[Note] Parameters to assess ChatBot products ?

context window:
- How much information can you feed into the chatbot at once?
- This is typically measured in words or tokens.
context remainment:
- How many back-and-forth exchanges can occur before the chatbot begins to forget earlier parts of the conversation?
- What’s the maximum conversation length it can effectively maintain?

hallucination:
- How accurate is the chatbot’s information?
- Can it resist confirming biased or incorrect assumptions in questions?
- Does it generate false or misleading information while presenting it as fact?
internet access :
- Can the bot retrieve real-time information from external sources?
- Does it automatically access websites, academic papers, or other online resources when needed ?
source transparency :
- Does the bot cite its sources?
- Is it clear about where its information comes from?
- Does it explain the basis for its responses?

writting tone:
- How natural and human-like is the bot’s communication?
- Can it adjust its tone to match different situations?
multi-model:
- Can the bot process different types of input beyond text?
- Does it handle documents (PDF, DOCX), images, audio, and video?
- Can it effectively use information from these various formats in its responses?
response freshness:
- How consistent are the bot’s responses to identical questions?
- Does it offer helpful variations in its answers?
legal liability:
- Does the bot recognize and refuse inappropriate requests?
- How does it handle sensitive topics (political, legal, or ethical issues)?
- Are there clear guidelines about what it won’t discuss?

community:
- Is there an active community developing extensions or plugins?
- Are there readily available tools to enhance the bot’s capabilities using initial prompting? and can the bot be customized for specific professional needs (like writing, teaching, or specialized fields) ?

Reference