# US 2025/0156766 A1 · Systems and methods for soft model assertions

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

- Number: US 2025/0156766 A1
- Document ID: US20250156766A1
- Status: Published application
- Kind: A1
- Application number: 19/025,952
- Filed: 2025-01-16
- Published: 2025-05-15
- Earliest claimed: 2021-05-10
- Assignees: Toyota Research Institute, Inc.
- Applicants: Toyota Research Institute, Inc.
- Inventors: DANIEL D. KANG; NIKOS ARECHIGA GONZALEZ; Sudeep Pillai
- Web page: https://patentgrep.com/patent/US20250156766A1

## CPC

- G06F 11/3013: {where the computing system is an embedded system, i.e. a combination of hardware and software dedicated to perform a certain function in mobile devices, printers, automotive or aircraft systems (testing or monitoring of control systems or parts thereof G05B23/02)}
- G06N 20/00: Machine learning
- B60W 60/001: {Planning or execution of driving tasks}
- G06F 11/0706: {the processing taking place on a specific hardware platform or in a specific software environment}
- G06F 11/0751: {Error or fault detection not based on redundancy (power supply failures G06F1/30; network fault management H04L41/06)}
- G06F 11/0793: {Remedial or corrective actions (recovery from an exception in an instruction pipeline G06F9/3861; by retry G06F11/1402; for recovering from a failure of a protocol instance or entity H04L69/40)}
- G06F 11/302: {where the computing system component is a software system}
- G06F 11/3089: {Monitoring arrangements determined by the means or processing involved in sensing the monitored data, e.g. interfaces, connectors, sensors, probes, agents (software debugging using additional hardware using a specific debug interface G06F11/3656; performance evaluation by tracing or monitoring G06F11/3466)}
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- B60W: CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT

## Abstract

Systems and methods are provided for implementing soft model assertions (SMA) system and techniques designed to monitor and improve Machine Learning (ML) model quality by to detecting errors within the one or more ML models. SMA techniques and systems are distinctly designed to leverage: 1) a user's ability to specify features over data; and 2) large, existing datasets of organizations, in a manner that can improve the accuracy and quality of predicting potential errors in Machine Learning (ML) models. A SMA system can include a controller device receiving predictions generated based on the ML models and output from the SMA system. The controller performs autonomous operations of the system in response to determining that the one or more detected errors within the one or more ML models yield a high certainty of errors in the predictions. The SMA system also includes a domain specific language and a severity score module.

## Claims

### Claim 1

A method comprising:
generating one or more priors based on observations over predictions from one or more machine learning (ML) models;
generating one or more application objective functions;
generating a severity score over the one or more priors and the one or more application objective functions; and
detecting one or more errors within the one or more ML models based on the severity score.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
executing an evasive maneuver of a vehicle that corresponds to the one or more errors.

### Claim 3 (depends on 1)

The method of claim 1, wherein the observations are associated within a given time step in bundles or associated across time in tracks.

### Claim 4 (depends on 1)

The method of claim 1, wherein generating the one or more priors comprises:
analyzing the observations; and
based on the analyzing of the observations, outputting a probability of a transformation of an input, wherein the probability comprises an application-specific probability.

### Claim 5 (depends on 4)

The method of claim 4, wherein generating the one or more application objective functions comprises:
automatically fitting the one or more priors over a training data set based on specifications of the transformation, wherein the training data set is associated with the one or more ML models.

### Claim 6 (depends on 4)

The method of claim 4, wherein generating the severity score comprises:
calculating an application-specific score based on the application-specific probability; and
assigning the application-specific score as the severity score.

### Claim 7 (depends on 1)

The method of claim 1, further comprising generating one or more data associations based on the observations over predictions from the one or more ML models.

### Claim 8 (depends on 7)

The method of claim 7, wherein generating the one or more priors comprises generating the one or more priors based on the one or more data associations.

### Claim 9 (depends on 7)

The method of claim 7, wherein the one or more data associations, the one or more priors, and the one or more application objective functions are components of a probabilistic domain-specific language (DSL) for a soft model assertion (SMA) system.

### Claim 10

A vehicle comprising:
one or more processors configured to execute machine-readable instructions to cause the vehicle to:
generate one or more priors based on observations over predictions from one or more machine learning (ML) models;
generate one or more application objective functions;
generate a severity score over the one or more priors and the one or more application objective functions;
detect one or more errors within the one or more ML models based on the severity score; and
execute an evasive maneuver for the vehicle that corresponds to the one or more errors.

### Claim 11 (depends on 10)

The vehicle of claim 10, wherein the observations are associated within a given time step in bundles or associated across time in tracks.

### Claim 12 (depends on 10)

The vehicle of claim 10, wherein generating the one or more priors comprises:
analyzing the observations; and
based on the analyzing of the observations, outputting a probability of a transformation of an input, wherein the probability comprises an application-specific probability.

### Claim 13 (depends on 12)

The vehicle of claim 12, wherein generating the one or more application objective functions comprises:
automatically fitting the one or more priors over a training data set based on specifications of the transformation, wherein the training data set is associated with the one or more ML models.

### Claim 14 (depends on 12)

The vehicle of claim 12, wherein generating the severity score comprises:
calculating an application-specific score based on the application-specific probability; and
assigning the application-specific score as the severity score.

### Claim 15 (depends on 10)

The vehicle of claim 10, wherein the one or more processors are further configured to execute machine-readable instructions to generate one or more data associations based on the observations over predictions from the one or more ML models.

### Claim 16 (depends on 15)

The vehicle of claim 15, wherein generating the one or more priors comprises generating the one or more priors based on the one or more data associations.

### Claim 17 (depends on 15)

The vehicle of claim 15, wherein the one or more data associations, the one or more priors, and the one or more application objective functions are components of a probabilistic domain-specific language (DSL) for a soft model assertion (SMA) system.

### Claim 18

A soft model assertion (SMA) system comprising:
one or more processors configured to execute machine-readable instructions to cause the SMA system to:
generate one or more priors based on observations over predictions from one or more machine learning (ML) models;
generate one or more application objective functions;
generate a severity score over the one or more priors and the one or more application objective functions; and
detect one or more errors within the one or more ML models based on the severity score.

### Claim 19 (depends on 18)

The SMA system of claim 18, wherein generating the one or more priors comprises:
analyzing the observations; and
based on the analyzing of the observations, outputting a probability of a transformation of an input, wherein the probability comprises an application-specific probability.

### Claim 20 (depends on 19)

The SMA system of claim 19, wherein generating the one or more application objective functions comprises:
automatically fitting the one or more priors over a training data set based on specifications of the transformation, wherein the training data set is associated with the one or more ML models.
