# US 2026/0073211 A1 · Altering manifolds for generative modeling

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

- Number: US 2026/0073211 A1
- Document ID: US20260073211A1
- Status: Published application
- Kind: A1
- Application number: 18/826,664
- Filed: 2024-09-06
- Published: 2026-03-12
- Earliest claimed: 2024-09-06
- Assignees: Microsoft Technology Licensing, LLC
- Applicants: Microsoft Technology Licensing, LLC
- Inventors: Harsh SHRIVASTAVA
- Web page: https://patentgrep.com/patent/US20260073211A1

## CPC

- G06N 3/08: Learning methods
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The present disclosure relates to systems and methods for modifying an output of a generative artificial intelligence (AI) model. The systems and methods control the loss function for samples under consideration by adjusting a value of the loss function for the samples and decreasing a probability that the generative AI model uses the samples in generating responses.

## Claims

### Claim 1

A method comprising:
learning, by a neural network, an initial manifold over a dataset in response to the neural network performing manifold learning using a first loss function over the dataset;
generating, by the neural network, samples in response to the neural network performing manifold sampling over the initial manifold;
identifying a sample from the samples;
including the sample in a second loss function;
learning, by the neural network, a second manifold over the dataset in response to the neural network performing manifold learning using the second loss function over the dataset; and
generating, using the neural network, new samples in response to the neural network performing the manifold sampling over the second manifold.

### Claim 2 (depends on 1)

The method of claim 1, wherein the initial manifold is a function of the first loss function versus the dataset and the second manifold is a function of the second loss function versus the dataset.

### Claim 3 (depends on 1)

The method of claim 1, wherein the sample is an unwanted output of the neural network.

### Claim 4 (depends on 1)

The method of claim 1, wherein the second loss function increases a loss for the sample decreasing a probability that the neural network generates the new samples using the sample.

### Claim 5 (depends on 1)

The method of claim 1, wherein the second manifold increases a trench at an input data point in the dataset on the second manifold increasing a probability that the neural network generates the new samples from a region surrounding the trench, and
wherein the second loss function increases a loss of the sample and removes the sample from the region surrounding the trench in the second manifold.

### Claim 6 (depends on 5)

The method of claim 5, wherein generating the new samples occurs in the region surrounding trench at the input data point.

### Claim 7 (depends on 1)

The method of claim 1, wherein a plurality of input data points are identified in the dataset and the second manifold increases trenches in the second manifold, where each trench corresponds to an input data point of the plurality of data points.

### Claim 8 (depends on 7)

The method of claim 7, wherein the trenches increase a probability that the neural network generates the new samples from regions surrounding the trenches.

### Claim 9 (depends on 7)

The method of claim 7, wherein the second loss function increases a loss of the sample and removes the sample from regions surrounding the trenches, reducing a probability that the neural network generates the new samples from the sample.

### Claim 10 (depends on 1)

The method of claim 1, wherein the sample is in a region a distance from samples coming from an underlying distribution of input data.

### Claim 11 (depends on 10)

The method of claim 10, wherein as the distance increases from input data, samples generated by the neural network are out of distribution samples.

### Claim 12 (depends on 10)

The method of claim 10, wherein the new samples are closer in distance to the input data as compared to the sample identified in the region.

### Claim 13 (depends on 1)

The method of claim 1, further comprising:
identifying a set of samples from the samples; and
including the set of samples in the second loss function, wherein the second loss function increases the loss for each sample of the set of samples reducing a probability that the neural network generates the new samples from the sample.

### Claim 14 (depends on 1)

The method of claim 1, wherein the manifold is a function of the first loss function versus the dataset.

### Claim 15 (depends on 1)

The method of claim 1, wherein the neural network learns a probability distribution over the dataset.

### Claim 16 (depends on 1)

The method of claim 1, wherein the manifold learning and the manifold sampling occurs in dimensions higher than one dimension.

### Claim 17

A device comprising:
a memory to store data and instructions; and
a processor operable to communicate with the memory, wherein the processor is operable to:
learn, by a neural network, an initial manifold over a dataset in response to the neural network performing, using a first loss function, manifold learning over a dataset;
generate, by the neural network, samples in response to the neural network performing manifold sampling over the initial manifold;
identify a sample from the samples;
include the sample in a second loss function;
learn, by the neural network, a second manifold over the dataset in response to the neural network performing, using the second loss function, the manifold learning over the dataset; and
generate, using the neural network, new samples in response to the neural network performing the manifold sampling over the second manifold.

### Claim 18 (depends on 17)

The device of claim 17, wherein the second manifold increases a trench at an input data point in the dataset on the second manifold increasing a probability that the neural network generates the new samples from a region surrounding the trench, and
wherein the second loss function increases a loss for the sample removing the sample from the trench and decreasing the probability that the neural network generates the new samples using the sample.

### Claim 19 (depends on 17)

The device of claim 17, wherein the processor is further operable to:
identify a set of samples from the samples; and
include the set of samples in the second loss function, wherein the second loss function increases the loss for each sample of the set of samples reducing a probability that the neural network generates the new samples from the set of samples.

### Claim 20 (depends on 19)

The device of claim 19, wherein a plurality of input data points are identified in the dataset and the second manifold increases trenches in the second manifold, where each trench corresponds to an input data point of the plurality of data points, and
wherein increasing the loss for each sample of the set of samples removes the set of samples from the trenches in the second manifold reducing the probability that the neural network generates the new samples from the set of samples.
