LLMs: Internals, Hallucinations, and Applications | Data Brew | Episode 33
Data Brew by Databricks
00:00
Risks of Hallucinations in Language Models and Mitigation Strategies
This chapter explores the risks of hallucinations in language models and provides strategies to mitigate them, including adjusting the temperature of the ORM, guiding models with prompts, improving model architectures and data quality, conducting offline and online evaluations, enhancing prompt information, and leveraging the chain of thought process for reasoning improvement.
Transcript
Play full episode
