disclosure.json

{
    "cap": "The models are adept at pulling the latest information from the internet to provide you with the most up-to-date answers possible. While designed to reduce errors, sometimes inaccuracies can slip through, particularly with very new or niche topics. Service providers frequently update their knowledge base to ensure information is fresh and reliable.",
    "time": "The models use real-time data from the web to respond to queries, ensuring information is current as of the moment it's provided. The data sources for the models are regularly updated, which helps maintain the accuracy and relevance of the information provided.",
    "source": "Responses are generated using high-quality, credible web snippets that are regularly vetted for accuracy. They are based on robust AI models (mistral-7b, llama2-70b), enhanced with Perplexity’s own technology to utilize web knowledge effectively.",
    "data": "Responses are fine-tuned with diverse, accurate data sets to optimize performance across helpfulness, factuality, and freshness. Regular updates and refinements help the models stay effective and reliable over time.",
    "mech": "The system integrates real-time internet data to enhance responses, making them relevant and timely. The models are continuously tested and improved based on feedback from realistic use cases and human evaluators to ensure quality and accuracy."
}