IP Library Granted Patent US 12682184
Granted Patent B2
US 12682184 · App. 18/828,735 · Granted Jul 14, 2026

Modular reasoning, knowledge, and language systems

Inventors: Yoav Shoham (Tel Aviv, IL); Barak Lenz (Tel Aviv, IL); Opher Lieber (Sde Warburg, IL); Yoav Levine (Tel Aviv, IL); Amnon Shashua (Tel Aviv, IL); Shai Shalev-Shwartz (Tel Aviv, IL); Kevin Leyton-Brown (Vancouver, CA); Ehud Karpas (Rehovot, IL); Erez Schwartz (Tel Aviv, IL); Noam Rozen (Tel Aviv, IL); Hofit Bata (Tel Aviv, IL); Gal Shachaf (Tel Aviv, IL); Dor Muhlgay (Tel Aviv, IL); Yoel Zeldes (Jerusalem, IL); Ori Ram (Tel Aviv, IL); Itay Dalmedigos (Tel Aviv, IL); Daniel Jannai-Epstein (Jerusalem, IL); Nir Ratner (Tel Aviv, IL)
Assignee: AI21 LABS
G06F40/40G06F40/103
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12682184
App. No.
18/828,735
Granted
Jul 14, 2026
Kind
B2
Abstract

The presently disclosed embodiments may include a computer readable medium including instructions that when executed by one or more processing devices cause the one or more processing devices to perform a method. The method may include receiving a user input, wherein the user input includes a natural language question, using a trained language model to decompose the natural language question into two or more information requests, routing each of the two or more information requests to at least one information resource, receiving two or more information responses from the at least one information resource, wherein the two or more information responses correspond to the two or more information requests, generating a natural language response to the user input based on the two or more information responses and providing the natural language response to the user.

Claims (28)

1 . A non-transitory computer-readable medium including instructions that when executed by one or more processing devices cause the one or more processing devices to perform a method including:

receiving a user input, wherein the user input includes a natural language question;

using at least one trained large language model to decompose the natural language question into one or more information requests;

identifying, using the at least one trained large language model, at least one information resource based on the one or more information requests;

formatting, using the at least one trained large language model, the one or more information requests for at least one input format associated with the at least one information resource;

routing each of the one or more information requests to the at least one information resource;

receiving one or more information responses from the at least one information resource, wherein the one or more information responses correspond to the one or more information requests;

generating a natural language response to the user input based on the one or more information responses; and

providing the natural language response to the user.

2 . The non-transitory computer-readable of claim 1 , wherein the at least one trained large language model is static.

3 . The non-transitory computer-readable of claim 1 , wherein the natural language response is provided to the user via a display.

4 . The non-transitory computer-readable of claim 1 , wherein the at least one information resource includes a calculator application.

5 . The non-transitory computer-readable of claim 1 , wherein the at least one information resource includes a search engine.

6 . The non-transitory computer-readable of claim 1 , wherein the at least one information resource includes one or more of a news outlet, a Wiki page or a website available via an Internet connection.

7 . The non-transitory computer-readable of claim 1 , wherein the at least one information resource includes an expert module including a model trained to generate a specialized result based on an expected class of model input information.

8 . The non-transitory computer-readable of claim 1 , wherein the at least one information resource includes a database.

9 . The non-transitory computer-readable of claim 1 , wherein the one or more information requests relate to information generated after a time at which training of the at least one trained large language model was completed.

10 . The non-transitory computer-readable of claim 1 , wherein the one or more information requests relate to current events information.

11 . The non-transitory computer-readable of claim 1 , wherein the one or more information requests relate to dynamically changing information.

12 . The non-transitory computer-readable of claim 11 , wherein the dynamically changing information includes at least one of weather information, currency exchange rates, stock prices, or news.

13 . The non-transitory computer-readable of claim 1 , wherein the method performed by the one or more processing devices includes: outputting, to the user, indicators associated with both the one or more information requests decomposed from the natural language question and the one or more information responses received from the at least one information resource.

14 . The non-transitory computer-readable of claim 1 , wherein the method performed by the one or more processing devices includes: outputting, to the user, a log of tasks associated with generating the natural language response to the user.

15 . The non-transitory computer-readable medium of claim 14 , wherein the log of tasks comprises a list of operations including decomposing the natural language question into the one or more information requests, receiving the one or more information responses from the at least one information resource, and generating the natural language response.

16 . The non-transitory computer-readable medium of claim 14 , wherein the log of tasks includes an explanation of how the natural language response was generated.

17 . The non-transitory computer-readable of claim 1 , wherein the natural language question included in the user input includes a multi-part question, and wherein each of the one or more information requests is associated with a different part of the multi-part question.

18 . The non-transitory computer-readable medium of claim 1 , wherein the one or more information requests relate to dynamically changing information.

19 . The non-transitory computer-readable medium of claim 1 , wherein the at least one information resource includes a trained machine learning model.

20 . The non-transitory computer-readable medium of claim 1 , wherein using the at least one trained large language model to identify the at least one information resource includes analyzing the natural language question.