# US 12,762,169 B2 · Summary content generation and validation for parts on listing platforms > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,762,169 B2 - Document ID: US12762169B2 - Status: Granted - Kind: B2 - Application number: 18/962,924 - Filed: 2024-11-27 - Published: 2026-05-28 - Granted: 2026-10-13 - Earliest claimed: 2024-11-27 - Assignees: eBay Inc. - Applicants: eBay Inc. - Inventors: Sachin Joshi; Rajkumar Varadharajan - Web page: https://patentgrep.com/patent/US12762169B2 ## CPC - G06F 16/338: Presentation of query results - G06F 16/2228: {Indexing structures} - G06F 16/242: Query formulation - G06F 16/24578: {using ranking} - G06F 16/285: {Clustering or classification} - G06F 16/345: {Summarisation for human users} - G06N 3/045: Combinations of networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/0475: Generative networks - G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections - G06N 3/094: Adversarial learning - G06N 5/022: Knowledge engineering; Knowledge acquisition - G06N 20/00: Machine learning - G06Q 30/0641: {utilising user interfaces specially adapted for shopping} - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR ## Abstract Some aspects relate to technologies for generating and validating content providing summary information for parts on a listing platform, such as compatibility information and item characteristic information for the parts. In some aspects, input data comprising attribute-value pairs based on one or more items is accessed for content generation. The input data and one or more pre-defined generation prompts are provided to a first generative model, which generates content comprising natural language text as output. In some aspects, the content is validated by providing the content and one or more pre-defined validation prompts to a second generative model. If approved, the content can be provided for presentation on a user device. If rejected, the content is not provided for presentation. In further aspects, the rejected content is used by a third generative model for tuning the pre-defined generation prompts. ## Claims ### Claim 1 One or more computer storage media storing computer-useable instructions that, when used by one or more servers of a listing platform, cause the one or more servers to perform operations, the operations comprising: accessing, by at least one of the one or more servers of the listing platform, a first input data from an item listings data store of the listing platform, the first input data comprising a first plurality of attribute-value pairs for attributes of a first part item from a first part item listing on the listing platform, the first part item comprising a physical item having one or more compatibility requirements with one or more other physical items, the first plurality of attribute-value pairs for the attributes of the first part item providing information regarding the compatibility requirements; causing, by at least one of the one or more servers of the listing platform, a first generative model to generate a first content for the first part item listing by providing one or more pre-defined generation prompts and the first input data including the first plurality of attribute-value pairs to the first generative model, the first content comprising natural language text providing compatibility information for the first part item; causing, by at least one of the one or more servers of the listing platform, a second generative model to determine whether the first content is valid based on consistency of the natural language text of the first content with the compatibility requirements by providing one or more pre-defined validation prompts and the first content and the first plurality of attribute-value pairs to the second generative model; when a first output from the second generative model indicates the first content is valid, providing, by at least one of the one or more servers of the listing platform, the first content for presentation on a user device in association with the first part item listing; and when the first output from the second generative model indicates the first content is not valid, withholding the first content from presentation on the user device in association with the first part item listing. ### Claim 2 (depends on 1) The one or more computer storage media of claim 1, wherein the first plurality of attribute-value pairs comprises one or more attribute-value pairs for fitment data for the one or more part item listings and one or more attribute value pairs for item attribute data for the one or more part item listings. ### Claim 3 (depends on 2) The one or more computer storage media of claim 2, wherein the one or more attribute-value pairs for the fitment data and/or the one or more attribute-value pairs for the item attribute data are based on at least one selected from the following: search input, search results, browse node attributes, user data, and system knowledge. ### Claim 4 (depends on 1) The one or more computer storage media of claim 1, wherein the operations further comprise: accessing a second input data comprising a second plurality of attribute-value pairs for one or more second part item listings; causing the first generative model to generate a second content for the one or more second part item listings by providing the one or more pre-defined generation prompts and the second input data to the first generative model; causing the second generative model to validate the second content by providing the one or more pre-defined validation prompts and the second content to the second generative model; responsive to a second output from the second generative model rejecting the second content, causing a third generative model to generate a first tuned generation prompt by providing the second content and at least a portion of the second output from the second generative model to the third generative model; and adding the first tuned generation prompt to a repository of pre-defined generation prompts. ### Claim 5 (depends on 4) The one or more computer storage media of claim 4, wherein the first tuned generation prompt is generated for a first pre-defined generation prompt from the one or more pre-defined generation prompts; and wherein adding the first tuned generation prompt to the repository of pre-defined generation prompts comprises replacing the first pre-defined generation prompt in the repository of pre-defined generation prompts with the first tuned generation prompt. ### Claim 6 (depends on 4) The one or more computer storage media of claim 4, wherein the first generative model, the second generative model, and the third generative model comprise a same generative model. ### Claim 7 (depends on 4) The one or more computer storage media of claim 4, wherein the operations further comprise: accessing a third input data comprising a third plurality of attribute-value pairs; causing the first generative model to generate a third content by providing the one or more pre-defined generation prompts including the first tuned generation prompt and the third input data to the first generative model; causing the second generative model to validate the third content by providing the one or more pre-defined validation prompts and the third content to the second generative model; and responsive to a third output from the second generative model approving the third content, providing the third content for presentation on a user device. ### Claim 8 (depends on 7) The one or more computer storage media of claim 7, wherein the third input data comprises the second input data. ### Claim 9 (depends on 1) The one or more computer storage media of claim 1, wherein the operations further comprise: determining a first item category for the first input data; and selecting the one or more pre-defined generation prompts based on the first item category. ### Claim 10 (depends on 9) The one or more computer storage media of claim 9, wherein the operations further comprise: selecting the one or more pre-defined validation prompts based on the first item category. ### Claim 11 (depends on 10) The one or more computer storage media of claim 10, wherein the operations further comprise: accessing a second input data comprising a second plurality of attribute-value pairs for one or more second part item listings; determining a second item category for the second input data; selecting a second set of one or more pre-defined generation prompts based on the second item category; causing the first generative model to generate a second content for the one or more second part item listings by providing the second set of one or more pre-defined generation prompts and the second input data to the first generative model; selecting a second set of one or more pre-defined validation prompts based on the second item category; and causing the second generative model to validate the second content by providing the second set of one or more pre-defined validation prompts and the second content to the second generative model. ### Claim 12 A computer-implemented method comprising: accessing, by at least one of one or more servers of the listing platform, a first input data from an item listings data store of the listing platform, the first input data comprising a first plurality of attribute-value pairs for attributes of a first part item from a first part item listing on the listing platform, the first part item comprising a physical item having one or more compatibility requirements with one or more other physical items, the first plurality of attribute-value pairs for the attributes of the first part item providing information regarding the compatibility requirements; causing, by at least one of the one or more servers of the listing platform, a first generative model to generate a first content for the first part item listing by providing one or more pre-defined generation prompts and the first input data including the first plurality of attribute-value pairs to the first generative model, the first content comprising natural language text providing compatibility information for the first part item; causing, by at least one of the one or more servers of the listing platform, a second generative model to determine whether the first content is valid based on consistency of the natural language text of the first content with the compatibility requirements by providing one or more pre-defined validation prompts and the first content and the first plurality of attribute-value pairs to the second generative model; responsive to a first output from the second generative model rejecting the first content, withholding the first content from presentation on a user device in association with the first part item listing and causing, by at least one of the one or more servers of the listing platform, a third generative model to generate a first tuned generation prompt by providing to the third generative model one or more selected from the following: the first content, and at least a portion of the first output from the second generative model; and adding, by at least one of the one or more servers of the listing platform, the first tuned generation prompt to a repository of pre-defined generation prompts. ### Claim 13 (depends on 12) The computer-implemented method of claim 12, wherein the method further comprises: accessing a second input data comprising a second plurality of attribute-value pairs; causing the first generative model to generate a second content by providing the one or more pre-defined generation prompts including the first tuned generation prompt and the second input data to the first generative model; causing the second generative model to validate the second content by providing the one or more pre-defined validation prompts and the second content to the second generative model; and responsive to a second output from the second generative model approving the second content, providing the second content for presentation on a user device. ### Claim 14 (depends on 13) The computer-implemented method of claim 13, wherein the second input data comprises the first input data. ### Claim 15 (depends on 13) The computer-implemented method of claim 13, wherein the second input data is based on one or more second part item listings. ### Claim 16 (depends on 12) The computer-implemented method of claim 12, wherein the method further comprises: selecting the one or more pre-defined generation prompts based on one or more selected from the following: an item category of the one or more part item listings, a use case for which the first content is being generated, user data, and system knowledge. ### Claim 17 (depends on 12) The computer-implemented method of claim 12, wherein the method further comprises: selecting the one or more pre-defined validation prompts based on one or more selected from the following: the one or more pre-defined generation prompts, an item category of the one or more part item listings, a use case for which the first content is being generated, user data, and system knowledge. ### Claim 18 A listing platform having one or more servers comprising: a processor; and a computer storage medium storing computer-useable instructions that, when used by the processor, causes the listing platform to perform operations comprising: accessing, by at least one of the one or more servers of the listing platform, input data from an item listings data store of the listing platform, the input data comprising a plurality of attribute-value pairs for attributes of a first part item from a first part item listing on the listing platform, the plurality of attribute-value pairs comprising one or more attribute-value pairs for fitment data for the first part item listing and one or more attribute-value pairs for item attribute data for the first part item listing, the part item comprising a physical item having one or more compatibility requirements with one or more other physical items; selecting, from a repository of pre-defined generation prompts, one or more pre-defined generation prompts for the input data; causing, by at least one of the one or more servers of the listing platform, a first generative model to generate content by providing the one or more pre-defined generation prompts and the input data including the fitment data and the item attribute data to the first generative model, the content comprising natural language text providing summary information for the a first part item listing including compatibility information and item characteristic information; causing, by at least one of the servers of the listing platform, a second generative model to determine whether the content is valid based on consistency of the natural language text of the content with the fitment data and the item attribute data by providing one or more pre-defined validation prompts and the content and the plurality of attribute-value pairs to the second generative model; when a first output from the second generative model indicates the content is valid, providing, by at least one of the one or more servers of the listing platform, the content within a web page communicated to a user device for presentation; and when the first output from the second generative model indicates the first content is not valid, withholding the first content from presentation on the user device in association with the first part item listing. ### Claim 19 (depends on 18) The listing platform of claim 18, wherein the operations further comprises: causing a second generative model to validate the content by providing one or more pre-defined validation prompts and the content to the second generative model, wherein the content is provided within the web page communicated to the user device responsive to an output from the second generative model approving the content.