# US 2026/0187057 A1 · Multilinguistic Query Response Agent With Agent Core

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0187057 A1
- Document ID: US20260187057A1
- Status: Published application
- Kind: A1
- Application number: 19/004,830
- Filed: 2024-12-30
- Published: 2026-07-02
- Earliest claimed: 2024-12-30
- Assignees: Oracle International Corporation
- Applicants: Oracle International Corporation
- Inventors: Amit Agarwal; Tao Sheng; Hansa Meghwani
- Web page: https://patentgrep.com/patent/US20260187057A1

## CPC

- G06F 16/2452: Query translation
- G06F 16/3337: {Translation of the query language, e.g. Chinese to English}
- G06F 16/3329: Natural language query formulation
- G06F 16/90332: {Natural language query formulation or dialogue systems}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Techniques for responding to multilinguistic queries using an agent core are disclosed herein. An agent core is trained and/or fine-tuned in a first language to generate instructions (i.e., commands) for answering a query. A language identification and/or translation model receives queries, identifies languages associated with the queries, and translates the queries to the first language. The agent core generates instructions in the first language based on the translated queries. The instructions include instructions, in the first language, to perform actions such as retrieval, generation, contextual understanding, or calculation, in both the first language and the second language. The results of executing the instructions, including an action performed in the first language and an action performed in one or more second languages, are combined, ranked and/or reranked to generate an answer to the query.

## Claims

### Claim 1

A method, comprising:
accessing a first query;
responsive to determining that the first query is in a first language that is different than a target language:
generating a translated query by translating the first query from the first language to the target language;
generating a first command based on the translated query by providing the translated query to a first model configured for the target language and receiving the first command as output from the first model;
generating a first response in the first language using a second model by providing the first query and an input based on the first command to the second model; and
storing the first response;
wherein the method is performed by at least one device including a hardware processor.

### Claim 2 (depends on 1)

The method of claim 1, wherein:
the second model is a language model configured for the first language;
the method further comprising:
responsive to determining that the language of the first query is the target language:
generating a second command based on the first query by providing the first query to the first model;
generating a second response in the target language by providing the first query and an input based on the second command to a third model, the third model being a language model configured for the target language; and
storing the second response.

### Claim 3 (depends on 1)

The method of claim 1, wherein:
the first command comprises an identification of information including first language information in the first language and second language information in the target language, the method further comprising:
ranking the first language information and the second language information, and
generating the first response based on the ranking.

### Claim 4 (depends on 3)

The method of claim 3, wherein:
ranking the first language information and the second language information comprises:
ranking the first language information with a first ranking method configured for the first language; and
ranking the second language information with a second ranking method configured for the second language.

### Claim 5 (depends on 1)

The method of claim 1, wherein:
the first command comprises a command in the target language to access information from a data source; and
generating the first response comprises providing the first query and the information as context with the first query as input into to the second model.

### Claim 6 (depends on 1)

The method of claim 1, further comprising:
using a language detection model to determine the language of the first query.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
accessing an identification of a selected language, wherein the second model is configured for the selected language.

### Claim 8 (depends on 1)

The method of claim 1, further comprising:
determining a language of the first query using a specialized language determination model;
wherein said generating the translated query is performed using the language determination model; and
wherein the specialized language determination model executes operations on one or more central processing units.

### Claim 9 (depends on 8)

The method of claim 8, further comprising:
receiving feedback for the first response; and
fine-tuning the specialized language determination model using the feedback.

### Claim 10 (depends on 1)

The method of claim 1, wherein:
the first model is an agent model comprising a thought, action, observation framework.

### Claim 11 (depends on 1)

The method of claim 1, wherein:
the first model comprises a thought language model (LM), an action LM, and an observation LM;
the thought LM has a first number of parameters;
the action LM has a second number of parameters;
the observation LM has a third number of parameters; and
the second model has a fourth number of parameters that is greater than the first number, the second number, and the third number.

### Claim 12 (depends on 11)

The method of claim 11, wherein:
the thought LM has less than seven billion (7 B) parameters, the action LM has less than seven billion (7 B) parameters, and the observation LM has less than seven billion (7 B) parameters.

### Claim 13 (depends on 1)

The method of claim 1, wherein:
the second model is a language model configured to provide output in the first language;
prior to generating the translated query, the first model accesses a determination of whether to generate a response to the first query using (a) a language model configured to provide output in the target language or (b) a language model configured to provide output in the first language; and
the steps of generating the translated query, generating the first command, and generating the first response using the second model are performed responsive to determining to generate a response to the first query using a language model configured to provide output in the first language.

### Claim 14 (depends on 1)

The method of claim 1, wherein:
a number of parameters for the second model is at least one hundred billion (100 B) parameters.

### Claim 15 (depends on 1)

The method of claim 1, wherein:
the first model is a first agent model associated with the target language, the method comprising:
responsive to determining that the language of the first query is a second language different from the first language and the target language, providing the first query to a second agent model, the second agent model being configured for the second language.

### Claim 16 (depends on 1)

The method of claim 1, wherein:
the first model has at least one order of magnitude less parameters than the second model.

### Claim 17

One or more non-transitory computer readable media comprising:
instructions which, when executed by one or more hardware processors, cause performance of operations comprising:
accessing a first query;
responsive to determining that the first query is in a first language that is different than a target language:
generating a translated query by translating the first query from the first language to the target language;
generating a first command based on the translated query by providing the translated query to a first model configured for the target language and receiving the first command as output from the first model;
generating a first response in the first language using a second model by providing the first query and an input based on the first command to the second model; and
storing the first response.

### Claim 18 (depends on 17)

The computer readable media of claim 17, wherein:
the second model is a language model configured for the first language;
the method further comprising:
responsive to determining that the language of the first query is the target language:
generating a second command based on the first query by providing the first query to the first model;
generating a second response in the target language by providing the first query and an input based on the second command to a third model, the third model being a language model configured for the target language; and
storing the second response.

### Claim 19

A system comprising:
at least one device including a hardware processor;
the system being configured to perform operations comprising:
accessing a first query;
responsive to determining that the first query is in a first language that is different than a target language:
generating a translated query by translating the first query from the first language to the target language;
generating a first command based on the translated query by providing the translated query to a first model configured for the target language and receiving the first command as output from the first model;
generating a first response in the first language using a second model by providing the first query and an input based on the first command to the second model; and
storing the first response.

### Claim 20 (depends on 19)

The system of claim 19, wherein:
the second model is a language model configured for the first language;
the method further comprising:
responsive to determining that the language of the first query is the target language:
generating a second command based on the first query by providing the first query to the first model;
generating a second response in the target language by providing the first query and an input based on the second command to a third model, the third model being a language model configured for the target language; and
storing the second response.
