Anti-patterns while working with LLMs
13 days ago
- #AI Security
- #Context Management
- #LLM Best Practices
- Avoid sending redundant information in the same LLM session to optimize context usage.
- Use LLMs for tasks they are good at, like coding, rather than tasks they struggle with, such as counting letters.
- Be cautious with long sessions exceeding token limits, as LLMs may forget or misremember critical information.
- Expect lower accuracy from LLMs on obscure or post-training-cutoff topics.
- Stay vigilant with LLM-generated code to prevent security risks like unnecessary data exposure.