# US 2026/0278304 A1 · Generating summaries of texts using large language models selected based on metrics associated with a classification score, processing and reading times and words in a deny list

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

- Number: US 2026/0278304 A1
- Document ID: US20260278304A1
- Status: Published application
- Kind: A1
- Application number: 19/673,453
- Filed: 2026-05-11
- Published: 2026-09-17
- Earliest claimed: 2023-11-29
- Assignees: Sorcero, Inc.
- Applicants: Sorcero, Inc.
- Inventors: Walter Bender; Nithi Vivatrat; Richard Graves; Tomáš Valena; David McMinn
- Web page: https://patentgrep.com/patent/US20260278304A1

## CPC

- G06F 16/345: {Summarisation for human users}
- G06F 40/42: Data-driven translation
- G06F 18/241: relating to the classification model, e.g. parametric or non-parametric approaches
- G06F 3/0484: for the control of specific functions or operations, e.g. selecting or manipulating an object, an image or a displayed text element, setting a parameter value or selecting a range
- G06F 16/332: Query formulation
- G06F 16/34: Browsing; Visualisation therefor (browsing or visualisation for clustering or classification G06F16/358)
- G06F 40/157: using dictionaries or tables
- G06F 40/205: Parsing
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 5/04: Inference or reasoning models
- G06Q 10/103: {Workflow collaboration or project management}
- 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

Systems, methods, and computer-readable storage media. A system can include one or more processors to identify a plurality of sections of an article, determine metrics for a plurality of generative models, select one or more generative models of the plurality of generative models, provide a prompt including instructions to generate a section summary for the section, determine a concept found in the article summary that is missing from the article, and provide an output including the article summary and an indicator corresponding to the concept.

## Claims

### Claim 1

A system, comprising:
one or more processors coupled to memory, the one or more processors configured to:
identify a plurality of sections of an article;
determine metrics for a plurality of generative models, the metrics comprising at least two of a processing time, a classification score, a reading score associated with a reading duration, a reading time, or a number of words included on a deny list;
select one or more generative models of the plurality of generative models based on at least one of (i) a comparison of the metrics of the selected one or more generative models to the metrics of non-selected generative models of the plurality of generative models, or (ii) a comparison of the metrics of the selected one or more generative models to one or more thresholds;
provide, to the selected one or more generative models, for each section of the plurality of sections, a prompt including instructions to generate a section summary for the section;
determine, from an article summary generated based on the section summary for each section, a concept found in the article summary that is missing from the article; and
provide, for presentation at a client device, an output including the article summary and an indicator corresponding to the concept.

### Claim 2 (depends on 1)

The system of claim 1, wherein the metrics of the selected one or more generative models indicate at least two of a lowest processing time, a highest classification score, a lowest score associated with the reading duration, a lowest reading time, or a lowest number of the words included on the deny list.

### Claim 3 (depends on 1)

The system of claim 1, wherein the one or more processors further:
detect, using at least one generative model, one or more hallucinations in the article summary;
tag one or more segments of the article summary with labels configured to simulate feedback associated with the one or more hallucinations; and
provide, to the selected one or more generative models, an updated prompt to cause the selected one or more generative models to regenerate at least one section summary in accordance with the simulated feedback.

### Claim 4 (depends on 1)

The system of claim 1, wherein the one or more processors further:
identify a target audience associated with the article summary;
generate, based on the target audience, an updated prompt to cause the one or more generative models to adjust a complexity of the article summary; and
provide the updated prompt to the one or more generative models.

### Claim 5 (depends on 1)

The system of claim 1, wherein the metrics comprise at least one of:
a minimization of the processing time;
a minimization of the reading score associated with the reading duration;
a minimization of the reading time; and
a minimization of the number of the words included on the deny list.

### Claim 6 (depends on 1)

The system of claim 1, wherein the one or more processors further:
provide, during a plurality of iterations, a plurality of prompts to cause the selected one or more generative models to generate a plurality of section summaries for each section, wherein the selected one or more generative models perform a plurality of processing stages during each of the plurality of iterations;
compare at least one section summary of the plurality of section summaries to one or more predetermined thresholds;
determine, based on the comparison of the at least one section summary to the one or more predetermined thresholds, to incorporate an optimization during an iteration of the plurality of iterations; and
augment at least one of the plurality of prompts to cause the selected one or more generative models to implement the optimization by bypassing at least one of the plurality of processing stages during the iteration.

### Claim 7 (depends on 1)

The system of claim 1, wherein the one or more processors further:
determine a source persona associated with the article and a target persona associated with the article summary;
provide data corresponding to the source persona to the selected one or more generative models via the prompt; and
refine the article summary by updating sentences included in the at least one section summary based on the target persona.

### Claim 8 (depends on 1)

The system of claim 1, wherein the one or more processors further:
extract, from the article, a document structure comprising the plurality of sections;
determine, based on the document structure, a mapping between the plurality of sections of the article and each section summary generated for the article summary; and
verify that one or more conclusions in at least one section summary of the article summary are supported by a corresponding section of the plurality of sections of the article based on the mapping.

### Claim 9 (depends on 1)

The system of claim 1, wherein the one or more processors further:
determine a first tone for a first section summary and a second tone for a second section summary, wherein the first tone is different from the second tone; and
provide, to the selected one or more generative models, a first prompt to cause the one or more generative models to generate the first section summary in the first tone and a second prompt to cause the one or more generative models to generate the second section summary in the second tone.

### Claim 10 (depends on 1)

The system of claim 1, wherein the one or more processors further:
transmit, to the client device, data to cause the client device to render a graphical diagram associated with the article summary for display on a graphical user interface;
receive, from the client device, one or more user preferences; and
transmit, to the client device, data to cause the client device to dynamically update the graphical diagram in accordance with the one or more user preferences.

### Claim 11 (depends on 1)

The system of claim 1, wherein at least one of the metrics corresponds to a reading level.

### Claim 12 (depends on 1)

The system of claim 1, wherein the one or more processors further:
generate an article summary based on the section summary generated by the selected one or more generative models for each section;
determine, using a classifier, for a first sentence included in the article summary, a confidence score indicating a likelihood that the first sentence is not supported by the article; and
provide, for presentation at the client device, a second indicator corresponding to the first sentence.

### Claim 13 (depends on 1)

The system of claim 1, wherein the one or more processors further:
interface with at least one external application via a network;
determine an output format for the output, the output format associated with the at least one external application; and
provide, to the selected one or more generative models, the prompt comprising information associated with the output format.

### Claim 14 (depends on 13)

The system of claim 13, wherein the output format comprises at least one of a video format, a clinical trial format, or a study format, and wherein the one or more processors generate at least one of a dictionary or guide for presentation via the client device in association with the output.

### Claim 15 (depends on 1)

The system of claim 1, wherein the one or more processors further:
access the plurality of generative models using a remote server;
provide, for display on an interface of the client device, the plurality of generative models for selection;
receive, from the client device via the interface, an input associated with the plurality of generative models; and
select the one or more generative models based at least in part on the input and the metrics.

### Claim 16

A method, comprising:
identifying, by one or more processors coupled to memory, a plurality of sections of an article;
determining, by the one or more processors, metrics for a plurality of generative models, the metrics comprising at least two of a processing time, a classification score, a reading score associated with a reading duration, a reading time, or a number of words included on a deny list;
selecting, by the one or more processors, one or more generative models of the plurality of generative models based on at least one of (i) a comparison of the metrics of the selected one or more generative models to the metrics of non-selected generative models of the plurality of generative models, or (ii) a comparison of the metrics of the selected one or more generative models to one or more thresholds;
providing, by the one or more processors, to the selected one or more generative models, for each section of the plurality of sections, a prompt including instructions to generate a section summary for the section;
determining, by the one or more processors, from an article summary generated based on the section summary for each section, a concept found in the article summary that is missing from the article; and
provide, for presentation at a client device, an output including the article summary and an indicator corresponding to the concept.

### Claim 17 (depends on 16)

The method of claim 16, wherein the metrics of the selected one or more generative models indicate at least two of a lowest processing time, a highest classification score, a lowest score associated with the reading duration, a lowest reading time, or a lowest number of the words included on the deny list.

### Claim 18 (depends on 16)

The method of claim 16, further comprising:
detecting, by the one or more processors, using at least one generative model, one or more hallucinations in the article summary;
tagging, by the one or more processors, one or more segments of the article summary with labels configured to simulate feedback associated with the one or more hallucinations; and
providing, by the one or more processors, to the selected one or more generative models, an updated prompt to cause the selected one or more generative models to regenerate at least one section summary in accordance with the simulated feedback.

### Claim 19 (depends on 16)

The method of claim 16, further comprising:
identifying, by the one or more processors, a target audience associated with the article summary;
generating, by the one or more processors, based on the target audience, an updated prompt to cause the one or more generative models to adjust a complexity of the article summary; and
providing, by the one or more processors, the updated prompt to the one or more generative models.

### Claim 20

A non-transitory computer-readable medium (CRM) having one or more instructions stored thereon, the one or more instructions executable by one or more processors to:
identify a plurality of sections of an article;
determine metrics for a plurality of generative models, the metrics comprising at least two of a processing time, a classification score, a reading score associated with a reading duration, a reading time, or a number of words included on a deny list;
select one or more generative models of the plurality of generative models based on at least one of (i) a comparison of the metrics of the selected one or more generative models to the metrics of non-selected generative models of the plurality of generative models, or (ii) a comparison of the metrics of the selected one or more generative models to one or more thresholds;
provide, to the selected one or more generative models, for each section of the plurality of sections, a prompt including instructions to generate a section summary for the section;
determine, from an article summary generated based on the section summary for each section, a concept found in the article summary that is missing from the article; and
provide, for presentation at a client device, an output including the article summary and an indicator corresponding to the concept.
