US 12,353,457 B2
System and method for training a multi-tenant language model
Kfir Aharon Tishbi, Herzliya (IL); Raanan Raz, Tel Aviv (IL); and Amir Sheffer, Tel Aviv (IL)
Assigned to Avalor Technologies, Ltd., Tel Aviv (IL); and Zscaler, Inc., San Jose, CA (US)
Filed by Avalor Technologies, Ltd., Ramat Gan (IL)
Filed on Jun. 22, 2023, as Appl. No. 18/339,846.
Prior Publication US 2024/0427810 A1, Dec. 26, 2024
Int. Cl. G06F 16/334 (2025.01); G06F 16/34 (2025.01); G06F 16/901 (2019.01); G06F 40/284 (2020.01)
CPC G06F 16/3344 (2019.01) [G06F 16/34 (2019.01); G06F 40/284 (2020.01); G06F 16/9024 (2019.01)] 27 Claims
OG exemplary drawing
 
1. A method for reducing false responses from a large language model, comprising:
mapping a data field from a first source to a data field of a predefined semantic layer, the predefined semantic layer including a plurality of data fields;
storing data from the first source in a database based on the predefined semantic layer;
tokenizing each data field of the plurality of data fields for a first large language model (LLM);
fine-tuning the first LLM based on the tokenized predefined semantic layer;
providing a prompt to the first LLM, which configures the first LLM to generate an output answer;
providing the output answer to a second LLM, which configures the second LLM to generate a query for the database;
executing the query on the database to generate a database output based on the stored data;
providing the output answer in a user interface (UI) in response to determining that the database output and the output answer are within a predefined threshold; and
fine-tuning the first LLM further, in response to determining that the database output and the output answer are not within the predefined threshold.