Search
Most LLMs are trained once and not updated, lacking the ability to dynamically adapt to ever-changing world. These models often hallucinate and provide factually incorrect information, reducing trustworthiness of their responses, especially in settings where accurate and up-to-date information is critical.
SUTRA can use, infer, and process real-time knowledge from the internet and leverage it to provide the most up-to-date information when forming responses. SUTRA-ONLINE
can accurately respond to time-sensitive queries, extending its knowledge beyond a static training corpus.
The following example shows how SUTRA-ONLINE
responds to queries with up-to-date information, without suffering from knowledge cut-off dates from pretraining training data.
SUTRA-ONLINE
response for “Who won the India vs England cricket match from yesterday?” (as of 02/26/2024)
Along with the above response, SUTRA-ONLINE
also provided following 5 sources, so that user can get further information and have better transparency on how the model arrived
GPT-3.5 response for “Who won the India vs England cricket match from yesterday?” (as of 02/20/2024)
The provided Times of India is factually incorrect and refers to a match from Oct 29, 2023.
Llama-70B-chat
response for “Who won the India vs England cricket match from yesterday?” (as of 02/20/2024)
Mistral
response for “Who won the India vs England cricket match from yesterday?” (as of 02/20/2024)