGPT Answer:  Complex contextual dependencies, such as those in repository-level code generation involving user-defined functions or non-code resources, can induce hallucinations due to the model’s limited context awareness. 

System: This sentence is attributed to  --- "In contrast, we investigate hallucinations in repository-level code generation within more practical development contexts, adopting a holistic perspective that encompasses phenomena ,mechanism , and mitigation ."

USER: ok

ChatGPT Answer:  Additionally, the hallucinations may arise when the model lacks proper intention understanding, leading to task requirement conflicts. 

System: This sentence is attributed to  --- "Most hallucinations associated with Task Requirement Conflicts andFactual Knowledge Conflicts can be, to a certain extent, attributed to data quality issues in the training corpora."

USER: ok

ChatGPT Answer:  Lastly, low-quality or insufficient training data related to the specific domain or task can cause the model to generate hallucinated content. 

DONE 

System: This sentence is attributed to  --- "This hallucination may result from the model’s incorrect parsing or inaccurate internal representation of the input information, causing the output content to deviate from the intent and context of the source input."

