System and method for tailoring prompts for generative models

Patent metadata

Field	Value
Publication number	US-2025335776-A1
Application number	US-202418646546-A
Country	US
Kind code	A1
Filing date	Apr 25, 2024
Priority date	Apr 25, 2024
Publication date	Oct 30, 2025
Grant date	—

Abstract

A method for modifying prompts includes generating, via large language model, a first group of prompts based on receiving a first user prompt from a first user. The method also includes receiving, from the first user, a first input selecting a first selected prompt of the first group of prompts. The method further includes generating, via a first generative model, a first output based on the first user selecting the first selected prompt. The method still further includes receiving, from a second user, a first rating associated with the first output.

First claim

What is claimed is:

1 . A method for modifying prompts, comprising:
generating, via large language model, a first group of prompts based on receiving a first user prompt from a first user; receiving, from the first user, a first input selecting a first selected prompt of the first group of prompts; generating, via a first generative model, a first output based on the first user selecting the first selected prompt; and receiving, from a second user, a first rating associated with the first output.

2 . The method of claim 1 , further comprising:
identifying a subset of stored prompts from a set of stored prompts based on receiving a second user prompt from a third user, each stored prompt of the subset of stored prompts associated with a rating; generating a second group of prompts based on the subset of stored prompts and the second user prompt; receiving, from the third user, a second input selecting a second selected prompt of the second group of prompts; generating, via a second generative model, a second output based receiving the second input selecting the second selected prompt; and receiving, from a fourth user, a second rating associated with the second output.

3 . The method of claim 2 , wherein the subset of stored prompts are identified based on an embedding of the second user prompt.

4 . The method of claim 2 , wherein the subset of stored prompts identified based on the respective rating of each stored prompt in the set of stored prompts.

5 . The method of claim 2 , wherein the subset of stored prompts is identified based on a quantity of stored prompts in the set of stored prompts being greater than a stored prompt threshold.

6 . The method of claim 2 , wherein the first user is the same user as the third user and/or the second user is the same user as the fourth user.

7 . The method of claim 1 , wherein:
the large language model is trained to generate the first group of prompts; and the first group of prompts is generated in response to a second prompt received at the large language model.

8 . An apparatus for modifying prompts, comprising:
one or more processors; and one or more memories coupled with the one or more processors and storing processor-executable code that, when executed by the one or more processors, is configured to cause the apparatus to:
generate, via large language model, a first group of prompts based on receiving a first user prompt from a first user;
receive, from the first user, a first input selecting a first selected prompt of the first group of prompts;
generate, via a first generative model, a first output based on the first user selecting the first selected prompt; and
receive, from a second user, a first rating associated with the first output.

9 . The apparatus of claim 8 , wherein execution of the processor-executable code further causes the apparatus to:
identify a subset of stored prompts from a set of stored prompts based on receiving a second user prompt from a third user, each stored prompt of the subset of stored prompts associated with a rating; generate a second group of prompts based on the subset of stored prompts and the second user prompt; receive, from the third user, a second input selecting a second selected prompt of the second group of prompts; generate, via a second generative model, a second output based receiving the second input selecting the second selected prompt; and receive, from a fourth user, a second rating associated with the second output.

10 . The apparatus of claim 9 , wherein the subset of stored prompts are identified based on an embedding of the second user prompt.

11 . The apparatus of claim 9 , wherein the subset of stored prompts identified based on the respective rating of each stored prompt in the set of stored prompts.

12 . The apparatus of claim 9 , wherein the subset of stored prompts is identified based on a quantity of stored prompts in the set of stored prompts being greater than a stored prompt threshold.

13 . The apparatus of claim 9 , wherein the first user is the same user as the third user and/or the second user is the same user as the fourth user.

14 . The apparatus of claim 8 , wherein:
the large language model is trained to generate the first group of prompts; and the first group of prompts is generated in response to a second prompt received at the large language model.

15 . A non-transitory computer-readable medium having program code recorded thereon for modifying prompts, the program code executed by one or more processors and comprising:
program code to generate, via large language model, a first group of prompts based on receiving a first user prompt from a first user; program code to receive, from the first user, a first input selecting a first selected prompt of the first group of prompts; program code to generate, via a first generative model, a first output based on the first user selecting the first selected prompt; and program code to receive, from a second user, a first rating associated with the first output.

16 . The non-transitory computer-readable medium of claim 15 , wherein the program code further comprises:
program code to identify a subset of stored prompts from a set of stored prompts based on receiving a second user prompt from a third user, each stored prompt of the subset of stored prompts associated with a rating; program code to generate a second group of prompts based on the subset of stored prompts and the second user prompt; program code to receive, from the third user, a second input selecting a second selected prompt of the second group of prompts; program code to generate, via a second generative model, a second output based receiving the second input selecting the second selected prompt; and program code to receive, from a fourth user, a second rating associated with the second output.

17 . The non-transitory computer-readable medium of claim 16 , wherein the subset of stored prompts are identified based on an embedding of the second user prompt.

18 . The non-transitory computer-readable medium of claim 16 , wherein the subset of stored prompts identified based on the respective rating of each stored prompt in the set of stored prompts.

19 . The non-transitory computer-readable medium of claim 16 , wherein the subset of stored prompts is identified based on a quantity of stored prompts in the set of stored prompts being greater than a stored prompt threshold.

20 . The non-transitory computer-readable medium of claim 16 , wherein the first user is the same user as the third user and/or the second user is the same user as the fourth user.

Assignees

• Toyota Res Inst Inc
• Toyota Motor Co Ltd

Inventors

• Chen Francine R
• Carter Scott
• Filipowicz Alexandre Leo Stephen
• Huynh Brandon

Classifications

• G06N3/091 Primary

  Active learning · CPC title